Reddit user DallasGrave, who recently soldered a dual 8-pin board to an RTX 5090's 16-pin header, is back with a more ambitious take on the same idea. The new "Quad 2x8 Header Conversion" mounts four 8-pin PCIe connectors right on the card, tapping the contact points of the 12V-2x6 header, so the 16-pin connector can either stay in use alongside it or be skipped entirely. Each 8-pin carries three 12 V conductors, which spreads the load over twelve wires instead of six.

DLSS Multi-Frame Generation and DLSS 5 Neural Rendering can now run on hardware never intended for it, at least according to Nvidia. Now, two of the modders foundational to the release of these community advancements speak out.

Taiwanese outlet BenchLife has added more detail to last week's report that NVIDIA is pulling GB202 silicon away from the GeForce RTX 5090. According to BenchLife, board partners are no longer receiving GB202 chips for RTX 5090 production. The same applies to the China-specific RTX 5090 D and RTX 5090 D V2 variants. NVIDIA hasn't said anything officially yet. Apparently, the company doesn't see this as an end-of-life (EOL) move either, since the chips are simply being shifted to more profitable GB202-based RTX PRO cards like the RTX PRO 6000 and the recently unveiled RTX PRO 5500. In the meantime, any RTX 5090 cards still reaching shelves are built from chips partners already have. The card immediately sold out at Walmart for $4,299 last month, so buyers shouldn't expect prices to come down anytime soon.

Everyone points at someone else.

NVIDIA has raised the price of the Shield TV Pro from $199.99 to $299.99, effective October 2. The streaming box launched in October 2019 at $199.99 and held that price for almost seven years. Earlier this year, NVIDIA promised to keep updating the Shield TV, with hardware engineering SVP Andrew Bell saying the company had no plans to abandon it anytime soon. NVIDIA has also stopped making the standard, tube-shaped Shield TV, which had reportedly been out of stock for months. That leaves the Pro as the only Shield model in production. NVIDIA confirmed both changes in a statement.

Games could be quicker to make and run better when they go on sale.

Brazilian YouTube group TecLab has bypassed the RTX 5090's notorious melting problem in the most direct way possible. The group removed the 16-pin connector from the equation entirely and soldered power wires straight onto the PCB. The demonstration used a Galax RTX 5090D HOF OC LAB XOC, an extreme-overclocking variant with two 16-pin inputs and a custom BIOS rated for up to 2,000W, far above the 575W TDP of a standard RTX 5090. TecLab framed the mod as a sarcastic jab at NVIDIA, saying that enthusiasts are tired of hearing about connectors melting and that they wanted to show how it should have been done from the start.

Even when you think you're not buying Nvidia, you might still be buying Nvidia

A newly disclosed Rowhammer attack called GPUThor can bypass error-correcting code (ECC) protections on NVIDIA GPUs, enabling denial-of-service (DoS) and root-level privilege escalation.

Kyiv wants tighter controls to keep foreign silicon out of Moscow's weapons

Do more than just game with your GeForce NOW subscription.

What about the risks of an AI infrastructure bubble?

The primary focus of this release seems to be adding support for two major Xbox games alongside the new extraction game Mistfall Hunter.

NVIDIA's next generation of graphics cards, under the "Blackwell-Next" name, began their integration into the Linux kernel some time ago. The latest set of patches now includes updates for handling Compute Express Link (CXL). This particular patch is intriguing because it refers to the next generation of NVIDIA GPUs with the "Blackwell-Next" architecture, even though NVIDIA has publicly confirmed the codenames for its upcoming GPU architectures multiple times. From the "Blackwell" generation, we have seen another variant called "Blackwell Ultra" in the B300 series. However, the next-generation architecture has been publicly confirmed to be called "Rubin." Therefore, it is unclear why NVIDIA's engineers are referring to the next generation GPU design as "Blackwell-Next," given that the codenames have been public for some time.

Jensen can't risk semiconductor supply chains derailing the AI hype train

Chinese cloud firms are reportedly testing more than 300 Vera servers overseas.

cross-posted from: https://piefed.ca/c/fuck_ai/p/759175/nvidia-s-hostile-takeover

• Invidious
• Youtube

"But ultimately our customers, our partners don't want to buy computers, they want to build AI factories. All of the CPUs until now were created for people. We were the users, we were the users. We were the renters. This is CPU for the agents (AI/LLMs). All the CPUs in the past we built for humans. This CPU is built for agents. And these agents are going to be using the CPUs with very little patience. There'll be a lot more agents then there are people. And then the agents are very impatient."

• Jensen Huang; CEO of NVIDIA, Computex or GTC 2026.

Nvidia's new chips will power laptop workstations and mini desktop PCs at first.