Newegg has refreshed its bundle deals and have recently come out with a couple of solid deals that help take some of the sting out of building a PC in these trying times. The latest combo pairs one of the fastest gaming CPUs in the Ryzen 5 9800X3D, 32GB of Corsair DDR5-6000 RAM, and a premium motherboard with MSI’s X870E Carbon Wifi for only $1,250.98, saving $119 compared to buying the parts individually. In fact, buying this RAM kit alone would cost $509.99, so if you apply the discount to the RAM, it’s just $390 and cheaper than any comparable standalone kit, which looks to be around $460 at the time of this writing. Newegg also throws in a free Rosewill 360mm AIO ($90 value) to sweeten the pot.

• Check out all the RAM Combo deals at Newegg

The RAM portion of the bundle, Corsair’s Vengeance RGB 32GB (2x16GB) DDR5-6000 CL36 RAM kit, is currently available on Newegg for an eye-watering $509.99 when bought separately. The good news is that the discount brings this RAM down to something reasonable for the times. This black kit matches the MSI X870E Carbon board in the combo and features a frosted RGB light bar that runs the length of the sticks. If RGB lighting isn’t your thing and you prefer a stealthy, unassuming look, you can turn it off. The DDR5-6000 speed lands in the sweet spot for the AM5 platform and is a worthwhile, good-looking option. The SK Hynix ICs under the heatsink pair nicely with the MSI board and offer plenty of overclocking headroom.

Next is one of the most popular and fastest current-generation gaming CPUs: the Ryzen 7 9800X3D. This 8-core/16-thread Zen5-based processor has a 4.6 GHz base clock and a 5.2 GHz max boost, plenty fast for any task. The X3D variant's 96MB of L3 cache helps with gaming (among other things), making it one of the fastest gaming CPUs available. It is also a well-rounded performer that excels in games and suits most applications.

Last but certainly not least is the premium MSI X870E Carbon Wifi motherboard (one we reviewed in the past. We loved the robust power delivery (perfect for the 9800X3D or greater), the dual Ethernet ports, EZ features, storage options (4x SATA, 4x M.2 sockets - one PCie 5.0) and ample connectivity on the rear I/O, including four USB Type-C ports (2x 40 Gbps). The all-black motherboard sports a cool RGB design element of the MSI dragon on the VRM heatsink to break up the monotony and lend some bling to the inside of your chassis. It’s a great motherboard to build your AM5 rig around.

In the end, we like this combo as this, and others like it, help take some of the pain out of buying a PC today. Priced at $1,250.98 with a savings of $118.99 and a free 360mm Rosewill Frostmax Digital ARGB AIO ($90 value), it’s one of the better bundles available on Newegg right now. Get it while supplies last.

President Donald Trump has said that retired Apple CEO Tim Cook should replace Jensen Huang as the CEO of Nvidia, just hours after it was announced that the latter had resigned in the wake of a series of scandals. Taking to his Truth Social platform, the President said, "NOW THAT Nvidia CEO JENSEN HUANG, who has done a fantastic job, has resigned, there is only one man to take his place - TIM COOK of Apple. A Legend replaces a Legend! President DONALD J. TRUMP."

Nvidia's board of directors announced on Wednesday that Huang had resigned following the revelation that he conducted an extramarital affair with a Megaspeed executive suspected of being a Chinese intelligence asset, and a further affair with Nvidia Vice President of Health Care, Kimberly Powell.

Colette Kress, Executive Vice President and Chief Financial Officer, will serve as Interim CEO while the company looks for a replacement. Naturally, names have been floated as potential successor candidates, including Apple's Chief Hardware Officer Johny Srouji.

With Nvidia at the center of the global explosion in AI, Huang and Trump have enjoyed a close relationship that routinely makes news headlines. Trump previously described Huang as 'utterly visionary' in regard to his views on tariffs and rescinding President Biden's AI Diffusion Rule.

President Trump's name has been mentioned in the ongoing Huang affair scandal. Notably, it is alleged that while U.S. intelligence believed that Huang had been the target of a Chinese honeypot operation, reports claim that President Trump was briefed on the matter but chose not to do anything with the information, so as not to affect Nvidia's sales or antagonize China.

Congress has summoned Jensen Huang to appear before the Select Committee on the CCP, with lawmakers threatening to issue Huang with a subpoena if he doesn't agree to attend a closed-door meeting voluntarily.

Tim Cook is currently the Executive Chair of Apple's board, having served as the company's CEO from 2011 to 2026. During his tenure, he oversaw the launch of a plethora of Apple services, including Apple Music and Apple TV, while expanding into new product categories such as Apple Watch and AirPods.

Nvidia's share price plunged more than 10 percentage points at the open on Wednesday, but recovered much of that ground as traders looked to take advantage of the dip in Nvidia's share price. The company will report its third-quarter fiscal 2027 financial results next week, on Tuesday, November 17.

Jensen Huang resigned as Nvidia CEO two weeks after his relationship with a suspected Chinese spy was revealed by intelligence sources and a day after his affair with an Nvidia vice president became publicly known. But hours before this move, when it was just rumored that the Nvidia board was planning to oust him, Elon Musk had already stated his willingness to hire the former Nvidia chief on X. A random X user posted, “Elon please hire Jensen,” to which the billionaire replied, “That would be an honor.”

Despite the scandals that Jensen is currently facing, he still has the engineering know-how, deep knowledge of the industry and its inner workings, plus multiple connections across the world, which would be extremely valuable for any company looking to break into the semiconductor industry.

This is exactly what Elon Musk is doing with his Terafab project, which he first mentioned in November 2025. Jensen Huang even warned that this was going to be a challenge, pointing out that “the engineering, the science, and the artistry of doing what TSMC does for a living is extremely hard.” Nevertheless, Musk poured his resources into the project, with construction on the Terafab facility in full swing and with plans to hit volume chip production by the second half of 2028. Musk himself said that the site targets a production capacity of a million wafers a month by 2032. Tesla has an ambitious AI accelerator roadmap, and SpaceX has even touted making its own GPUs in the future.

We still don’t know what Jensen Huang’s plans are, but even though he’s no longer a part of the Nvidia board and its leadership, Jensen and his wife, Lori, still jointly own approximately 800 million Nvidia shares through various trusts and LLCs, while Huang himself personally owns around 70 million Nvidia shares.

Aside from that, he still has an outstanding invitation from Congress to discuss Nvidia’s security protocols, which could turn into a subpoena if he snubs them, although we don’t know how his resignation would affect the closed-door meeting. There’s also a reported FBI investigation into his relationship with Alice Huang, the Megaspeed executive who was allegedly instrumental in the import of $7.5 billion worth of Nvidia AI GPUs into China and is suspected of working with China’s Ministry of State Security.

Jensen Huang is just the latest in the long list of executives who resigned or were ousted from their leadership position because of a relationship or an affair with a subordinate. This includes former Boeing CEO Harry Stonecipher in 2005, HP CEO Mark Hurd in 2010, Intel CEO Brian Krzanich in 2018, and former McDonald’s CEO Steve Easterbrook in 2019. Musk himself is known to have had relationships with subordinates from his various companies, and that he has even fathered four children with Shivon Zilis, a top executive at Neuralink who started working at Tesla in 2017, through IVF.

It’s still not confirmed if Musk’s reply is a serious offer for Jensen Huang. But if the former Nvidia chief ever decides to work for Musk on his Terafab project, then Jensen could put his effort and expertise into turning it into a formidable competitor to Nvidia.

Corsair needs very little introduction to anyone who has built a PC over the past two decades. Founded in California in 1994 as a manufacturer of memory modules, the company has since grown into one of the largest brands in the PC component and peripheral market, with a catalog that covers cases, cooling, keyboards, mice, streaming gear, and of course, power supplies. Corsair entered the PSU market in the second half of the 2000s and built its reputation there by choosing its OEM partners carefully rather than manufacturing in-house. Channel Well Technology (CWT) has been one of the most important of these partners, producing a large portion of the RMx and HXi units over the years. The HXi series in particular has long been Corsair's high-end, digitally monitored family, positioned just below the flagship AX models.

We take a closer look at the HX1000i Shift Crystal to see if it deserves a spot in our best power supplies list. The HX1000i Shift Crystal is a bit of an odd bird, and Corsair is clearly fine with that. First shown at Computex 2026 and launching on November 12th, it combines the side-mounted connector layout of the SHIFT models with the digital HXi platform, and then wraps everything in a chassis with clear polycarbonate panels on three sides, turning the internals of the PSU into a part of the build's aesthetics. Beneath the show, the platform is built by CWT. It holds a Cybenetics Platinum certification at both 115 VAC and 230 VAC, and it is fully compliant with the ATX 3.1 and PCIe 5.1 design guides. It also integrates an iCUE LINK System Hub and Corsair's PinProtect+ per-pin current monitoring for the native 12V-2x6 connector. The 1000W model that we are reviewing today is rated to deliver its full power output continuously at ambient temperatures up to 50°C and comes with a ten-year warranty.

