Intel Arc cards have always leaned heavily on Resizable BAR, to the point where Intel once told Alchemist buyers without it to stick with other vendors, and TechPowerUp's B580 review listed it as required for good performance. As spotted by Videocardz, developer Rod Burlamaqui has now released arc-rebar, a Linux tool that turns it on for Arc GPUs on motherboards that never got a ReBAR option. It runs from the initramfs at every boot, before the GPU driver loads, and doesn't touch the BIOS, kernel, or Intel's drivers. That sets it apart from earlier workarounds like ReBarUEFI, which has to be added to the motherboard's firmware.

The fix targets the PCIe switch built into Arc cards, whose small upstream BAR blocks Linux from enlarging the GPU's memory window. arc-rebar writes the card's ReBAR register, then removes and rescans the root port. Intel engineers have proposed a kernel-level fix for this, though it hasn't been merged yet. On a Supermicro X9DRH-7TF with dual Xeon E5-2670 v2 CPUs and a 2015 BIOS, a B580 went from 256 MiB of CPU-visible VRAM to a 16 GiB window covering all 12 GB. Measured VRAM bandwidth jumped from 15 GB/s to 392 GB/s, and a 4K Vulkan blur that used to stall ran at 39 to 50 FPS.

Intel is preparing to launch its upcoming "Nova Lake" processors in multiple versions, and the latest leaks highlight a particularly interesting Xeon E lineup. According to well-known leaker Jaykihn, Intel will be releasing its Xeon E-series based on "Nova Lake" silicon in two versions: big P-Core only or efficiency E-Core only SKUs. Reportedly, the P-Core only lineup will include five configurations with 4, 6, 8, 12, and 16 P-Cores. Meanwhile, the E-Core lineup will be slightly smaller, featuring four SKUs with 4, 8, 16, and 32 E-Core versions. At the top of the range, Intel Xeon E-series buyers can choose between a 16 P-Core design or a 32 E-Core configuration. There will be no hybrid SKUs with both P-Cores and E-Cores, as those will be reserved for the regular "Nova Lake-S" desktop processors.

Intel's new Xeon E designs will be powered by the W980 chipset platform. Motherboards with the W980 chipset will support native DDR5 memory with ECC, up to 48 PCIe lanes, including 12 PCIe 5.0 lanes directly from the chipset alongside 12 PCIe 4.0 lanes, and Intel vPro technology, along with numerous other optimizations typical of workstation chipsets. The DMI chipset bus bandwidth is set to 128 Gbps per direction, thanks to the implementation of DMI 5.0 x4, which is another great platform addition. It also offers two 40 Gbps USB4 or Thunderbolt 4 ports from the CPU. The PCI-Express 5.0 x16 PEG slot from the processor can be segmented into x8/x8 or other combinations using lane switches. Both CPU-attached M.2 NVMe slots can be configured as Gen 5.

JGINYUE has put AMD's B650 chipset on a PCIe add-in card called the Xscale.io, and it costs just 388 yuan in China, VideoCardz reports. That works out to roughly $58, although some conversions put it closer to $54. The card uses the "Promontory 21" chip to add four M.2 slots, a 2.5 GbE port, two 10 Gbps USB Type-A ports, and one 10 Gbps USB Type-C port to any system with a spare PCIe slot. There are also two internal USB 2.0 headers. The four M.2 slots are not equal, however. The two on the front are PCIe 4.0 x4, while the two on the back are limited to PCIe 3.0 x2 and PCIe 3.0 x1, so the rear slots are probably best left for slower bulk storage.

The card takes a 6-pin PCIe power connector, and it has a 4-pin fan header for an optional 40 mm fan. Apparently, that header runs at a fixed speed, and the fan is not included. This isn't the first card of its kind, since Lekuo launched a B650 add-in card in August with four M.2 slots and four SATA ports, and WisdPi sells a $199 option with OCuLink. JGINYUE's version undercuts both by a wide margin. The company reportedly plans variants with 20 Gbps USB Type-C, 10 GbE, and a half-height design, though there's no word on availability outside China.

