Power used for data movement is power that isn’t used to feed GPUs, and with demands for the absolute maximum compute density and efficiency in every rack these days, a savings of that magnitude is a huge deal, so it's no surprise that Meta is hopping on board as it continues to expand its footprint.
Beyond hardware, Nvidia will offer its considerable in-house expertise in designing AI models to Meta’s engineers to help the company tune and boost the performance of its own core AI applications.
All of this just goes to show that as the AI revolution continues, Nvidia’s reach into tech far beyond gaming graphics cards is so extensive that people are likely to use software powered or shaped by its models and accelerators, whether they realize it or not, and that’ll only grow more true by the day as Meta expands its use of Nvidia's platforms and tech.
Follow Tom's Hardware on Google News , or add us as a preferred source , to get our latest news, analysis, & reviews in your feeds.
As the Senior Analyst, Graphics at Tom's Hardware, Jeff Kampman covers everything to do with GPUs, gaming performance, and more. From integrated graphics processors to discrete graphics cards to the hyperscale installations powering our AI future, if it's got a GPU in it, Jeff is on it.\u00a0 ","collapsible":{"enabled":true,"maxHeight":250,"readMoreText":"Read more","readLessText":"Read less"}}), "https://slice.vanilla.futurecdn.net/13-4-16/js/authorBio.js"); } else { console.error('%c FTE ','background: #9306F9; color: #ffffff','no lazy slice hydration function available'); } Jeffrey Kampman Senior Analyst, Graphics As the Senior Analyst, Graphics at Tom's Hardware, Jeff Kampman covers everything to do with GPUs, gaming performance, and more. From integrated graphics processors to discrete graphics cards to the hyperscale installations powering our AI future, if it's got a GPU in it, Jeff is on it.
bit_user It looks like ARM CPUs were the straw that broke the camel's back. AMD should've kept going with its ARM ambitions. Maybe they didn't have the resources to complete the K12, but they should've made successors when their financial picture started to improve. I had thought their Sound Wave SoC would feature Zen-based ARM cores, but rumors indicate they just had IP core licensed from ARM. I guess we'll have to wait a bit longer to see how well the Zen cores can compete, without being bogged down by the legacy of x86. Reply
thestryker If Neoverse cores were good enough to enhance their workloads they could have had this at pretty much any time. Seems like they just didn't want to do it themselves which they absolutely have the scale to have done. Unfortunate that with all the acquisition stuff going on Ampere wouldn't be an option as this seems like the type of market they were targeting. Reply
bit_user thestryker said: If Neoverse cores were good enough to enhance their workloads they could have had this at pretty much any time. All of the other Hyperscalers did. Microsoft, Amazon, and Google. thestryker said: Seems like they just didn't want to do it themselves which they absolutely have the scale to have done. Maybe Meta just thought it was extending themselves to far to build their own CPUs, when they were already struggling to build their own AI accelerators? So far, the only off-the-shelf ARM server chip has been Ampere, but they haven't really been competitive for like 4-5 years. thestryker said: Unfortunate that with all the acquisition stuff going on Ampere wouldn't be an option as this seems like the type of market they were targeting. AmpereOne just isn't very competitive. I think it missed its market window. It's very hard to beat the big boys at their own game. Reply
thestryker bit_user said: AmpereOne just isn't very competitive. I think it missed its market window. It's very hard to beat the big boys at their own game. The acquisition situation seemed to ensure it was never going to happen. It never seemed to have a real launch despite existing in 2024. Reply
bit_user thestryker said: The acquisition situation seemed to ensure it was never going to happen. It never seemed to have a real launch despite existing in 2024. No, Phoronix benchmarked it. It simply wasn't competitive enough. That's why you didn't hear much about it. https://www.phoronix.com/review/ampereone-a192-32x They even followed through with launching AmpereOne M, although I don't know where else it surfaced besides Oracle. https://www.phoronix.com/news/AmpereOne-M-Oracle-Cloud-A4 Post-acquisition, it seems they're not quite dead, yet: https://www.phoronix.com/news/LLVM-Clang-Ampere1C Reply
bit_user I think RAM could be playing a big role in this, in two ways. First, Nvidia is somewhat well-known for buying up capacity in advance. However many Grace CPUs they had secured wafers for, they probably bought up enough LPDDR5X to populate them. Because it's a fixed amount per CPU, that's an easy calculation to make. Having made these DRAM purchases in advance could be making them more cost-competitive, right now. The second point is about energy-efficiency. Here's where LPDDR5X has a huge advantage over DDR5 DIMMs. Given how much efficiency was highlighted in this article, it seems like that might be a factor at play. As for the relatively small core & thread-count, NVLink can mitigate that, by scaling up servers well beyond the typical dual sockets you tend to find in mainstream cloud compute. I'd be really interested in knowing just how many Grace CPUs per machine Meta is deploying. That could go some ways toward further improving the value proposition. Reply
TerryLaze bit_user said: It looks like ARM CPUs were the straw that broke the camel's back. AMD should've kept going with its ARM ambitions. Maybe they didn't have the resources to complete the K12, but they should've made successors when their financial picture started to improve. I had thought their Sound Wave SoC would feature Zen-based ARM cores, but rumors indicate they just had IP core licensed from ARM. I guess we'll have to wait a bit longer to see how well the Zen cores can compete, without being bogged down by the legacy of x86. Does nvidia have any arm core only product?!? My understanding was that the arm cores are just there to feed the GPUs and that the GPUs are the main attraction. AMD would have to get their GPU business up to snuff first. Reply
thestryker TerryLaze said: Does nvidia have any arm core only product?!? https://www.nvidia.com/en-us/data-center/grace-cpu-superchip/ scroll down to: NVIDIA Grace CPU C1 Reply
bit_user TerryLaze said: Does nvidia have any arm core only product?!? My understanding was that the arm cores are just there to feed the GPUs and that the GPUs are the main attraction. The main point of the article is that they're deploying machines containing only Grace CPUs and no GPUs. But yes, Grace was ostensibly created with the primary purpose of supporting Hopper & Blackwell GPUs. That's why it's noteworthy they're being deployed without either. Reply
Key considerations
- Investor positioning can change fast
- Volatility remains possible near catalysts
- Macro rates and liquidity can dominate flows
Reference reading
- https://www.tomshardware.com/pc-components/cpus/SPONSORED_LINK_URL
- https://www.tomshardware.com/pc-components/cpus/meta-will-deploy-standalone-nvidia-grace-cpus-in-production-with-vera-to-follow-company-sees-perf-per-watt-improvements-of-up-to-2x-in-some-cpu-workloads#main
- https://www.tomshardware.com
- India Fuels Its AI Mission With NVIDIA
- Over 80% of companies report no productivity gains from AI so far despite billions in investment, survey suggests — 6,000 executives also reveal 1/3 of leaders
- Save $110 on this Asus ROG Xbox Ally handheld, back at a record-low price — just $489.99 for this entry-level Steam Deck rival with 7-inch touchscreen, AMD Ryze
- AMD denies report of MI455X delays as Nvidia VR200 systems are rumored to arrive early — company says Helios systems 'on target for 2H 2026'
- Turtle Beach's rotating button Stealth Pivot wireless controller falls to just $99.99 — this hall effect gamepad hits its lowest ever price
Informational only. No financial advice. Do your own research.