Specifications and Design

In the Box

The HX1000i Shift Crystal ships in a very large and sturdy black cardboard box, with a large render of the unit on the front that does a good job of showing off the transparent panels and the lit fan. The model name is printed in big yellow and white letters across the bottom half, while the badges in the top right corner - ATX 3.1 with PCIe 5.1 and 12V-2x6 support, the Cybenetics certifications, and the iCUE LINK logo- tell you most of what you need to know before even opening the package.

The bundle is rather spartan for a product of this class. Inside the box, we found the AC power cord, four mounting screws, a handful of black cable ties, and a leaflet with safety and compliance information. There are no Velcro straps, no cable combs beyond those already on the cables, and no other accessories of any kind. It is not a deal breaker by any means, but a premium unit like this that is focused on aesthetics could have been a little more generous in this department.

All of the cables are fully modular and black, with individually sleeved wires and low-profile cable combs. There is a 24-pin ATX cable, two 4+4-pin EPS cables, three PCIe cables carrying four 6+2-pin connectors in total, and a native 12V-2x6 cable rated for 600W, alongside two SATA and two Molex cables with four connectors each. Eight Molex connectors on a modern 1000W unit is extremely generous, frankly more than any reasonable build will ever need. Corsair also supplies two iCUE LINK cables of different lengths and a USB cable that connects the PSU's monitoring interface to the motherboard.

External Appearance

With dimensions of 150 x 86 x 200 mm, the HX1000i Shift Crystal is a long ATX unit, which is a non-issue if you are getting the matching Corsair 4500X case anyway. In any other case, you will need to check your case's specifications before buying one. The side-mounted SHIFT connectors also require some free space next to the PSU, as the cables plug into the side of the unit instead of its face, which is great for cases with a hidden PSU compartment but not suitable for every chassis. The design is unlike anything else on the market. The steel frame is black, with Y-shaped cutouts forming the fan grille and a debossed pattern of the same shape on the opposite panel, while three of the sides are clear polycarbonate panels that reveal the white fan frame, the bulk capacitors, the coils, and the white PCBs. With the RGB fan lit, the unit very much looks like a piece of display hardware rather than a power supply.

The rear of the unit is a little more conventional, with the AC receptacle, a rocker on/off switch marked with the "HX1000i SHIFT" logo, and a large perforated area with the same Y-shaped cutouts for the exhaust. The modular connector bay is on the side, behind the clear panel. The face opposite the AC receptacle hosts the white control PCB, also visible through its clear panel, with two iCUE LINK ports, a USB Type-C port, and a small port marked "TACH". The hexagonal specifications sticker is on the solid steel top side of the unit and is an envelope-type that can be closed to hide the text.

The CRYSTAL is fully compatible with Corsair's iCUE software, which can be used to monitor the PSU, adjust its fan speed, and switch the 12V overcurrent protection between single- and multi-rail modes. Corsair also integrated an iCUE LINK System Hub that, according to the company, supports up to 24 iCUE LINK devices without the need for a separate hub. The most interesting addition, however, is PinProtect+, Corsair's per-pin current monitoring for the 12V-2x6 connector. If the current on a pin exceeds a user-customizable threshold, the PSU cuts power to the connector, and the RGB fan lighting turns red to signal the fault. Given the history of melting connectors, this is a genuinely useful feature rather than marketing fluff. We can confirm that it is working as intended, as we tried pulling 20A from a single wire of the 12V-2x6 connector and the PSU shut down instantly.

Internal Design

Cooling is handled by a 140 mm Corsair NR140HL-RGB fan. Corsair calls it a magnetic dome bearing design, and in practice it is a rifle-bearing fan with a half-domed shaft magnet, rated at 0.55A on 12V. The fan is white with addressable lighting, which also doubles as the visual alarm for PinProtect+. It is a decent fan, and the magnetic shaft should help with longevity, but it is worth mentioning that the regular HX1000i SHIFT uses a fluid dynamic bearing fan, which we would normally consider the better choice for a premium unit.

The OEM behind the HX1000i Shift Crystal is Channel Well Technology (CWT), Corsair's long-time partner for its HXi units. Almost every part inside the unit has been relabeled as "Corsair", including the capacitors and the magnetic components, so the OEM is not obvious at a glance, but the Cybenetics report confirms it. The layout is extremely clean and obviously optimized for aesthetics, which also explains the excessive length of the chassis. The bridge rectifiers have their own heatsink, the APFC components sit on another large heatsink, and the primary switchers have their own heatsink as well, while the secondary side components are on a vertical PCB right after the main transformer. All heatsinks are of good size and quality.

The transient filtering stage has four Y capacitors, three X capacitors, and two filtering inductors. Two bridge rectifiers sharing a large heatsink feed the APFC stage, which relies on two Infineon 60R099P6 MOSFETs and two fully encased filtering coils. The APFC stage also includes two bulk capacitors, one Corsair 560 μF / 420 V capacitor and one Corsair 470 μF / 420 V capacitor, for a combined capacity of 1030 μF. According to the Cybenetics report, these capacitors are supplied by Elite, a relatively reputable Taiwanese manufacturer, and they are rated for 105°C. They are not capacitors that we frequently encounter in anything above mid-tier, though, and we feel they are not a great match for a premium unit.

The primary inversion stage uses two more Infineon 60R099P6 MOSFETs in a half-bridge LLC resonant topology. On the secondary side, ten Infineon 014N04LS MOSFETs handle the synchronous rectification of the 12V rail, and the minor rails are derived from DC-to-DC converters. All of the capacitors on the secondary side are rebranded as Corsair, with black and white sleeves that match the color scheme of the unit, and are also made by Elite. These are mostly good parts, but it is a bit ironic that Corsair advertises 105°C Japanese electrolytic capacitors for the regular HX1000i SHIFT, while the version built to show its capacitors off uses Taiwanese ones.

Cold Test Results (25°C Ambient)

Cold Test Results (25°C Ambient)

For the testing of PSUs, we are using high-precision electronic loads with a maximum power draw of 2700 Watts, a Rigol DS5042M 40 MHz oscilloscope, an Extech 380803 power analyzer, two high-precision UNI-T UT-325 digital thermometers, an Extech HD600 SPL meter, a self-designed hotbox, and various other bits and parts.

At 25°C ambient, the HX1000i Shift Crystal is a very efficient power supply. It averaged 92.1% at 230 VAC and 90.5% at 115 VAC across the nominal load range, with the efficiency peaking at 93.6% (230 VAC) and 92.2% (115 VAC) at around 40% load. The curve is relatively flat after that, with the unit still achieving 91% at 230 VAC and 88.5% at 115 VAC while delivering its full 1000W. The efficiency at very low loads is good as well, at just over 88% with a 50W load on 230 VAC. These figures are in line with the unit's Cybenetics Platinum certification at both input voltages. There is no 80 PLUS certification for this unit.

The acoustics at room temperature are very good. The fan stays completely off up to 50% load, so a typical gaming system with a powerful graphics card will rarely ever wake it up. Once it starts, the unit registers 31.6 dB(A) at 500W, and the fan ramps up very gently, staying around 33 dB(A) at 700W and below 37 dB(A) at 900W. Only at full load does the noise climb to about 39 dB(A), which is still a low figure for a 1000W unit. Thermal performance gives no reason for concern at all, with the primary and secondary heatsinks reaching just 55°C and 58°C respectively at full load.

Hot Test Results (~45°C Ambient)

Hot Test Results (~45°C Ambient)

Inside our hotbox, at roughly 45°C ambient and still within the unit's 50°C rating, the HX1000i Shift Crystal hardly lost any efficiency at all. It averaged 91.7% at 230 VAC and 90.1% at 115 VAC, a drop of just 0.4 percentage points compared to the cold test. The peak efficiency dropped to 93.3% and 91.9%, respectively, and the full-load efficiency to 90.6% and 88.1%. That is a very strong result that shows how well the platform tolerates high ambient temperatures.

The fan is the part that changes its behavior the most in the heat. It starts spinning a bit earlier, at about 400W, registering 32.2 dB(A), and the noise climbs relatively fast to about 36 dB(A) at 600W and 39.5 dB(A) at 700W. Beyond 800W, the fan profile becomes much more aggressive, with the unit reaching 45.7 dB(A) at 900W and 46.8 dB(A) at its maximum output. That is definitely loud, though it takes very adverse conditions to get there. Internal temperatures rose considerably, with the primary heatsink reaching 84°C and the secondary heatsink 89°C at full load, figures that are high but comfortably within what the platform can handle.