The ICY DOCK EXLink MB308A5 is a PCIe 5.0 x16 NVMe host adapter card with dual MCIO 8i (SFF-TA-1016) connectivity, designed for next-generation high-speed NVMe cabling. With an integrated PCIe Gen 5 redriver and 100 MHz clock buffering, it helps maintain signal integrity for 32 GT/s links across longer trace and cable routing commonly found in servers, workstations, and validation platforms.

PCIe 5.0 x16 Lane Bifurcation for Multi-Drive NVMe Topologies
With a PCIe 5.0 x16 interface delivering up to 512 Gbps of total bandwidth, the MB308A5 is designed for platforms that support PCIe lane bifurcation (x4 / x4 / x4 / x4). When enabled, the PCIe x16 interface can be allocated into four independent NVMe x4 links, allowing up to four NVMe SSDs to be addressed through MCIO cabling.

Intel's Core Ultra 400-series "Nova Lake" CPUs are reportedly nearing launch as Intel has successfully passed USB and PCIe compliance testing. This means the platform has officially passed testing with the two organizations, and all the advertised specifications are working correctly. Starting with the USB Implementers Forum (USB-IF), a governing body of the USB standard, there are two "Nova Lake-H" device IDs certified for compliance with USB4. These include devices with IDs D331 and D333. Additionally, USB-IF lists ID 6E6E for the "Nova Lake" PCH-S chipset for desktop motherboards, compliant with USB 3.2 Gen 2. Intel likely chose a desktop chipset and a laptop CPU to achieve Qualification by Similarity, where testing one product can earn certification for other products easily. To avoid revealing too much before the launch, Intel opted for this strategy instead of testing a full flagship desktop CPU.

For the PCI-SIG, a governing body for PCIe compliance, Intel's 900-series chipset has long been present with the IDs 6E38-6E3F, 6E30-6E35, and 6E40-6E47. These were added back in April, but in August, another set of entries for CPUs specifically appeared. Certifying the PCIe 5.0 x16 root complex were device IDs D461, D465, and D467-D46A, which are likely "Nova Lake-S" CPUs or some mobile variation. Regardless, the verification process is the final stage that a platform and the CPU usually go through before the launch is finalized.

After showing off its Gorgon Halo mini workstations at IFA 2026, Acemagic has officially launched the F9A mini PC as an AI workstation powered by the AMD Ryzen AI Max+ APU with its Radeon 8065S iGPU, for 131 TOPS of AI compute and up to 192 GB of unified memory. The Ryzen AI mini PC's all-metal chassis measures in at just 158.5 x 158.5 x 81.5 mm, or 2 L volume, and pricing starts at $6,499 for 192 GB of memory and a 2 TB SSD.

The mini PC features a modern ports selection, with front I/O consisting of dual USB 3.2 Gen 2 Type-A ports, a USB4 Type-C port, an SD card reader, and a 3.5 mm audio combo jack. The back of the mini PC, meanwhile, features another USB4 port, DisplayPort 2.1, HDMI 2.1, dual 2.5 GbE, a USB 2.0 port, another USB 3.2 Gen 2 port, and an OCuLink port for eGPU expansion in the event that you need more power than the Radeon 8065S can put out. The 192 GB of LPDDR5X-8355 unified memory can be configured so that 160 GB is dedicated to video memory.

The Asahi Linux project development team has long been working on bringing Linux to Apple Silicon MacBooks and desktops, going so far as to show off games on the Arm-powered laptops. The latest patches to Asahi add proper support for USB4 and Thunderbolt on MacBooks with the M1, M2, and M3 SoCs, and that support is now making its way to the mainline Linux kernel, as per an upstream patch by prominent Asahi developer, Sven Peter.

This series adds initial USB4/Thunderbolt support for Apple Silicon M1, M2 and M3 SoCs. Each USB4 capable Type-C port comes with an ACIO ("Apple Converged I/O") block which contains a Cortex-M3 coprocessor and the hardware implementing the USB4 router.

