The Hot Take: More GPU security issues, #FUN.
Two new Rowhammer attacks for GPUs have been discovered that can cause bit flips in VRAM to gain arbitrary read/write access over it. These attacks target page files and the page directory that are otherwise protected from electrical disturbance by the driver. By "massaging" these data structures into vulnerable regions where a bit flip can occur, the attacker can access even the CPU memory.
Read the full article
The Hot Take: Google, Nvidia and not Intel all the suddenly make this amazing new tech at around the same time? Not buying it.
Intel is advancing texture compression techniques with its newly introduced Texture Set Neural Compression (TSNC) technology, a neural network-based approach designed to significantly reduce the size of texture assets used in modern graphics workloads.
Read the full article
The Hot Take: Following intels steps on the Arc? Also, how much space are the pre-compiled shaders going to consume of diskspace?
NVIDIA has introduced a new beta feature called Auto Shader Compilation, or ASC, through the latest NVIDIA App update, and it targets a familiar pain point in modern PC gaming: long initial loading phases and shader compilation stutter in DirectX 12 titles.
Read the full article
The Hot Take: Let's milk the architecture untit the pleebs scream, beg and plead for a new architecture... All while ringing out as much cash from the Ai market......
Thereās an easier way: A manufacturer could simply release the most expensive gaming graphics card in the series, and the market would eventually settle down. For NVIDIA, however, that moment seems to be a long time coming. Since early February, reports have been circulating that an even more powerful Blackwell modelāpositioned above the GeForce RTX [ā¦]
Source
Read the full article
The Hot Take: I'm glad, but will be actually be able to afford or get any in our hands? Also what games are we going to need this for, as game releases have definitely stagnated along with the market.
NVIDIA's GeForce RTX 60 Series GPUs will be powered by the Rubin architecture, which exists only for AI and data center use thus far. The Rubin CPX, for example, is built around NVIDIA's GR212 chips, but new information shared with YouTuber RedGamingTech claims the RTX 60 Series chips will be the GR202 (RTX 6090), GR203 (RTX 6080), and GR205
Read the full article
The Hot Take: We need more competition, AMD seems to be very quiet lately and might come out of no where with a beast but they haven't yet. So intel coming back in even to do an Ai bubble grab it still helps us all. Especially when that bubble pops.
Intel's Arc Pro B70 is designed to offer accessible local inference for AI users, delivering more memory at half the price of the competition. Intel Arc Pro B70 vs NVIDIA RTX PRO 4000 Blackwell: 32 GB vs 24 GB, $949 vs $1800, More AI Context, 2x Tokens Per Dollar So we talked about the unveiling of the Intel Arc Pro B70 graphics card in our other post, where we highlighted the specifications, availability, and prices of the product. The B70 is going to be the flagship Pro & AI product from Intel within its Arc Pro stack, and they have [ā¦]Read full article at https://wccftech.com/intel-arc-pro-b70-outclasses-nvidia-rtx-pro-4000-in-ai-at-half-the-cost/
Read the full article
The Hot Take: Nvidia starting to feel the heat of competition and see those $ evaporate as they try other vendors.
Nvidia is preparing to launch a new chip designed to speed up AI responses, breaking with its long-running habit of flogging the same processor for every job.
Nvidia chief executive Jensen Huang is expected to unveil a chip focused on āinferenceā, meaning running models rather than training them.
According to people familiar with the plans for GTC next week, the chip is the first new product to emerge from Decemberās $20bn deal to hire the founders of Groq, a start-up building ālanguage processing unitsā tuned for high-speed answers to complex AI queries.
Three months after that deal, Nvidia is expected to debut a Groq-based LPU to sit alongside its forthcoming flagship Vera Rubin graphics processing unit. It is part of a product family meant to head off challengers and meet new kinds of AI applications.
The move lands as the worldās most valuable company gets grief from start-ups and customers, such as Google, all busy cooking up their own AI chips. This week, Meta announced a new family of four inference-focused processors.
One Silicon Valley venture investor said: āWe are entering an interesting phase that is not āNvidia dominantā,ā
For the past three years, Nvidiaās $4.5tn market capitalisation has been built on its GPUs, which have become the backbone of generative AI. They train models such as the ones behind OpenAIās ChatGPT.
Huang has insisted that a single system can handle training and then run the chatbots and coding tools built on top. Big Tech has spent hundreds of billions deploying these boxes while funding their own specialised silicon.
But the growing sophistication of AI tools, including āagenticā coding systems, is pushing Huang to ditch the mantra that one GPU fits every workload.
The Groq deal was worth about $20bn, according to people familiar with the transaction, making it one of the biggest deals in Nvidiaās 33-year history. It includes licensing and the hiring of key talent, including Groq founder and former Google chip executive Jonathan Ross.
Groq, which had been working with Samsung to manufacture its products, previously bragged that its LPUs were faster and more efficient than Nvidiaās GPUs for inference. Nvidia clearly listened.
Nvidiaās flagship Blackwell and Rubin systems lean on high-bandwidth memory to cope with the massive data loads that AI models fling around. But HBM is expensive and in increasingly short supply as SK Hynix and Micron struggle to keep up with demand.
The Groq-style chip will use SRam rather than the dynamic Ram used for HBM, according to people familiar with Nvidiaās plans, because SRam is more available and better suited to speeding up AI āreasoningā tasks.
Bank of America reckons that by 2030, inference will account for 75 per cent of AI data centre spending, up from about 50 per cent last year, and it expects a ābroadened AI portfolioā at GTC.
Ā
Read the full article