PSU Quality and Bottom Line

Power Supply Quality

Voltage regulation is very good. Across the 20% to 100% load range, the 12V rail deviated by 1%, from 12.10V down to 11.98V; the 5V rail by 1% as well; and the 3.3V rail by just 0.7%. These are not the sub-0.5% figures of the very best digital platforms on the market, but all of the rails are well controlled, and the 12V rail stays very close to its nominal voltage across the whole load range. The ripple suppression is excellent. The 12V rail peaked at 48 mV under maximum load and 44 mV during the 12V cross-load test, while the 5V and 3.3V rails both topped out at 26 mV. Compared to the ATX design limits of 120 mV and 50 mV, that leaves a margin of 60% on the 12V rail and about 48% on the minor rails, which is very comfortable headroom for a 1000W unit.

During our thorough assessment, we evaluate the essential protection features of every power supply unit we review, including Over Current Protection (OCP), Over Voltage Protection (OVP), Over Power Protection (OPP), and Short Circuit Protection (SCP).

All protection mechanisms were verified and triggered correctly during testing. With the unit hot, OCP tripped at 148% (37A) on the 3.3V rail, 144% (36A) on the 5V rail, and 132% (110A) on the 12V rail, with OPP shutting the unit down at 130% of its rated output, or 1300W. These thresholds are on the high side, especially for the minor rails, but they are not unusual for an ATX 3.1 unit.

Bottom Line

The HX1000i Shift Crystal is a power supply designed to be looked at, and that is not something we often get to say. Corsair took a CWT-built HXi platform, cleaned up its layout to the point that it can be shown off, and placed it behind three polycarbonate panels with an RGB fan on top. The result is unique, and it is backed by a genuinely modern feature set, with ATX 3.1 compliance, full iCUE support, an integrated iCUE LINK System Hub, and a native 12V-2x6 connector with PinProtect+ per-pin current monitoring. The latter is more than a bullet point on the box, as the unit shut down instantly when we pulled 20A from a single wire of the 12V-2x6 connector.

Most importantly, the looks did not come at the expense of performance. The efficiency is excellent, with averages of 92.1% at 230 VAC and 90.5% at 115 VAC, and it barely moved when we raised the ambient temperature to 45°C. Voltage regulation is tight on every rail, and the ripple suppression is excellent, leaving a large margin against the ATX limits. At room temperature, the fan does not even spin up before the load reaches 500W and it stays quiet all the way up to the maximum output.

The compromises are few, but some of them are hard to ignore on a unit like this. The capacitors are all made by Elite and rebranded as Corsair. Elite is a relatively reputable Taiwanese manufacturer, but its capacitors are not something we frequently encounter above the mid-tier, and they are not a great match for a premium unit when the regular HX1000i SHIFT advertises Japanese capacitors. The fan uses a rifle bearing with a magnetic shaft instead of a fluid dynamic bearing, and it becomes loud under heavy load in a hot environment. The bundle is also spartan for an aesthetics-focused premium product. Finally, the 200 mm long chassis and the side-mounted connectors mean that not every case will accommodate the unit, although most buyers are likely to pair it with Corsair's matching 4500X case anyway. On the bright side, the ten-year warranty covers the unit for a period that very few users will ever need, and it makes the component choices easier to accept.

At the time of this review, Corsair had not made the price of the HX1000i Shift Crystal public, which makes a definitive value judgment impossible. The regular HX1000i SHIFT is the obvious comparison, as it offers the same side connector layout, iCUE integration, and iCUE LINK hub for $339.99 at Corsair's own store, along with a fluid dynamic bearing fan and Japanese capacitors. Compared to other showy high-end units, such as the Asus ROG Thor 1000W Platinum II or the Lian Li Edge 1000, the CRYSTAL is the only one that actually puts its internals on display, and its Cybenetics Platinum efficiency and tight electrical performance mean that it does not need its looks to justify itself. If it arrives at a reasonable premium over the regular HX1000i SHIFT, it is an easy recommendation for anyone building a showcase system, and a well-performing 1000W unit for everyone else who simply likes how it looks, as long as they can live with its parts selection.

MORE: Best Power Supplies

MORE: How We Test Power Supplies

MORE: All Power Supply Content

Nvidia’s board of directors has confirmed today that Jensen Huang has resigned as the company's CEO. The news follows revelations that Huang had been involved in not one but two extramarital affairs, one with a suspected Chinese intelligence asset, and another with his direct report, Nvidia’s VP of health care, Kimberly Powell.

"NVIDIA (NASDAQ: NVDA) Board of Directors announced today that Jensen Huang has resigned as President, Chief Executive Officer, and a member of the Board, effective immediately," the board stated. "Colette Kress, current Executive Vice President and Chief Financial Officer of NVIDIA, has agreed to serve as Interim Chief Executive Officer. Kress will continue to serve as Executive Vice President and Chief Financial Officer while serving as Interim Chief Executive Officer until a permanent successor has been named. The Board is initiating a search for a permanent Chief Executive Officer and will consider internal and external candidates."

In recent weeks, stories have come to light alleging that Huang was involved in an overly familiar relationship with Megaspeed executive Alice Huang, whom U.S. intelligence agencies suspect may be a Chinese intelligence asset. It was reported that the AI GPU giant’s CEO may have been the target of a spy plot — Megaspeed is under investigation for its involvement in the smuggling of $2 billion worth of export-controlled AI GPUs into China.

Nvidia shares plunged more than 10 percent in early market trading following the news, and are now down more than 23% in the last month.

Over the weekend, Jensen Huang admitted that he had been unfaithful to his wife and issued a company-wide apology. Separately, Nvidia’s Board of Directors launched an independent investigation into reports that Huang’s entanglement may be a matter of national security, given his role in the global AI revolution.

Colette Kress has been Nvidia's Executive Vice President and CFO since September 2013; prior to that, she spent 25 years at Cisco and Microsoft, among others.

Just a couple of days later, a second bombshell report emerged claiming that Jensen Huang was also romantically involved with Powell, one of his direct reports at the company. Reports indicate the affair was known within the company, and that the two were even filmed being affectionate in a San Jose hotel at Nvidia GTC. Security allegedly sequestered a bystander who filmed the interaction on camera and made the person an offer to sign a nondisclosure agreement.

Nvidia also announced that Harvey C. Jones will serve as the company's first-ever Chair of the Board; he has previously served as CEO and Executive Chairman of Synopsys.

Austrian cooling outfit Noctua is officially rolling its new NT-CP1 AM5/4 Carbice carbon nanotube thermal pad for AMD AM4 and AM5 processors into the retail market. After announcing its partnership with advanced nanotechnology company Carbice back in June 2026, the premium thermal pad can now be purchased for $24.95 in the US and €24.95 in the European region.

• Grab the Noctua NT-CP1 AM5/4 Carbice carbon nanotube thermal pad on Amazon

Conventional thermal pastes and most thermal pads gradually lose performance over time due to pump-out, delamination, cracking, and other forms of degradation. However, Noctua says that its new NT-CP1 thermal pad takes a different approach. It makes use of Carbice's vertically-aligned carbon nanotube technology, originally developed for high thermal reliability in aerospace systems. The carbon nanotubes gradually conform to microscopic imperfections of a CPU’s IHS over hundreds to thousands of thermal cycles.

This allows heat transfer performance to improve during the initial activation period rather than gradually deteriorating. The thermal pad can also maintain stable, maintenance-free performance for years once the initial activation phase is complete. Thus, making it a suitable option for users who want consistent thermal performance without having to periodically replace degraded thermal paste.

"Carbice pads offer a unique combination of ultimate performance stability and convenient installation. This makes the NT-CP1 the perfect choice for everyone who’s looking for a thermal pad that’s super easy to apply and never needs to be replaced," said Noctua CEO Roland Mossig in a press note.

Unlike conventional carbon or graphite-based thermal pads, which can be brittle, slippery, and electrically conductive, the NT-CP1 features an aluminum backbone sandwiched between layers of carbon nanotubes. The surface is also protected by a nanoscale polymer coating, which Noctua says makes the pad more robust, easier to handle, and less likely to move during installation. The company recommends a simple peel, press, and peel installation process, while the slightly tacky coating helps keep the pad in place and offers electrical insulation as long as the polymer layer remains intact.