Lexar has been exploring various technologies to help consumers achieve faster data throughput and more reliable storage. The company now envisions a shift as the PC evolves from a traditional personal computer to a local AI-enhanced experience. We interviewed Lexar's Chief Technical Officer, Daniel Guo, about the technology Lexar is developing to help reduce some DRAM demand and shift the balance towards integrated hardware-software AI storage solution with NAND Flash. According to Guo, DRAM is about six times more expensive to manufacture than NAND Flash, and AI SSDs present opportunities to reduce DRAM requirements for running AI models on local hardware. This is where the Lexar AI Storage Solution comes into play, as the company is creating new storage solutions to support local AI deployments in constrained DRAM capacity by offloading some parts of large language models to SSDs. This approach allows larger and more powerful LLMs to fit into a PC build, reducing DRAM capacity requirements by up to 40% in specific scenarios.

Based on internal testing, Lexar managed to run the Qwen 3.5 122B AI model on a local PC. Traditionally, users would need to invest about $4,500 in a PC with a decent CPU and 128 GB of DRAM to run this model. Through hardware and software optimization, the Lexar AI suite with the Lexar AI SSD can reduce the DRAM requirement to 32 GB and run the model with 35 billion parameters at 15.6 tokens per second, compared to only 5.2 tokens per second using traditional frameworks. When attempting to load the 122B model on 32 GB of DRAM, the traditional Llama.cpp fails to load and crashes, while Lexar's SSD offloading provides about 4.4 tokens per second.

We've already covered a number of DeepCool's latest innovations at Computex 2026, from its latest CPU and AIO coolers to its new silent daisy-chain fans, but the brand has also shown off a new prototype SFF PC that flips the script on PC building. The as-yet-unnamed mini PC packs the motherboard and main hardware into a small chassis with a cutout over the CPU where one of DeepCool's new Assassin V coolers sticks out into the open air. This protruding CPU tower cooler presents two main advantages: The CPU always has unfettered access to fresh, cool air, and CPU performance is no longer constrained by the size of the cooler that can fit into the case.

Of course, there are obvious drawbacks, namely that there is a giant hole in the case, where dust will undoubtedly enter the mini PC, and that the air being forced through the tower cooler isn't filtered, so dust and debris will eventually accumulate on the CPU cooler. In addition to the CPU cooler that sticks out of the case, DeepCool's prototype mini PC has a GPU module, which is a separate compartment entirely that connects to the CPU/motherboard module with a PCIe riser cable and slots in under the base module. The compartmentalization of components would likely help with cooling, but also serves to enhance the aesthetics of the mini PC somewhat. It's unclear if and when the prototype mini PC will launch, but it will likely have to compete with the likes of the Steam Machine if it does make it to the mass market. DeepCool also didn't give any indication of what sort of internals we could expect from this prototype, but given the standard cooler compatibility and full-length PCIe riser, it seems likely that the modular mini PC is hiding an ITX or microATX motherboard.

China appears to be the first in line to receive laptops powered by Intel's "Wildcat Lake" Core 300 series processors, and the initial models have already arrived. The most affordable option is the CHUWI UniBook, which features a five-core Intel Core 3 304 processor. This is the lowest in the series and the only five-core "Wildcat Lake" processor available. The laptop comes with 8 GB of LPDDR5X memory running at up to 7,467 MT/s and 256 GB of SSD storage. All of this is housed in a 14-inch body with a 1920x1200 IPS display. Interestingly, this model poses a direct challenge to Apple's MacBook Neo, operating on Windows 11 Pro and a different x86 platform.

Additionally, other laptop manufacturers have started listing their "Wildcat Lake" models on the Chinese retailer JD, including ASUS, HP, and Honor. These offerings include the ASUS Fearless 14SE 2026, HP OmniBook 3, and Honor Notebook X14 2026 Combat Edition. All three models are powered by the Intel Core 5 320 processor, which features a six-core design. They also come with 16 GB of LPDDR5X memory and 512 GB of SSD storage. The prices for these models are $597 for the ASUS, $571 for the Honor, and $662 for the HP OmniBook 3. These laptops offer twice the storage and memory compared to the popular Apple MacBook Neo, aiming to surpass Apple in the entry-level PC market. Interestingly, sales tax is not included in these prices.

