Monday, May 27th 2024
NVIDIA's Arm-based AI PC Processor Could Leverage Arm Cortex X5 CPU Cores and Blackwell Graphics
Last week, we got confirmation from the highest levels of Dell and NVIDIA that the latter is making a client PC processor for the Windows on Arm (WoA) AI PC ecosystem that only has one player in it currently, Qualcomm. Michael Dell hinted that this NVIDIA AI PC processor would be ready in 2025. Since then, speculation has been rife about the various IP blocks NVIDIA could use in the development of this chip, the two key areas of debate have been the CPU cores and the process node.
Given that NVIDIA is gunning toward a 2025 launch of its AI PC processor, the company could implement reference Arm IP CPU cores, such as the Arm Cortex X5 "Blackhawk," and not venture out toward developing its own CPU cores on the Arm machine architecture, unlike Apple. Depending on how the market recieves its chips, NVIDIA could eventually develop its own cores. Next up, the company could use the most advanced 3 nm-class foundry node available in 2025 for its chip, such as the TSMC N3P. Given that even Apple and Qualcomm will build their contemporary notebook chips on this node, it would be a logical choice of node for NVIDIA. Then there's graphics and AI acceleration hardware.
It's being speculated that NVIDIA's AI PC processor could integrate an iGPU based on the "Blackwell" graphics architecture. This would give the chip first-rate graphics acceleration capabilities, including for DirectX 12 Ultimate and Vulkan. As for AI acceleration—the whole point behind this chip, we predict that the iGPU with its tensor cores would double up as the AI coprocessor. This would free up silicon real-estate for NVIDIA, the company can simply expose the GPU as an NPU virtual device to Windows, and use a series of translation layers to TensorRT. Given that Microsoft's requirement for an NPU capable of accelerating a local session of Copilot+ is just 45 AI TOPS, this should be easily done for NVIDIA. Lastly, 2025 will see LPDDR6 hit mainstream for mobile applications, such as smartphones, tablets and ultraportable PCs. It's likely that NVIDIA will implement this memory standard.
Sources:
XpeaGPU (Twitter), VideoCardz
Given that NVIDIA is gunning toward a 2025 launch of its AI PC processor, the company could implement reference Arm IP CPU cores, such as the Arm Cortex X5 "Blackhawk," and not venture out toward developing its own CPU cores on the Arm machine architecture, unlike Apple. Depending on how the market recieves its chips, NVIDIA could eventually develop its own cores. Next up, the company could use the most advanced 3 nm-class foundry node available in 2025 for its chip, such as the TSMC N3P. Given that even Apple and Qualcomm will build their contemporary notebook chips on this node, it would be a logical choice of node for NVIDIA. Then there's graphics and AI acceleration hardware.
25 Comments on NVIDIA's Arm-based AI PC Processor Could Leverage Arm Cortex X5 CPU Cores and Blackwell Graphics
Depending on how beefy it shall be, I wonder if (LP)DDR6 will be a limiting factor, or how they plan to get around this (fat L3 cache? HBM even?).
ARM A78C and Ampere-based GPU (with some bits from Ada)
We now have five potential strong CPU and GPU players in the laptop and desktop market: Apple, Qualcomm, Nvidia, Intel and AMD. Mediatek is there too but I don't think we will see much of them. Still only two (Intel and AMD) are available for the CPU DIY market. Of course we have three in the DIY GPU market (Nvidia, AMD and Intel).
If ARM does eventually surmount X86 in the PC market Intel will have no one to blame but themselves.
pretty certain all those chip design companies are working hard to get risc v on par with their arm offerings so they can move away from them and their licensing and this will happen before ram pushed out x86 from the pc space.
That doesn't mean they couldn't make something worthwhile and cheap...only that it might take them a little bit longer (a couple years?), which makes sense given WoA will take a bit to get off the ground.**laughs at the history of Tegra**
---------
FWIW, where's Thor (other than auto) that was supposed to be in the new Tegra? I thought they cancelled a <8nm design that would have made Nintendo's next console suck less by focusing on that.
I could imagine this will use some version of that (but again, just a hunch), although the current iteration of Neoverse only supports DDR5.
I think jumping to the conclusion it will use DDR6 is not one that should be assumed.
Overall I'd say it's been a bit all over the place for Tegra, but from my perspective when they've been able to leverage the GPU they shine.
Nintendo bought that chip for cheap after nobody wanted to use it for phones/tablets because it drew too much power (almost certainly because of the GPU).
It sat around until Nintendo took it off their hands (and by that time was outdated).
Nintendo can make literally anything work.
I would not call the success of Switch anything to do with the help of nVIDIA other than likely selling them the chip (which was a failure in intention) for cheap, which they down-clocked a whole bunch to make work.
Look, listen, I'm not saying it won't be a good design (because it probably will be), but just let's not romanticize the history of Tegra and that power envelope, especially that chip.
Without the bloat of X86, the future is bright. Considering nVIDIA now designs from the bottom up, not the top down, meaning mobile (and efficiency) first, there is indeed hope for some very nice solutions.
Look man, I'm not trying to argue! I just think using Tegra's current and past (undesirable) adoption in mobile computing is not what should be used to hype expectations. :toast:
Soundwave rumor
www.extremetech.com/computing/184323-intel-stuck-with-1-45-billion-fine-in-europe-for-unfair-and-damaging-practices-against-amd#:~:text=The%20EU%20found%2C%20in%20part%3A%20That%20Intel%20paid,AMD%20hardware%2C%20including%20Dell%2C%20Acer%2C%20Lenovo%2C%20and%20NEC.
You can find some if their systems with AMD cpus, but the lucrative business lines (Latitude, Optiplex and Precision) are strictly limited to intel.
Yes, they have some if those systems advertised with AMD but good luck getting them. I remember an Optiplex with an AMD cpu that was only available via phone.
About Ngreedia’s push, I would not be surprised of a similar arrangement.
Nvidia: 'We think this will be a terrible move for the gaming industry and will oppose it'.
Microsoft: 'We have discussed with nVIDIA the willingness for all MS and ABK titles to be available on Geforce Now.'
(Literally the next day)
Nvidia: 'We see no problems with this merger'.
-----
IDK man...Some people just don't get it. The people that do, and still support them...That's fine, if shitty.
But, to me, it's somewhere between funny and extremely sad bc to think those things don't happen is ignorant, and to know they do but ignore that reality is probably even worse.
In that sense, this partnership seems destined....you have one party who loves anti-competitive practices and using financial power to do so (Nvidia) and you have another party who will gladly engage in such anti-competitive practices so long as the price is right (Dell)...seems like a perfect match.
greatwoeful battery life o_O