With an ultra-thin 90µm profile and precut 33x33mm size, the thermal pad fits both AM5 and AM4 socket processors. Noctua also highlights its manufacturing process, noting that the pad is produced in the US using recycled aluminum and waste carbon gas to grow the carbon nanotubes. The company claims that its manufacturing process is more energy-efficient compared to conventional graphite and graphene-based pads, while the aluminum structure allows the product to be recycled through standard aluminum recycling channels.

Alice Huang, the person allegedly involved in an affair with Nvidia CEO Jensen Huang, has reportedly surfaced in Beijing, marking the first sighting of the Megaspeed executive since the scandal broke out. According to Bloomberg, it’s unclear when the suspected Chinese spy arrived in the capital or why she’s in the area, but she was last spotted leaving Singapore for Shanghai in 2025 after she was detained and questioned during the city-state’s investigation into the $2-billion smuggling case involving Megaspeed.

The Nvidia CEO is now under intense scrutiny after the revelation of his indiscretion with the Megaspeed executive, with Congress threatening to issue a subpoena if he does not respond to their invitation to “a joint, closed-door briefing” with the heads of the House committees on China, Foreign Affairs, and Intelligence. The FBI is also looking into the relationship between the head of the largest AI hardware company in the world and a person suspected of helping sanctioned Chinese companies secure hundreds of thousands of banned H100 and H200 chips. Overnight, a second bombshell report claims that Jensen Huang was also involved in an extramarital affair with Nvidia's VP of health care, Kimberly Powell.

Although the Chinese embassy in Washington has denied any links to Alice Huang, calling it nothing but “groundless smears and slanders” against Beijing, there is speculation that she’s under the control of the Ministry of State Security (MSS), which is the primary agency that handles China’s foreign intelligence and counterintelligence operations. The U.S. intelligence community suggests that she may be passing on information to the MSS, but it remains unclear if she was specifically tasked by Chinese intelligence to target Jensen Huang or if they were only leveraging her relationship with the Nvidia chief to get insights into the U.S. AI industry.

A deeper investigation into Alice Huang, known by her Chinese name as Huang Le, revealed that she supposedly worked in media before jumping into banking and finance. The Wall Street Journal reports that it’s unclear when she met her husband, Kang Wei, but the couple ran a private equity and investment management firm called Shanghai Hexi in 2016. This allowed the couple to invest in companies across industries, including gaming, pharmaceuticals, and, eventually, data centers and cloud computing.

The couple is known for riding trends and isn’t a stranger to scandals. Their firm, Shanghai Hexi, invested in gaming companies in the 2010s, especially when mobile gaming started to boom across the world, while Chinese regulators fined Kang and the company in 2018 for disclosure failures and other violations while it was in the midst of a takeover battle for a corporation. Kang’s moves in the gaming industry are what allowed him to meet Jensen Huang, which eventually led to a close relationship between the couple and the Nvidia chief. Sources told the WSJ that Kang apparently accepted Alice and Jensen’s extra-marital relationship, even sometimes jokingly telling people that she was Jensen’s “personal assistant.”

It seems that the couple is used to leveraging their relationships to gain business advantage, and with investments in AI data centers expected to hit $32 trillion by 2050, Alice Huang’s “close” relationship with Jensen Huang put Shanghai Hexi in an excellent position to take advantage of the trend through Megaspeed and make big money. And with U.S. sanctions making it harder for competitors to acquire Nvidia chips, it appears that their access has turned into a gold mine for the company. But whether Alice Huang acted solely out of business interest with the supposed procurement of Nvidia chips or was leveraged by Chinese authorities to import these chips, the fact remains that $7.5 billion worth of banned chips reportedly made their way into China despite Washington’s efforts.

Best Buy has just slashed $350 off this all-AMD CyberPowerPC gaming desktop, scoring you a powerful 9800X3D build for less. Right now, this gaming PC with an RX 9070 16GB is just $1949.

● Check out this deal at Best Buy

This is a fantastic gaming PC for 1440p, with a strong supporting cast of specs. The 32GB of RAM in this PC alone is worth $450 right now, and a 1 TB SSD will set you back $250.

The Ryzen 7 9800X3D is the headline spec of this build. With its generous 3D cache, it's one of the very best gaming CPUs on the market right now, edged out only by the 9850X3D in terms of gaming performance, as you can see from our bench database below. With 8 cores and 16 threads, this CPU will chew through any gaming title thanks to that 3D cache, which means the chip has to rely much less heavily on RAM, drastically improving performance.

The GPU in this build is the RX 9070 with a generous 16GB of VRAM. Sadly, it isn't the more potent XT version, but this is still a great gaming GPU, averaging nearly 75 FPS in our 19-game rasterization benchmark suite at 1440p. 4K performance was around 41 FPS without frame generation, too, so there is some 4K potential.

The PC also comes with 32GB of DDR5 5600 memory, a slight compromise over 6000 speeds but one you shouldn't notice too much. You also get a 1TB PCIe 4.0 SSD for your operating system and storing games, plenty to get you started.

It's all held together by a B650 motherboard, and ships inside one of CyberPower's more impressive cases, along with fans aplenty. You also get Wi-Fi 6, Bluetooth, lots of USB ports, HDMI and DisplayPort connectivity, and a liquid cooler for the CPU, as well as an 850-watt power supply. And, of course, a copy of Windows 11.

With Black Friday still a couple of weeks away, this is one of the best gaming PC deals going right now and one not to be missed.

Google's eighth-generation TPU family has now expanded further. Back in April, the company announced TPU 8t for training AI models and TPU 8i for general-purpose inference workloads. Google has now introduced TPU 8tx, built on the same silicon as the previously announced 8t, for confidential, hardware-isolated inference aimed at regulated enterprises and governments, and TPU 8i Max, which combines two 8i chips on one package for long-context agentic work.

TPU 8tx swaps the Arm-based Axion hosts of the 8t and 8i for a 5th Gen AMD EPYC CPU, adopting a Boardfly network, and the 8i Max doubles the compute, SRAM, and HBM bandwidth of the original 8i. The original split, the TPU 8t and 8i, made natural sense. The 8t targets large-scale training, while the 8i leans toward low-latency inference and reasoning. TPUs have become a fast-growing business, with the company claiming nearly 100 customers, including Meta’s recent move to TPUs. The more modular approach is flexible and can deliver specialized systems “in a matter of months rather than years,” helping that popularity along.

For the new parts, the TPU 8i Max isn't innovative on the surface, as it’s just two TPU 8i chips combined into a single package. The result is twice the theoretical compute, but more importantly, it fits twice the silicon within the “existing rack footprint,” and Google claims more compute per square foot and per watt.

Doubling both SRAM and HBM bandwidth helps keep things scaling, which would put the 8i Max at about 20.2 FP4 PFLOPS, 768 MB SRAM, and about 17.2 TB/s HBM bandwidth. This compares to Nvidia’s Rubin R200 GPU at a rated 50 FP4 PFLOPS for inference and 22 TB/s of HBM4 per socket, or AMD’s MI455X with 40 FP4 PFLOPS. The compute gap is closing.

The memory bandwidth doubling is also potentially one way around moving to HBM4. The memory wall, as Google calls it, or the bandwidth bottleneck, is a real concern “when dozens of AI agents work together across an entire software codebase.” Larger context windows — with larger models targeting 1 million tokens — mean larger KV caches, partly held in SRAM, and each new token has to read that cache back from memory, which is where the extra HBM bandwidth helps. It’s a different route from Nvidia’s LPX: more bandwidth in one package, rather than a second chip type for decoding and a storage tier for context.

Google also provides useful software for agents to get the most out of its hardware. TPU Sync is an open-source “model-state layer for TPU clusters” that moves KV caches off the chip’s HBM, “automatically tiering massive KV caches off the chip in milliseconds” down to host DRAM and, experimentally, shared storage. Google says this can cut KV transfer latency by “more than 90%,” with time-to-first-token gains that vary by workload. The company has also announced the Agent Compute Engine, a Google Cloud service that pauses idle agent sessions and moves their working memory aside until they’re needed again, another memory-saving effort.

Google credits its vLLM-TorchTPU stack with “state-of-the-art price-performance marks” on InferenceX and AgentX, which test realistic agentic coding and long-context work. However, the only published third-party numbers so far are for the last-generation Ironwood TPU. In SemiAnalysis’s InferenceX tests in September, running Qwen3.5 397B in FP8, Ironwood delivered 50.4% more tokens per dollar than Nvidia’s B200 and 96.0% more than the B300 at 20 tokens per second per user. One issue with that, however, is that Ironwood lacks native FP4, although newer November results show its FP8 beating Nvidia’s FP4 GB300 NVL72 on the same model, at 1.67 times the performance per dollar. There is also no eighth-gen result yet to compare how the new generation differs.