GPD has officially announced its Box mini PC and its G-series eGPUs, which both now feature faster external PCIe interfaces for gaming and AI use cases. The GPD Box itself is a fairly standard Panther Lake mini PC, although not too much has been revealed about the Box aside from its new eGPU interface. The GPD Box swaps out the traditional USB4 or Thunderbolt 5 port with an MCIO 8i (Mini Cool Edge IO) port, which is capable of full PCIe 5.0 ×8 bandwidth. MCIO 8i aside, the Box looks to have a healthy assortment of ports, with dual USB Type-A and Type-C ports, alongside a 3.5 mm audio jack on the front and two USB Type-A ports, DP and HDMI for display, and dual Ethernet ports on the back of the mini PC. That MCIO ×8 port is theoretically capable of bidirectional bandwidth of up to 512 Gbps, although there are legitimate concerns about rated insertion cycles and wear on the ports, given that MCIO was initially designed for server applications, where hardware would be plugged and unplugged far less frequently than in consumer applications. It should also be noted that it is limited to PCIe ×8, even on PCIe Gen 5, meaning it will be limited to 256 Gbps bidirectional bandwidth on PCIe Gen 5.

The GPD G2 itself is the eGPU enclosure launching alongside the Box mini PC, and it is capable of connecting either over USB4 v2.0 or MCIO 8i, making it useful for more than just the GPD mini PC. GPD claims that the G2 can run an NVIDIA RTX 4090 with only 2% performance loss, which is a level of throughput that neither Thunderbolt 5 nor OCuLink can achieve. The G2 eGPU dock itself has dual USB Type-A ports for peripherals and a USB4 port with 100 W PD charging, and there's a built-in PCIe SSD expansion slot in the eGPU dock for extra storage. The G2 also has its own on-board PSU that connects to the GPU with an externally routed 12v-2×6 cable—GPD has not yet revealed the output of that unit, but presumably, it will at least be enough to power an RTX 4090 if that's what the company is staking its performance claim on. Availability and pricing have not yet been announced, but the Weibo announcement post lists 2026 as the release date, so we will likely find out more, including more specifications, later this year.

If your workstation build has no budget limits, there's now a way to get 16 TB of NVMe PCIe 4.0 SSD storage on a single M.2 2280 SSD. However, this small storage drive will set you back nearly $16,000, with Amazon listing a brand new Exascend PE4 16 TB M.2 SSD at $15,935. What makes this drive so expensive is the fact that you are getting 16 TB of NVMe SSD storage in a super-dense configuration that fits into a single M.2 slot on a motherboard, requiring no additional drives to achieve this capacity. The Exascend PE4 drive uses a PCIe 4.0 interface and can reach sequential read speeds of up to 3,270 MB/s, while write speeds can reach up to 2,980 MB/s, just shy of 3 GB/s. The drive uses TLC 3D NAND Flash storage modules from an unknown manufacturer.

The drive is designed to withstand a TBW of 16,640 TB written to it and about two million hours of MTBF, meaning its usable life is guaranteed to last through any known task, running 24/7 in systems where reliability is the number one factor. Interestingly, Exascend only offers a five-year warranty but claims its hardware will significantly outlast that. Rigorous testing and massive capacity justify the high price point, so some users might find the need for such a drive. Additionally, idle power consumption remains under 1.3 W, while this SSD can reach up to 7.2 W of active power when under full load, meaning that efficiency for such a drive is rather good.

When it first launched the Q-Release Slim PCIe mechanisms on its high-end gaming X870 Gaming motherboards, the new connector ended up causing scratches on the PCIe contacts of GPUs after repeated insertion and removal cycles. The connector was later quietly revised in March 2025 to remove the metal bracket that was blamed for the scratched contacts. Now, ASUS has once again revised the Q-Release Slim PCIe mechanism design for the latest Neo series motherboards, once again requiring a button to release PCIe cards from the motherboard.

