Qualcomm has introduced the AI200 and AI250 data center accelerators with a focus on memory capacity and data movement efficiency. The AI200 card supports 768 GB of LPDDR memory , which allows large language models and multimodal models to run with fewer parts and without constant off-card transfers.
The AI250 uses a near-memory compute layout designed to increase effective memory bandwidth during inference, which is where workloads often stall due to
MSI has listed two compact GeForce RTX 5050 graphics cards from its INSPIRE ITX lineup. Both models use the same small dual-slot cooler and share identical core specifications, with the only difference being the factory tuning.
The standard GeForce RTX 5050 8G INSPIRE ITX (G5050-8II) features a boost clock of 2572 MHz , while the INSPIRE ITX OC (G5050-8IIC
Gamers were not provided access to the 32GB version. Professionals, however, were.
Today marks the third embargo lift for Radeon AI PRO R9700. The card was announced in May and released to OEMs and system integrators in July. Three months later, AMD is launching its flagship RDNA4 workstation GPU for the DIY market.
The Radeon AI PRO R9700 is not a gaming card. So far, every
It has been a busy period for external GPUs, where the main focus was on Thunderbolt 5 support after a long wait since its announcement. Practical use for Thunderbolt 5 is still limited to newer desktops and mini-PCs, but the same can be said about OCuLink. This direct PCIe connection standard has been adopted by many Chinese mini-PC and handheld makers, and it appears to be a good alternative.
ASRock has introduced a new Radeon RX 9060 XT graphics card under the Challenger series called the Standard Edition . Despite not carrying the OC label, the card still ships with factory-overclocked clocks compared to AMD’s reference specifications.
The Standard Edition (RX9060XT CL 16G) is listed with:
While GeForce RTX 50 laptop pricing has reached a high point and RTX 5090 specs remain out of reach for most buyers, there is an alternative that requires some courage and ideally a system that is already out of warranty. This refers to shunt modding, but applied in a slightly different context.
RTX 5090 laptops now sell starting at around $3500, which is a steep price for a
AMD naming gets new level of confusion.
AMD’s product naming is now even more confusing. After the company added “AI” to the Ryzen branding, the lineup became harder to follow. Following the launch of the Ryzen 8000 series, the expected next step for laptops would have been the Ryzen 9000 series. However, AMD reserved the 9
It’s a perfect size to be a drink coaster.
Yesterday when I was looking for materials for my post about the RTX 5090 Founder Edition story, I came across something cool. I’ve been long considering buying a 3D printer, and perhaps this project might have sealed the deal for me.
As we remember, the RTX 5090/5080 cards use a three-board design, and the
The company is set to launch 6, 8 and 16-core variants.
AMD clearly sees the advantage of X3D technology, and so do gamers. The company is reportedly preparing two high-end Ryzen 9000X3D CPUs with either 16 or 8 cores, higher clocks, and possibly dual 3D V-Cache variants. That is as far as current rumors go.
We also have semi-official confirmation of a Ryzen
NVIDIA begins preparing Nova driver for next-gen GPUs
NVIDIA has started posting patches for its open-source Nova kernel graphics driver that reference support for ‘next-gen’ GPU architectures. A new patch series submitted this week mentions changes required for “next-gen GPUs,” which refers to post-Blackwell products. The next architecture expected after Blackwell is Rubin, followed later by Feynman.
One patch titled “gpu: nova: add boot42