In contrast to the 8i Max, which is designed for long-context work, the new TPU 8tx, “built on the same silicon as TPU 8t,” is “optimized for secure and confidential inference” as well as small-scale training. The security is built into the hardware with isolation to prevent snooping on sensitive data, including model weights. Normally, data has to be unlocked in memory during processing, which is the step this isolation is meant to cover.

This chip is intended for areas where security and confidentiality are high priorities, which means regulated enterprises and governments, particularly those who are using frontier or proprietary models. In that case, the operators may want to “fine-tune models using their private data, distill large models into smaller ones, and train their agents,” without having to move sensitive data between clusters. Agents and security meet here for “secure agent execution.” Back in April, the Pentagon and Google reportedly held discussions for TPU usage in classified settings, although the 8tx chip is not specifically named.

Google’s Private AI Compute, its “hardened TPU platform” with protections “comparable to a typical TEE,” per its September technical brief, already runs confidential serving on TPUs. The 8tx appears to be that capability turned into a product for customers. Alternatives include Nvidia’s NVL72, with one security domain across 72 Rubin GPUs, 36 Vera CPUs, and the interconnects, and Intel’s TDX, which is adding trusted I/O to cover data moving to accelerators.

The TPU 8tx hosts are based on 5th Gen AMD EPYC boards, instead of the Axion hosts on the 8t, which replaced x86 in April. There's one Axion for every two TPUs on the 8i, compared with one Xeon per four on Ironwood. Google gives no reason for the change, but its Confidential VM configuration list has both AMD and Intel series chips, with no Arm-based ones listed. Private AI Compute uses “an AMD-based hardware Trusted Execution Environment” for its CPU workloads, suggesting Google swapped something it already runs on its confidential stack.

The 8tx also adopts the 8i’s Boardfly network over the 8t’s torus, which Google claims can deliver “up to 2x lower latency compared to TPU 8t” with almost no performance penalty, although it’s not clear what measure Google is using. The alternative torus network suits training, since chips mostly swap data with neighbors in a predictable manner, but inference on large mixture-of-experts and reasoning models can send smaller messages between distant chips, with every hop adding delay. With Boardfly, it’s at most seven hops against 16 at 1,024 chips.

What is lacking is a full specifications table, the amount of HBM capacity, pod sizes, an availability date, pricing, and customers for either system. Details on attestation, that is, the ability for the customer to verify confidentiality, are also missing for 8tx. The 8tx and 8i Max are currently listed, along with the 8i and 8t, as “coming soon” on Google’s Cloud TPU page.

Delivering two new TPU systems nearly seven months after April’s pair shows Google wants to turn out specialized systems much more quickly than has been the case in the past. Now, it just needs to prove them, through benchmarks and an availability date when the hardware is actually in customers’ hands.

Enterprising techies looking to build their own local AI cluster have far more interesting options than buying DGX Sparks or Apple Mac Studios. Used data center GPUs are everywhere, and a server rack's garbage is an enthusiast's treasure. This option has become even more popular with CMPunlocker, a soft-mod for Nvidia CMP-170HX cards that unlocks abilities well beyond stock. Now, a techie has taken it a step further, blasting through the remaining limitations with voltage glitching and creating a card faster than the actual $20,000 Nvidia A100 GPU with 80GB of VRAM... but from an $2,500 8GB CMP-170HX mining card.

The volt-mod was done by Kate (GitHub link here), who got her starting point from a paper that described how one could exploit GA100 cards' microcode. She turned her attention to the normally protected onboard fuse controller. Although it's rarely discussed, a "fuse box" has been integral in most any CPU and GPU for a while — though it's normally used once at the factory to break physical connections in chips to parts that have tested poorly, thus disabling them so they can be sold as a lower-end model.

If you're wondering how people are getting 80-40GB of VRAM out of 8-10GB, the explanation is simple: the GA100 silicon actually includes all the corresponding HBM2e VRAM dies within; but they get disabled at the factory like the shader modules. The smaller version is more desirable, as it's based on the variant of GA100 with 80GB of HBM2e within, while the 10GB cards spawn from the 40GB GA100.

The Nvidia cards in question have quite an elaborate eFuse controller (16,384 in a 512 x 32 array) — which can be read to obtain the card's current configuration. Most every part of the card has related fuses: VRAM sections, individual shader modules, PCIe 4.0 and NVLink, and even video codec hardware. However, as Kate succinctly put it: Nvidia's design is insecure, as popping fuses can grant access to more features, not fewer.

Under normal conditions, these fuses can't be popped, or "written to," by the user — but a geek armed with a soldering iron isn't a normal user. Kate had GPT-6 Astra design a simple circuit board based on an $8 STM32G4 microcontroller to tap into the power signals on the card's PCB.

The voltage regulators (VRMs) on the card "speak" PWMVID, an Nvidia standard, and Kate's board intercepted the PWM signals sent into the VRM controller, forcing it to produce voltage glitches on the line feeding the GPU, where the fuse controller lies.

Remember when we said the fuse controller was read-only? With the voltage glitches, many fuse-status reads in a tight loop, and she managed to set the controller to a state in which its security bit flags allowed writes through. From that point, it was "easy" to reconfigure her card. She was careful to force-disable any shader modules previously marked as defective, because these are physical fuses — there's no "undo" button.

She applied the mod to her cards and, after testing with multiple workloads, found that nine had the full 80 GB of VRAM working, and one card "only" had 64 GB usable. Most interestingly, unlocking all the shader modules (SM) nets her FP16 FMA performance 20% beyond a regular A100 card. She thinks this is due to the chips often having 120 or more usable SMs — over the 108 on the spec sheet. The GA100 die starts with 128 SMs to account for defects, which is normal in chip manufacturing.

The CMPUnlocker soft mod is required for enabling the "new" features on the drivers, but it doesn't unlock the card's full potential: the VRAM upper limit is 40-64GB (depending on the variant), many shader modules remain locked, and PCIe bandwidth only goes to 2.0. Additionally, there's no ECC, NVLink, or multi-instance GPU support, all of which are critical in a server setup and are also used for product segmentation.

Kate's volt-mod does away with all of these limits, and her best test card is happily working at PCIe 4.0 speed with the full feature set. She didn't test the video encoder bits, but it's unlikely anybody using these cards cares much for that.

As for the cost, a CMP-170HX 8GB is currently hovering around $3,000, and the 10GB variant costs around $2,500. A full-fat Nvidia A100 80GB in a PCIe configuration can reach $20,000 or more. The SXM server-module versions are cheaper, but they require infrastructure around them. Note that these prices will likely vary significantly, especially as people become increasingly aware of the mods.

Kate says she's publishing her research now because another person independently found the same issue. She has informed Nvidia of the vulnerability, though there's not much the company can do for existing cards already on the market.

The AI data center boom is bringing hope to struggling cannabis farmers as they face increased competition, costs, and taxes amid reduced demand for marijuana. According to Arizona Public Media, the cannabis industry is in decline, with sales peaking in 2024 and oversupply driving prices down. And now, several growers are considering selling their cultivation facilities or converting them into “mini data center nodes.” These facilities already have many of the ingredients needed for AI data centers, such as the increasingly hard-to-get supply of electricity and water infrastructure.

"For some reason, there's a lot more cannabis grown in Arizona than Arizonans use. And so, a lot of folks are getting out of that business, like myself, at least on the cultivation side. But these facilities can potentially be repurposed to be mini data center nodes," Moe Asnani, the former owner of two Tucson D2 Dispensaries, told the public broadcaster. He also added, "The whole idea is if you create these smaller nodes where people control their own data, where they're not at the mercy of other people's updates and things like that — that's kind of where the idea came in, because I believe in decentralization."

Cannabis farming is both power- and water-intensive. Forbes looked at a 32,850-square-foot cannabis facility in Kalamazoo, Michigan, which is currently wired for 1.5 megawatts, which can be expanded to 2.5 megawatts in under a year and 15 megawatts in the long term. It's also provisioned for fiber internet and equipped with a state-of-the-art HVAC system for climate control, which should make it a tempting property for anyone wanting to put up a small data center — something that even OpenAI and Anthropic are interested in as massive hyperscale sites face opposition from the community.

This concept is dubbed "adaptive reuse," with Tucson City Council member Paul Cunningham saying that it’s a more palatable alternative to massive, purpose-built sites sitting on the outskirts of town. "Data centers are dirty words, and everybody knows that. And no one wants to build a brand-new data center anywhere," the politician said. "But if we have buildings that were using that amount of water and that amount of energy already in their previous function and [it] can be done without being invasive, I'm OK with exploring that option."