In a video previewing the updated 2026 motherboard line-up, ASUS North America explains that the Q-Release mechanism was revised because "some users were confused" by the connector, which didn't require any intervention beyond pulling the GPU out of the slot to remove it, while others simply wanted a button to be present—likely those who had heard about the aforementioned scratched PCIe connectors and didn't want the added risk. The new motherboard PCIe retention mechanism will be present on all of the new NEO motherboards ASUS announced for 2026, which includes more affordable gaming motherboards, like the ASUS TUF Gaming B850-Pro WiFi7 Neo and the creator-focused ProArt B850-Creator WiFi Neo and goes all the way up to the premium ROG Crosshair X870E Glacial motherboard.

Intel might be saving the CES 2026 show for something other than the expected "Panther Lake" laptop design showcase, as more evidence is growing that Intel's big Arc "Battlemage" GPU is getting ready for launch. This time, it is in Intel's software libraries, which are now enabling support for the yet non-existent Arc B770 GPU. Inside Intel's SYCL templates for linear algebra, which accelerate math with heavy software and hardware optimization, there is a mention of B770 inside "Battlemage" recognition software. This means that Intel is likely scheduling the launch of the big "Battlemage" BMG-31 GPU to coincide with the CES 2026 launch of "Panther Lake," likely as a surprise to gamers and those wanting to upgrade their older "Alchemist" systems.

The Arc B770 GPU will be equipped with roughly 32 Xe2 cores and 16 GB of GDDR6 memory on a 256-bit interface. The design is reported to offer a full PCIe 5.0 x16 connection and will be available in multiple SKUs, allowing Intel to cater to both gamers and professional users. A workstation-oriented Arc Pro variant could further expand memory capacity for single-GPU configurations. This GPU is expected to have a 300 W TDP, marking Intel's highest-ever rating in the consumer Arc discrete GPU sector. A 300 W board-level rating significantly exceeds Intel's recent mainstream offerings. Previous high-end cards, like the Arc "Alchemist" A770 and several B-series "Battlemage" models, have board power figures ranging from 190 to 225 W.

Intel plans to ship "Nova Lake" in a five-SKU lineup that covers desktops and several mobile segments, with its Xe3P next-generation GPU generation serving as the integrated GPU. The lineup includes "Nova Lake-S" for desktops, "Nova Lake-U" for standard low-power laptops, "Nova Lake-UL" for ultra-low-power devices, "Nova Lake-H" for gaming laptops, and "Nova Lake-HX" for high-performance mobile systems. This range indicates that Intel aims to extend Nova Lake's reach beyond the primarily mobile-focused Panther Lake generation, offering clearer product differentiation for OEMs. The company seems to be selectively enabling advanced GPU features across these SKUs rather than providing a uniform feature set.

Initial driver listings suggest that ray tracing will be available on select models. Nova Lake-U and Nova Lake-H are expected to support ray tracing, while Nova Lake-S, Nova Lake-HX, and Nova Lake-UL may not. This type of segmentation is a common strategy to differentiate products using the same silicon, influencing purchasing decisions for gamers, creators, and laptop buyers. On the open-source front, preliminary work to integrate Xe3P identifiers and basic support into Linux and Mesa has begun. Patches have been introduced to add Nova Lake PCI IDs and entries for the Iris Gallium3D OpenGL driver and the ANV Vulkan driver, though the support is currently preparatory and marked as experimental. Early enablement suggests that we will be hearing a lot more about these SKUs soon, and how Intel plans to separate Xe3P iGPU into ray tracing and non-ray tracing SKUs.

External graphics enclosures rely on external PCIe delivery to make sure that desktop-class GPUs aid laptop users in their quest for maximum performance. Usually, this is delivered over OCuLink or Thunderbolt 4/5 connectors. Independent testing by Try Some Tech found that a direct PCIe connection over OCuLink delivers the highest sustained throughput, while Thunderbolt 4 and 5 trade some performance for convenience and compatibility. In lab measurements, OCuLink reached roughly 6.6 GB/s host to device and 6.7 GB/s device to host. Thunderbolt 5 recorded about 5.6 GB/s host to device and 5.8 GB/s on the return. Those bandwidth gaps translated into measurable differences in real applications, especially storage workloads and sustained transfers, where sustained throughput matters more than peak bursts.

