Hardware · Page 2

GeForce RTX 5080 24 GB Model Reportedly Canceled
According to a report cited by TechPowerUp, NVIDIA has canceled plans to launch a 24 GB version of the GeForce RTX 5080 in January 2027. The claim cites difficulties sourcing 3 GB GDDR7 memory modules.
AI imageQSAN to Showcase 3U60 NVMe Storage for On-Premises AI at SC26
QSAN will showcase its 3U60 All-Flash Array with JetStor at SC26, taking place November 15–20. Designed to house up to 60 NVMe SSDs in a 3U chassis, the system targets on-premises AI and HPC workloads.
AI imageLenovo announces ThinkSystem servers with 6th Gen AMD EPYC 9006 CPUs
On 10 November 2026, Lenovo announced ThinkSystem servers powered by 6th Gen AMD EPYC 9006 processors, along with new storage options. The portfolio targets AI and high-performance computing workloads.
AI imageAMD EPYC 9006 Venice launches with Zen 6 and up to 256 cores
AMD launched its Zen 6-based EPYC 9006 “Venice” server processors on November 10. The new SP7 platform supports up to 256 cores per socket, PCIe 6.0 and 16-channel DDR5-8000 memory.
AI imagePuget Systems to Show Liquid-Cooled 4U GPU Server at Supercomputing 2026
Puget Systems will showcase a liquid-cooled 4U server developed with Comino at Supercomputing 2026 in Chicago next week. The system features eight NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs.
AI imageGoogle Introduces TPU 8tx and TPU 8i Max: Two New Systems
On November 9, Google added the TPU 8tx, focused on confidential computing, and TPU 8i Max, designed for memory-intensive workloads, to its eighth-generation TPU family. The company aims to open the platform to external customers with TorchTPU and inference software.
AI imageRadeon RX 7500 Added to HWiNFO v8.55-6115 Beta
HWiNFO v8.55-6115 beta adds recognition support for the Radeon RX 7500, which AMD has not announced yet. The card’s specifications and launch plans remain unknown.
AI imageSupermicro Super AI Station Combines Desktop and 5U Rack Use
ServeTheHome reviewed the Supermicro Super AI Station in an article published on 5 November 2026. The system features a Grace processor and a Blackwell Ultra accelerator, and is designed for desktop or 5U rack installation.
AI imageQualcomm Confirms HBC Chip: Pre-Production Step for AI250
Qualcomm announced that it has validated the memory read-write path and near-memory computing on its first-generation High Bandwidth Compute (HBC) silicon at its San Diego lab. HBC will be integrated into the Dragonfly AI250, which is expected to enter sampling in the first half of 2027.
AI imageAMD lifts Q3 2026 revenue to $13.38 billion
AMD reported revenue of $13.377 billion in Q3 2026, with data center revenue up 97% year over year. The company expects revenue of approximately $17 billion in Q4.
AI imageClaim: AMD accounts for most orders for Samsung’s 2 nm production in Taylor
According to TechPowerUp, citing Maeil Business Newspaper, the planned production at Samsung’s second factory in Taylor has been reserved. AMD is said to account for most orders; Samsung did not confirm customer details.

GPU priority setting spotted in NVIDIA App; may be limited to RTX 5090
According to a 5 November report by MyDrivers, a setting that allocates GPU resources between foreground and background applications appeared in NVIDIA App version 11.0.9.585. The option was reported on some RTX 5090 systems.
AI imageLumai Iris Optical Inference System to Be Tested Under Real Data Center Conditions at ARIA
The UK’s ARIA will bring Lumai’s Iris optical inference servers into the Scaling Inference Lab. Testing under commercial data center conditions will assess the company’s energy-efficiency claims.
AI imageA challenging period for memory demand: Team Group points to late 2026–mid-2027
Team Group expects the most challenging period for memory demand to fall between the fourth quarter of 2026 and the second quarter of 2027. The company also warns that the global supply crunch could soon intensify.
AI imageHuawei asks some customers to source HBM for its AI chips
Huawei has reportedly asked some customers to source the HBM memory used in its AI chips themselves. The unit price of smuggled HBM3E has risen to around $4,000.
AI imageGeForce RTX 5080 24 GB Design Kits: Board Partners Still Waiting
According to a report by TechPowerUp citing BenchLife, NVIDIA’s board partners have not yet received design kits for the GeForce RTX 5080 24 GB. The lack of kits makes a near-term launch less likely.
AI imageMarvell–Lumentum deal: $115 billion enterprise value
Marvell announced a definitive agreement to acquire Lumentum in an all-stock transaction. The implied enterprise value for Lumentum is $115 billion, and the deal is expected to close in the first half of 2028.
AI imageNew price hike alleged for AMD Radeon graphics cards, with GDDR6 costs in focus
AMD may raise prices in November for GPU and GDDR6 memory packages sold to Radeon graphics card makers, according to a report. The claim has not been confirmed by AMD, and the potential impact on retail prices remains unclear.
AI imageNVIDIA Rubin GPU Specs Updated: HBM4 and NVLink Bandwidth Reduced
NVIDIA has updated the shipping specifications for Rubin GPUs and systems built around them. HBM4 bandwidth per GPU has fallen from 22 TB/s to 19.2 TB/s, while NVLink has dropped from 3.6 TB/s to 3 TB/s.
AI imageTuxedo Nano Pro Gen15: Mini PC with Panther Lake processor ships with Linux
Tuxedo has announced the Nano Pro Gen15 mini PC with a Core Ultra X7 358H processor. The system ships with Linux and stands out for its upgradeable components, support for up to four displays and compact chassis.
AI imageIntel to Convert Some Oregon 18A Capacity to Intel 3
During its Q3 2026 financial results call, Intel said it would convert some of its 18A resources in Oregon to Intel 3 to meet Xeon demand. The company expects the slowdown in PC demand to continue through at least the first half of next year.
AI imageNV UEFI Unlocker aims to remove GeForce RTX tensor performance limits
physicsdolphin’s NV UEFI Unlocker tool aims to remove the tensor computing limit on GeForce RTX 20, 30 and 40 series cards in Windows and Linux. Early tests report gains, but the tool is experimental and results may vary by model.
AI imageGeForce RTX 5080 Is Missing from NVIDIA’s US Store
The GeForce RTX 5080 is missing from the graphics card lineup in NVIDIA’s US store, along with partner models. Its product page still lists a starting price of $999, while the company has not commented on the card’s status.
AI imageIntel Fab 28 Disruption Could Cost $600 Million in Revenue Over Two Quarters
According to an estimate reported by TechPowerUp, a two-week halt to 10 nm production at Intel’s Fab 28 facility in Kiryat Gat could result in a total revenue loss of $600 million in the fourth quarter of 2026 and the first quarter of 2027.