There is massive demand for data centers, but it’s not a guaranteed payoff for struggling cash-strapped growers looking to unload their unprofitable properties. Niki Krear, vice president of acquisitions of real estate investment trust New Lake Capital Partners, said that there’s a sweet spot between connectivity, costs, and regulation that data center operators look for in a site.

"It’s a combination of finding affordable power, dense fiber networks, and favorable tax policies," Krear told MJBizDaily. Even though many cultivation facilities already boast higher-than-average power capacity for lighting and climate control, data centers often demand twice to thrice more, with upgrades possibly taking up to five years.

Data center operators looking for smaller sites may also have to contend with NIMBY — not in my backyard — fights from neighbors. Additionally, there may be zoning restrictions, moratoriums, and regulatory hurdles, and the lack of economies of scale could end up costing operators more.

This is particularly important as AI chips are expected to increase in power and efficiency in the next five years. As these new AI GPUs become available, data centers will need to sell their existing equipment and buy new hardware to remain competitive. Because suppliers usually offer favorable pricing for larger orders, upgrading these smaller sites will likely end up being more expensive on a per-unit basis. That said, growing opposition toward large new projects could make these former cannabis farms suitable alternatives for operators desperate to get more compute online quickly, provided they check all (or most of) the boxes they need.

Nvidia's Board of Directors is preparing to oust co-founder and chief executive Jensen Huang following revelations of two extramarital affairs, including one with a woman suspected of working for Chinese intelligence and another with an Nvidia executive, reports The Wall Street Journal. Although no final decision has been made yet, the company's directors could meet as early as November 11 to discuss Huang's future. Whoever succeeds Huang will be the company's second CEO in 33 years. Nvidia's stock dropped 8% in after-hours trading following the reports. Nvidia's board of directors launched an independent investigation into the matter on November 8.

The report from the WSJ comes hours after the New York Times revealed that Huang allegedly had an extramarital relationship with Kimberly Powell, Nvidia's vice president of health care, who reports directly to him. According to five people who worked at Nvidia, the relationship was known internally, while Huang and Powell were reportedly seen together at a San Jose hotel during Nvidia's GTC conference in March. The allegations follow an earlier report that Huang had a relationship with Alice Huang, a Chinese businesswoman whom U.S. intelligence believes to be a Chinese intelligence asset.

A similar controversy cost Intel chief executive Brian Krzanich his position in June 2018, when an internal investigation confirmed that he had maintained a consensual relationship with an employee, which violated the company's non-fraternization policy applicable to all managers. It is noteworthy that Nvidia's code of conduct explicitly identifies personal relationships as potential conflicts of interest and instructs employees to consult the company's legal department when such conflicts arise. Whether Huang violated Nvidia's internal policies remains unclear, as the code of conduct does not explicitly prohibit all relationships between managers and subordinates.

Huang's potential departure would mark the first CEO transition in Nvidia's 33-year history, and therefore his succession has become an immediate concern for Nvidia, which does not have anyone of Huang's caliber in the leadership team. Although chief financial officer Colette Kress could potentially serve as interim chief executive if Huang is ousted while the BoD searches for a permanent replacement, it remains to be seen how Wall Street would react to this assignment.

Nvidia co-founder Chris Malachowsky, who used to be the company's senior vice president of engineering and operations and remains a member of the company's board and an Nvidia Fellow, could also play a role in the succession process, given his extensive knowledge of Nvidia's technology, corporate culture, and history. However, has not served in an executive role at Nvidia for at least a decade, so the chances that he will succeed Huang are low.

Back on Sunday, directors expressed confidence in Jensen Huang and his leadership team, a choice of words that did not exactly guarantee his continued tenure.

Nvidia CEO Jensen Huang — already embroiled in a scandal for an extramarital relationship with a woman believed by U.S. intelligence to be spying on behalf of China — reportedly had a separate affair with one of his direct reports, according to the New York Times. Nvidia's board is now reportedly preparing to fire Huang.

The Times, citing five anonymous sources within Nvidia, says that Huang had a relationship with Kimberly Powell, the company's vice president of health care. The sources claim that Huang and Powell were "captured on video being affectionate" in March at a hotel in San Jose, Calif., where the company's GTC conference takes place.

That report says that someone at the hotel filmed the moment on their smartphone, but that Huang's security team "sequestered" them and "made the person an offer to sign a nondisclosure agreement."

Powell joined Nvidia in 2008 and has been steadily promoted. The Times highlights that personal relationships are mentioned as potentially causing conflicts of interest in Nvidia's code of conduct. Like Huang, she is married to someone else.

In 2018, former Intel CEO Brian Krzanich resigned following a relationship with an employee, which was also against the company's policies. Nvidia has not responded to a request for comment from Tom's Hardware. This story will be updated if we hear back.

Earlier this week, it came out that Jensen had an extramarital affair with Alice Huang, formerly a top executive at Singaporean firm Megaspeed and believed by U.S. intelligence to be a spy. Huang admitted to the affair in an email to Nvidia employees. China has denied claims of espionage.

Nvidia's board has an ongoing investigation into that affair, led by the law firm Wachtell, Lipton, Rosen & Katz. Huang is continuing to run the company while that investigation is ongoing. Law firms have been racing to sue the company on behalf of shareholders, including a proposed class-action lawsuit that Nvidia misled investors on material facts about the business.

The Congressional House committee on China has called on Huang for a closed-door session, with Republican lawmakers saying the reports "raise significant national security concerns."

Huang has been a leader of the AI industry, providing the most sought-after chips in the field. He has also had the ear of President Donald Trump, traveling with him and attempting to influence export controls.

Trump told reporters that the FBI was investigating Huang, but also that the Nvidia CEO "had absolutely no idea that the person in question was a spy.”

AMD missed out on deploying Turin-X CPUs in the previous generation, giving its 3D V-Cache tech a four-year hiatus from the data center following the launch of Genoa-X in 2023. In the second half of next year, 3D V-Cache is coming back for HPC with Venice-X, matching the 96 cores and 1,152 MB of L3 cache available on the EPYC 9684X (Genoa-X), but with vastly higher clock speeds, PCIe 6.0, and nearly four times the per-socket memory bandwidth.

After first bringing hybrid bonding to EPYC processors with Milan-X in 2022, AMD followed up a little over a year later with Genoa-X. Turin came and went without a chip sporting stacked SRAM, but now, it’s back with Venice-X. The critical change this time around doesn’t come from big jumps in core count — and consequently cache — as is the case with the standard SP7 lineup. Rather, it’s the 5.3 GHz boost clock speed, another pillar of Venice broadly, that’s particularly beneficial in the HPC workloads Venice-X targets.

As you can see from the top EPYC -X chips from previous generations, boost clocks were significantly lower for the X chips, and higher frequencies were available within the standard chips from each family. Milan scaled up to 4.1 GHz, while Genoa could climb to 4.4 GHz, just not on the EPYC -X chips. Venice-X brings up to the maximum 5.3 GHz boost clock available across the range, unifying a high frequency with a large pool of bonded L3 for HPC workloads.

There are still details missing about Venice-X, including the base frequency, how many SKUs are in the lineup and what their specs are, and the TDP (though the 400 to 600 W range available to other Venice CPUs is likely). The chips appear to use the SP7 socket, however, keeping with previous EPYC -X chips.

The platform comes with 96 PCIe 6.0 lanes, while the smaller SP8 platform targeting enterprises trades memory channels for PCIe lanes. Venice-X comes with the full 16-channel memory setup available on SP7, including support for MRDIMMs at 12,800 MT/s that can drive the quoted 1.6 TB/s of per-socket memory bandwidth.

The critical marriage with Venice-X for HPC workloads, according to AMD, is high frequency with a large pool of stacked cache. AMD currently maintains both 3D V-Cache and high-frequency EPYC chips in its HPC fleet to run its internal EDA workloads. AMD Fellow Phil Steinke says the company runs performance analysis on these workloads to characterize performance, optimize EPYC designs, and push partners to optimize their software.

The biggest consumer on the HPC fleet by far are function verification workloads at over half of AMD's used compute, according to Steinke. These are logic simulations that show up in tools like Cadence Xcelium, and Steinke says the EPYC team is running these workloads “24 x 7 x 365.”

An example of one such workload is above, measuring IPC with AMD's uProf monitoring utility. With 3D V-Cache, AMD says this workload completes 40% faster at around double the instructions per clock. AMD didn't clarify which chips it was comparing or what application it was using. However, the comparison is between an EPYC fleet with and without 3D V-Cache.