For testing, Try Some Tech mounted an NVIDIA GeForce RTX 5070 Ti in an enclosure and compared three setups: native PCIe, external via OCuLink, and external via Thunderbolt 5 using an Intel Core Ultra 7 265K testbed. The test included the following titles: Clair Obscur: Expedition 33, Indiana Jones, The Talos Principle 2, Days Gone, God of War, Red Dead Redemption 2, Cyberpunk 2077, Ghost of Tsushima, Hogwarts Legacy, Spider-Man: Miles Morales RT, and Spider-Man: Miles Morales. Compared to Thunderbolt 5, OCuLink increased gaming performance by up to 16.28% on average. Additionally, OCuLink is about 7-8% slower than a direct PCIe connection to the motherboard. When comparing direct PCIe to Thunderbolt 5, the TB5 connection lagged by roughly 19-25% on average, with worst cases approaching a 36% drop. OCuLink benefits from lower protocol overhead and lower latency, but it remains a specialist solution with limited laptop support, no hot swap, and no integrated power or display features. Thunderbolt is a more practical choice for most buyers because of its plug-and-play convenience, charging, and display passthrough.

ASUS unveiled a prototype this week with a new method for graphics cards to draw power from the motherboard's PCIe expansion slot, proposing a solution that could eliminate the need for card-mounted power cables for many entry-level and mid-range GPUs. The concept focuses on beefing up the physical PCIe power path: ASUS engineers widened and thickened the contact pads at the front edge of the card's PCIe interface, enlarged the copper area inside the corresponding slot, and fused several 12-volt lanes into larger, lower-resistance conductors. To share the load, the traditional 8-pin connector is relocated onto the motherboard, allowing the slot and onboard connector to work together. Higher-conductivity materials and a larger terminal cross-section are used to increase the continuous current capacity without modifying the card's external form factor.

This idea could be a good match for GPUs in the 180-220 W range, where customers might appreciate a cable-free card. Examples could include RTX 5060 Ti-class parts or comparable AMD models. The company has not released full schematics, continuous current ratings, thermal validation, or long-term endurance data, and the headline 250 W claim leaves some questions unresolved. A conventional slot plus a motherboard auxiliary connector typically totals 225 W, where the 8-pin supplies 150 W and the PCIe connector gives 75 W. ASUS is already known for experimenting with alternative power delivery, with an example being the Graphics Card High-Power (GC-HPWR) connector. This is a higher-power design, rivaling NVIDIA's 12V-2x6V connector.

PCI-SIG has announced a clear step forward for peripheral connectivity: PCIe 8.0 is being developed to run at 256 GT/s, which translates to about 512 GB/s in each direction and roughly 1 TB/s of simultaneous bidirectional bandwidth in a full x16 slot. Version 0.3 of the PCIe 8.0 specification has been released to member companies for review, marking the first formal draft from the working group. This draft is an early milestone on a multi-year path toward a final 1.0 release, and member feedback will shape further drafts as the consortium works to complete the standard by the targeted timeframe in 2028.

The draft reiterates several engineering priorities beyond raw speed. PCIe 8.0 will continue to use PAM4 signaling while doubling the bit rate over PCIe 7.0 and quadrupling it over PCIe 6.0, with a strong emphasis on keeping compatibility with past generations. Power efficiency and protocol optimizations for even higher rates are under active consideration, and the working group is testing new connector approaches. Optical links remain a topic of interest because they could extend reach and reduce energy use, but cost and interoperability questions will need to be resolved before they become mainstream in the PCIe ecosystem. Vendors and designers are already reviewing the draft as they plan silicon, switch, and system roadmaps that will leverage the next generation of bandwidth, particularly in high-performance computing and AI.

d-Matrix today announced the expansion of its AI product portfolio with d-Matrix JetStream, a custom I/O card designed from the ground up to deliver industry-leading, data center-scale AI inference. With millions of people now using AI services - and the rise of agentic AI, reasoning, and multi-modal interactive content - the industry's focus is quickly shifting from model training to deploying AI at ultra-low latency across multiple users.

When combined with d-Matrix Corsair accelerators and d-Matrix Aviator software, JetStream I/O accelerators can scale to state-of-the-art models exceeding 100B parameters, delivering 10x the speed, 3x better cost performance, and 3x greater energy efficiency than GPU-based solutions. JetStream's addition to the company's product portfolio makes d-Matrix one of the few AI infrastructure providers offering a complete platform that spans compute, software, and networking.

NVIDIA's top-end consumer GPU, the GeForce RTX 5090, and top-of-the-line ProViz SKU, the RTX 6000 PRO, have been plagued by a new virtualization bug. According to developers at CloudRift, who are creating a GPU cloud for AI developers, they have encountered a specific bug that renders the RTX 5090 and RTX PRO 6000 completely unresponsive. After a few days or weeks of constant usage, the GPU virtual machine can completely freeze without a sign of responsiveness. This occurs at random times, with no clear indication of why. The team has tested multiple GPUs, including the H100, B200, and older RTX 4090 models, all of which showed no issues. Not even the highest-performing server-grade B200 GPU from the "Blackwell" family experiences these issues, but consumer and ProViz SKUs do.

Behind the scenes, a more technical approach explains the process of locking the GPU. When a GPU is handed off to a virtual machine via KVM and VFIO, the host performs a PCIe function-level reset (FLR) as part of the normal cleanup process when the VM stops or the device is moved. Instead of coming back online after that reset, the card becomes unresponsive. The kernel times out and reports the failure with the message "not ready 65535ms after FLR; giving up." Hence, the only point of failure is the GPU itself, and CloudRift has even issued a $1,000 bug bounty for anyone who can resolve the issue.

Intel used Hot Chips to pull back the curtain on "Clearwater Forest", its next-generation E-core Xeon. The company has paired major architectural upgrades with its newest 18A process and advanced 3D Foveros packaging paired with EMIB. The chip replaces "Sierra Forest" and is built around the new "Darkmont" efficiency cores. Intel says Clearwater Forest will ship in a massive chiplet package using the 18A node for the compute pieces while relying on Foveros Direct 3D for stacking. The company is pitching this design as a showcase of what its foundry and packaging teams can now deliver.

Under the hood, the Darkmont E-core leans heavily on a wider, more parallel front end and a beefed-up out-of-order engine. Intel disclosed a 64 KB instruction cache per core and an expanded decoder that can handle more instructions per cycle than the previous "Crestmont" E-core design. Reorder and allocation structures have evolved, allocation units have been increased, and the out-of-order window has been enlarged to accommodate more in-flight work. Execution resources were also multiplied, with many more execution ports for integer and vector math, allowing the cores to sustain much higher parallel throughput.

According to a new patent filed by Sony Interactive Entertainment Inc. at the World Intellectual Patent Organization, the console maker may be moving to a simpler model for game development by putting the necessary hardware for game development into a host PC or server. Patent W02025170845 details a design that effectively puts "compute resources substantially similar to compute resources of a retail game console" onto a PCIe card and allows communication and translation between the host system and the on-board APU, storage, and other console hardware via a specialized management engine.

The patent details that this idea was partially inspired by the difficulties developing games for proprietary console hardware, but it also references cloud gaming, suggesting these slot-in cards may also come in handy as cloud compute modules that can be used in servers to support game developers who don't have access to dev kit hardware. A slot-in solution like this could also theoretically allow developers more freedom when it comes to testing and diagnosis, since they would be able to use hardware and software testing tools connected to the host system, instead of having to rely on built-in diagnostics and hardware support. The patent also specifically mentions being able to capture video from the host system, which could also help diagnostics and game optimization.

As a global leader in memory technology, Team Group Inc. has long been providing diverse storage solutions powered by cutting-edge technology. Today, the company announced the launch of it's brand-new NV series of M.2 solid-state drives, starting with the Team Group NV5000 M.2 PCIe 4.0 SSD, a high-efficiency entry-level drive utilizing the PCIe Gen 4x4 interface. The new NV series, short for Nova Value, symbolizes a shining new star in Team Group's product lineup, delivering an optimal combination of performance and quality at outstanding value. Fully compatible with the latest Intel and AMD platforms, the NV5000 is designed for both desktops and laptops, offering smooth performance for storing high-resolution photos, 4K videos, and large files, making it an ideal upgrade for seamless productivity and entertainment.