Steinke says AMD is able to now move toward a single chip with Venice-X, maintaining the cache advantages in workloads like functional verification while driving the EDA workloads more sensitive to frequency. "I mentioned we have a mix of high frequency and 3D V-Cache CPUs in our data center, and that's because through Milan and Genoa, the high frequency parts were always a little higher on the frequency side," Steinke said. "You don't have that dichotomy [with Venice-X]. You can get the best of both worlds in a single CPU, and this is what we're really excited to be picking up for our own use."

AMD hasn't shared any benchmarks for Venice-X yet, but it provided some more specific HPC benchmarks comparing its high-frequency EPYC 9686F to Intel's Xeon 6980P in several HPC workloads, such as computational fluid dynamics. In AMD's testing, the 9686F was 1.6x faster in these benchmarks while running 32 fewer cores.

We had a chance to test the EPYC 9686F along with the dense EPYC 9996, which you can read more about in our Zen 6 EPYC Venice deep dive and benchmarking article. And indeed, the EPYC 9686F looks very impressive in some of the HPC workloads we ran.

We were only able to test a 2P configuration, and as evidenced by the sole Xeon 6980P, CloverLeaf is sensitive to latency. Even then, the 2P EPYC 9686F completed the complex hydrodynamics simulation in about a third of the time as a single Xeon 6980P.

AMD says it's currently using MEXT, its AI-driven flash-as-DRAM technology, to bolster the amount of memory available for its EDA workloads. When it deploys Venice-X, Steinke says the company will continue using MEXT, which you can read more about in our recent MEXT breakdown on Tom's Hardware Premium.

Venice-X CPUs are launching in the second half of 2027, while SP7 and SP8 Venice CPUs are rolling out now. Also coming in the second half of next year is AMD's Verano, which it's pitching as an AI host node that leverages SOCAMM2 LPDDR5X memory.

Prusa released its new Prusament PETG ESD on Tuesday: a carbon-fiber and carbon-nanotube (CNT)- filled, electrostatic-dissipative (ESD) PETG filament meant to protect static-sensitive electronics. While it’s great for enclosures for a Raspberry Pi, ESP32, and the like, Prusa is targeting the filament more toward electronics production than hobbyist use; however, it’s cheaper than the non-ESD Prusament PETG Carbon Fiber at only $56.99.

• Prusament PETG ESD Black 1kg - $56.99

Prusa isn’t the only one selling ESD filament; 3DXTech and Polymaker were first to the scene, but Prusa is the first to include carbon fiber alongside carbon nanotubes to provide the ESD properties, improving dimensional accuracy and minimizing warping. 3DXTech sells its ESD-PETG at $105; Polymaker’s Fiberon PETG-ESD is $65.98, but the Prusament PETG ESD is only $56.99 and is the only one with carbon fiber reinforcement.

Non-ESD filaments are generally safe for electronics enclosures; many hobbyists use regular PLA, PETG, and ABS without any problems. And while the extra protection can be appreciated, Prusa positions this not for hobbyists, but for electronics production — PCB fixtures and holders for soldering, IC trays for handling and transport, holders for testing multiple boards at once, etc., as Prusa shows on the PETG ESD product page. It saves on manufacturing costs, reducing the need for CNC-machined ESD trays and fixtures, and can be quickly printed on-site with very little lead time. It’s not necessarily for big businesses only either, as indie hardware developers and smaller-scale manufacturers could use this filament to improve their production line without buying expensive ESD filament.

Prusa has not yet published the datasheet containing details about the strength of the filament, but Prusa claims that the filament has increased stiffness and reduced toughness, similar to Prusament PETG Carbon Fiber’s properties. However, the amount of carbon fiber in PETG ESD may be different from PETG Carbon Fiber, as the latter’s purpose is more for strength than for electronics manufacturing.

Prusa is known to use its own 3D printers to print its own components and use its own products for the 3D printers it sells, and on the product page, Prusa says it uses PETG ESD in its own production line to prevent overcoating the edges when powder coating its print sheets. While it isn’t the first or only ESD filament on the market, it’s certainly the cheapest, and when its strength profile is released or tested independently, it may potentially be more worth it than Prusa’s own PETG Carbon Fiber in certain cases.

Residents of Effingham County, Georgia, just filed a RICO lawsuit against their county commissioners, county leaders, and OpenAI and its affiliates, alleging that they committed various offenses in a conspiracy to get OpenAI’s “Project Camellia” approved. According to the Effingham Herald, the plaintiffs — Dr. Michelle Finch-Mincey, Lehman Mincey Jr., and Raymond Carver — are seeking civil damages, saying that the project has harmed their property values and even pointing out that purchase offers for nearby residential properties have dropped or been withdrawn since the data center was announced.

The lawsuit names all Effingham County Commissioners — Chairperson Damon Rahn, Forrest Floyd, Roger Burdette, Jamie DeLoach, Elizabeth Helmly, and Phil Kieffer — as defendants, alongside County Manager Tim Callanan, Effingham County Industrial Development Authority (IDA) CEO Brandt Herndon, the IDA itself, OpenAI, Octans GA LLC, the entity operating the data center on behalf of OpenAI, and Savannah Industrial Development LLC, the agency involved in the sale of IDA property to Octans and OpenAI.

It’s alleged that the defendants committed fraud and concealment through non-disclosure agreements and misleading public statements so that residents remain unaware of the nature of the projects for as long as possible, and that there was a civil conspiracy “to advance Project Camellia through confidential negotiations and closed-door decisions.” They also claimed that the parties engaged in improper land transactions, zoning violations, and civil racketeering.w

Project Camellia is one of OpenAI’s biggest data center projects planned for construction in Effingham County. The site is expected to cost somewhere between $20 billion and $30 billion, to be built across five phases, with a planned compute capacity of 3.2 gigawatts. The company also promised that it will not affect the community’s utility costs, as it will pay for the full cost of all its required infrastructure upgrades, and that it will have minimal water use due to its closed-loop water system. Despite those promises, it seems that some people are opposing the project because of how the county had allegedly tried to secure the land, tax incentives, and zoning approvals without informing the residents until key decisions were finalized and contracts had been signed.

The RICO Act — or Racketeer Influenced and Corrupt Organizations Act — was originally conceived as a way to counter organized crime, particularly Mafia leadership, who often used low-level members to commit individual crimes and do the dirty work for them. However, Georgia’s RICO Act is less stringent than the federal version, which is why it’s been used multiple times for cases that seemingly had nothing to do with the Mafia. For example, it was in the 2023 election racketeering case against Donald Trump and 18 others in the attempt to overturn the 2020 U.S. presidential election results in the state, and it was also used in Young Thug’s trial after the rapper was accused of running YSL Records as a criminal street gang.

Critics say that Georgia’s version is way too encompassing, though, and that it was being used to squeeze defendants involved in more mundane or common felonies. “When you have a hammer, everything looks like a nail,” Georgia State University professor of criminal justice and criminology Volkan Topalli said in Atlanta Magazine.

The actions of both state officials, OpenAI, and its affiliates are not new, though. There have already been multiple complaints of data center developers and local governments using non-disclosure agreements (NDAs) while they negotiate terms for possible data center projects. Developers say they use this tool to avoid “running afoul of insider trading rules,” but people have pointed out that West Feliciana Parish in Louisiana was able to close a data center deal without resorting to one. An AWS data center in Gilroy, California, is also facing scrutiny after residents only discovered the nature of the project when construction began after a five-year permitting saga that relied on zoning rules that were set 45 years ago.

As GTA 6, the most anticipated game of the year, perhaps the decade, nears its November 19 release date, one company has offered its employees the chance to take the day off to play it. Amid the expectation that gamers around the globe will call in sick to play the game instead of heading into the office, PC peripheral and hardware outfit Logitech is getting ahead of the game by giving 7,000 employees the chance to take a day off to play it. Or, if they prefer, they can play any other game instead, the company confirmed in a press release detailing the move.

The new initiative is called "Logi Game Day," and it isn't only available to the top brass. Interns and full-time employees alike can take a day off from now until this time next year, at a time when it's been estimated that the GTA 6 launch could see the U.S. economy take a $1 billion hit in terms of productivity. The game, which is a console exclusive ahead of a claimed early 2027 PC release, is already available for preorder, with pre-loading also up and running. It's a good job, too, with a 212GB PS5 download set to be a real downer for eager gamers with slow internet. The ongoing price hikes also mean that upgrading your PS5 storage for GTA 6 won't be a fun time, either.

As for Logitech, the company believes that it's time for people to move away from the belief that playing games is a waste of time, hence its new initiative. It makes sense for a company that makes gaming accessories to say that, of course, but CEO Hanneke Faber argues that the company sees "how engagement in one sphere supports performance in the other.” The hope will be that, by allowing employees to play games, a productivity boost will follow.

Whether that comes to fruition remains to be seen. But there's less doubt about GTA 6's popularity a little more than a week out from its release. If you've yet to get your order in, you can also grab Grand Theft Auto VI for $79.99 now. You'll get a box with a code that you'll need to redeem on your PS5 to be able to download and play the game.

The game's release brings an end to years of rumors and leaks, not to mention legal battles. Take-Two hit Microsoft with a subpoena in relation to a massive leak earlier this year in an attempt to find the perpetrator. Rockstar Games addressed the leaks in a post on the X social network shortly after the subpoena was made public, saying that it "would be an understatement to say that having videos of Grand Theft Auto VI gameplay leak in this way has been heartbreaking for our team."

SpaceX/xAI is developing a custom version of Nvidia's Vera Rubin NVL72 rack-scale AI system that replaces conventional NVLink cable assemblies with a massive 84-layer printed circuit board (PCB), in part for use in orbital data centers, reports SemiAnalysis. The system is codenamed Kessel and retains 36 Vera CPUs, 72 Rubin GPUs, and nine NVLink switch trays, but introduces a different solid-state interconnect architecture. While SpaceX/xAI will continue to buy Nvidia hardware directly from the company, Elon Musk's company will source other components independently, which could give it greater control over its hardware supply chain.

The Kessel system retains the fundamental configuration of Nvidia's Vera Rubin NVL72 platform, which comprises 72 GPU packages distributed across 18 compute trays and nine NVLink switch trays. However, instead of Nvidia's conventional arrangement with internal NVLink cable cartridges, SpaceX plans to position compute and switch trays back-to-back around a passive PCB midplane. The architecture resembles a smaller version of Nvidia's Kyber design with 576 AI GPUs (albeit in 144 packages), which also relies on a PCB-based interconnect rather than conventional cable assemblies, SemiAnalysis notes. Interestingly, the design is aimed at both terrestrial and orbital deployments, according to the analysts.

Arguably the most distinctive feature of Kessel is its 84-layer PCB midplane, which uses ultra-low-loss M10 copper-clad laminates (CCLs). By contrast, Nvidia's Kyber architecture relied on a backplane based on a considerably more complex 156-layer PCB featuring M9 CCLs. The reduction in layer count by nearly half could substantially improve manufacturability and production yields, but the more advanced copper-clad laminate will almost certainly increase material costs. In any case, an 84-layer PCB is a device that is exceptionally hard to build.

The move to a PCB midplane could simplify system assembly by eliminating numerous high-speed cable connections, but it would not necessarily reduce manufacturing costs. An 84-layer PCB that relies on ultra-low-loss M10 laminate is an exceptionally expensive component that requires sophisticated manufacturing processes and precise quality control. To put an 84-layer PCB number into context, modern high-end server motherboards use 15 – 20 layers, whereas server development boards use more than 20 layers to virtually eliminate any electrical issues with the board itself. Furthermore, manufacturing yields of an 84-layer PCB could become a major obstacle, as defects in internal layers may render an entire board unusable. Consequently, the economic viability of Kessel will depend on whether savings from cable elimination and independent component procurement can offset the cost of its complex midplane.

Perhaps more importantly, Kessel represents a departure from Nvidia's conventional system-level procurement model. Instead of purchasing Nvidia's assembled L6 compute boards or even L10 trays that Nvidia plans to sell with its Vera Rubin generation, SpaceX reportedly plans to acquire AI accelerators, CPUs, switches, and networking components directly from Nvidia and procure other components from third-party suppliers (one can wonder how easy it will be for SpaceX/xAI to procure memory amid today's shortages). This would enable SpaceX to select its own PCB manufacturers, component vendors, and manufacturing partners, optimize its hardware for its infrastructure (which will be important for orbital deployments), and perhaps reduce intermediary costs. Meanwhile, the company would assume greater responsibility for system design, validation, manufacturing quality, and supply-chain management, which is something that most hyperscalers have been doing in recent years.

For Nvidia, the implications are less straightforward. Although direct sales of CPU and AI GPU packages could reduce the company's revenue from complete compute-board assemblies, the arrangement could also encourage other large customers to develop their own Nvidia-based systems and eventually deploy more Nvidia-based machines, something that may be important in a world where custom silicon developed by hyperscalers is rapidly gaining traction. However, to make this happen, Kessel must be a smashing success that will offer clear advantages over systems designed by Nvidia.

By early 2026, Japanese tech investment outfit SoftBank had reserved more power generation capacity than the entire state of Connecticut after cornering a huge portion of the global gas turbine production capacity. Now, the same company has chosen to take a $3.7 billion hit to its earnings after canceling reservations for the highly sought-after power generation hardware. The WSJ reports that SoftBank had reserved more than 13 gigawatts of power-producing turbines, citing unnamed "people familiar with the orders." That's the equivalent of more than six Hoover Dams' worth of capacity, with SoftBank founder and chief executive Masayoshi Son believing that it would be vital to the company's earning potential moving forward. Notably, this development surfaced after SoftBank's CFO fired back at Elon Musk's recent revelation that the company had defaulted on turbine blade contracts, saying, "there is no such default."

The gas turbines were to power AI data centers, and SoftBank is set to build a 9.2-gigawatt power plant in Ohio, with construction scheduled for early 2027. It'll cost $33 billion to build and will provide power to a neighboring eight-gigawatt data center. That data center will be leased to OpenAI, albeit only after Nvidia entered discussions about guaranteeing around $250 billion of financing for the project.

But SoftBank got ahead of itself. After initially planning to use the turbines for future gas-powered plants, the company has now chosen to change direction. Orders for turbines, engines that convert natural gas into the electricity used by data centers, must be placed years in advance of them being required. The company had bet big, placing huge orders with the expectation that demand would greatly outstrip supply. That would leave SoftBank in an enviable position, holding huge orders for the vital components required as part of a data center rollout boosted by AI demand. But those reservations weren't cheap — turbine manufacturers regularly charge customers tens of millions of dollars per gigawatt.

It's now canceled its remaining reservations with GE Vernova, its turbine supplier, while also unwinding contracts for orders that were too far gone. SoftBank told investors on a recent earnings call that the orders had been placed “to secure power generation equipment in anticipation of shortages,” but it now appears it anticipated a little too much.

Crucially, SoftBank is also keen for everyone to know that this move was voluntary and that it did not default on its orders. That's contrary to what xAI CEO Elon Musk claimed in a post on his X social network. SoftBank Chief Financial Officer Yoshimitsu Goto was quick to counter Musk's claims. “This is simply not correct at all,” Goto reportedly said. “There is no such default.”

ASRock has rolled out a new firmware update for its Radeon RX 7000-series graphics cards. The latest VBIOS version 1.0.6 appears to be a follow-up to the previous 1.0.5 release. According to community reports, that older version was bricking some of the company's Radeon RX 7900 XT and RX 7900 XTX graphics cards. The new update has now quietly made its way to the ASRock official website, although the company has not published a changelog or provided any details on what the new update addresses.

Last month, AMD released a VBIOS patch for Radeon RX 7000-series graphics cards to address potential security vulnerabilities. While the company did not disclose any technical details about the vulnerabilities, it assured gamers that there had been no reports of them being exploited. According to a statement shared with ComputerBase, the company provided the following rationale for the update:

"AMD is releasing a firmware update as part of our usual approach to address potential security vulnerabilities. These corrections are already made available before the public announcement to give customers time to update their firmware. As a best practice, we recommend that users update the firmware of their graphics cards as soon as they are provided by the respective GPU manufacturer. There are no reports of exploiting these security vulnerabilities in practice."

ASRock subsequently listed the 1.0.5 VBIOS update for 10 graphics card models, spanning its Radeon RX 7600 XT, RX 7700 XT, RX 7800 XT, RX 7900 XT, and RX 7900 XTX families. However, several Radeon RX 7900 XT and RX 7900 XTX owners reported that their systems failed to boot after installing the firmware. Some lucky users were able to use external EEPROM programmers to recover their cards. Although ASRock has since removed the 1.0.5 update from its website, it has yet to provide clarity on what caused the failures or confirm whether the newly released VBIOS addresses the issue.

There are also unanswered questions around how ASRock plans to help customers who have already been affected. According to reports shared with VideoCardz, a user in Japan received a confirmation from ASRock that the failed update would be covered under warranty. However, another owner in Europe was reportedly directed to seek warranty service despite having an out-of-warranty graphics card. This essentially leaves affected owners uncertain of whether they will receive free repairs or replacements.

Additionally, the lack of a changelog or an official statement, along with the new firmware version, makes it difficult for users to determine whether installing the replacement firmware is safe. We can only hope that ASRock provides some clarity around what version 1.06