The Team Group NV5000 M.2 PCIe 4.0 SSD features advanced 3D NAND flash technology and utilizes a PCIe Gen 4x4 interface compliant with the NVMe 1.4 standard. Offering up to 2 TB of capacity and read speeds of up to 5,000 MB/s, it delivers performance that's over twice as fast as PCIe Gen 3 SSDs and more than nine times faster than traditional SATA SSDs. The NV5000 reduces load times and enhances multitasking performance for everyday tasks, including productivity activities such as document management and data backup, as well as entertainment activities like video editing and game launching. The SSD is equipped with Team Group's exclusively patented heat-dissipating graphene label which effectively lowers operating temperatures. Measuring under 1 mm thick, the label eliminates installation interference and works seamlessly with motherboard M.2 heatsinks to further improve cooling efficiency.

A lack of PCIe bandwidth can reduce RTX 5090 performance by as much as 25% in video editing and 3D workloads. Puget Systems conducted a series of tests on NVIDIA's GeForce RTX 5090, comparing different PCIe versions and lane configurations to assess the impact on content creation tools. At its launch, TechPowerUp's scaling tests noted that the GeForce RTX 5090 is the first high-end card to adopt PCIe 5.0 x16, delivering 64 GB/s of bidirectional bandwidth—twice the 32 GB/s that the RTX 4090 achieves on PCIe 4.0 x16. Our testing revealed a potential performance impact of up to 25% when utilizing limited PCIe bandwidth. Puget Systems today put this interface to the test, confirming real-world impacts on video editing workflows when lanes or generations are constrained.

In DaVinci Resolve benchmarks, Puget Systems found that configurations running at PCIe 5.0 x16, PCIe 5.0 x8, or PCIe 4.0 x16 yielded virtually identical render times. Dropping to PCIe 5.0 x4, PCIe 4.0 x8, or PCIe 3.0 x16 introduced a modest 10% slowdown. Further reducing bandwidth to PCIe 4.0 x4 or PCIe 3.0 x8 resulted in an increase of roughly 25% in render times. After Effects exhibited only minor slowdowns once the bandwidth dropped below 8 GB/s. Unreal Engine 5.5 virtual production tests recorded about a 7% drop in average frame rates at the lowest lane counts. By contrast, Blender offline renders and OctaneBench scores remained essentially unchanged, and Llama LLM benchmarks showed no measurable dependency on PCIe speed.

ASUS has quietly added a useful new feature called "Equipment Installation Check" to its ASUS ROG Astral GeForce RTX 5090 series graphics cards, and many users only noticed it after a recent GPU Tweak III update. This tool watches for any unwanted tilt or sag in the GPU and sends an alert if it spots more movement than ASUS thinks is safe. The secret sauce here is a tiny Bosch Sensortec BMI323 inertial measurement unit, or IMU. It's essentially an accelerometer and gyroscope in one, and it costs just a couple of US Dollars. Tracking the card's angle lets GPU Tweak III know when you might need to give your graphics card a little extra support. That's especially handy for models like the air-cooled ROG ASTRAL RTX 5090, which can weigh about 3 kg and put real stress on your motherboard's PCIe slot over time. Equipment Installation Check joins two other hardware monitoring tools ASUS already offers.

Additionally, some features like Power Detector+ monitors each of the 16-pin 12 V rails for any odd current spikes. Thermal Map shows temperatures from sensors scattered around the PCB so you can spot hotspots at a glance. Together, these features give you power, heat, and tilt monitoring all in one place. Most PC builders use third-party GPU brackets or reinforced PCIe slots to hold up heavy cards. Some of those supports might be optional because the software will remind you if it senses a problem or if a support bracket slips or loosens. That's a big deal if you only move your desktop once in a blue moon and don't notice slow shifts in component alignment. GPUs keep getting bigger and more demanding mechanically, so it's easy to forget that they're also a mechanical load on your system, especially when moving the PC for cleaning or part upgrades.