Arthur Hanson
Well-known member
Just out on CNBC that Nvidia and AMD are going after Intel's market in PC chips. Any thoughts or comments appreciated. This should be additional business for TSM.
NVIDIA, AMD planning on building ARM based pc chips
- Thread starter Arthur Hanson
- Start date
Originally it was supposed to be QCOM that built ARM PC chips but it seems like they dropped the ball creating this opening for NVDA and AMD.
I think AMD is interesting, in that they are willing to pivot towards building an ARM PC chip even if it competes with it's x86 business.
This is really great news for ARM, and not so great news for Intel.
I think AMD is interesting, in that they are willing to pivot towards building an ARM PC chip even if it competes with it's x86 business.
This is really great news for ARM, and not so great news for Intel.
hist78
Well-known member
Exclusive-Nvidia to make Arm-based PC chips in major new challenge to Intel
(Reuters) -Nvidia dominates the market for artificial intelligence computing chips. Now it is coming after Intel’s longtime stronghold of personal computers. Nvidia has quietly begun designing central processing units (CPUs) that would run Microsoft’s Windows operating system and use...
delong.height
Active member
how can AMD pay fee to ARM when ARM perform inferior to AMD, and AMD can do this on their own with their zen architecture. What's the reason to inspire new competitors (like Nvidia, Qualcomm, Mediatek) into this market when AMD and Intel are the dual participants of the PC market? If they really need energy efficient, they can opt in zen 4c for energy efficient, not need for ARM.
And speaking of Nvidia, it's even more laughable. Just watched how they try to shift all budget to AI/Data center now. The question is will they have enough resources for ARM PC when you have a volume silicon competitor like Intel and historical legacy of x86.
Are you sure ARM cost/performance is worse than x86? Apple has already proven that ARM can have better cost/performance.
Great news for TSMC as well but we have heard this all before. I'm looking forward to the first Arm investor call. The Q&A should be interesting.
Do we really know anything about the cost of Apple Silicon?
This is a very complicated topic. AMD and Nvidia will use their current merchant CPU vendor model, which has several inherently higher costs than Apple's proprietary CPU model. Some key differences:
1. AMD and Nvidia (and Intel) have large and very expensive product management, product marketing, and sales organizations. Apple does not have to research and accommodate other PC designers' requirements, or sell their CPUs to computer systems vendors. These factors lower Apple's M-series CPU costs dramatically, IMO.
2. Apple can target their CPUs to their specific products in only Apple's target markets (which are narrower than the general client CPU market), rather than the sum of the target computer systems vendors' markets. Intel, AMD, and Nvidia could do this, but it's not clear to me that narrowing is practical without a completely vertical architecture stack. This targeting allows Apple to use in-package memory, just to name one example, rather than DIMMs. This makes Apple's performance higher in some parameters, its board costs lower, power consumption is lower, system packaging can be simpler, and cooling more effective. AMD and Nvidia could copy some of these concepts, or conceive analogous ones, but that will either limit their markets or create many more SKUs.
3. Apple has its own OS. I doubt AMD and Nvidia will stray from Windows, though I suppose Linux is a possibility. Owning the OS allows further narrowing / targeting of CPU features. While x86 CPUs do share an instruction set, there are numerous features / designs outside of the instruction set that are different between AMD and Intel. (Virtualization and Networking are two that come to mind.) Microsoft and Intel/AMD do co-development, but it is much easier and more efficient when there is one management structure. I doubt Rosetta 2 would have been done if Apple didn't own the entire stack.
4. Apple has an Arm architecture license, and essentially designs its own proprietary CPUs that only have Arm's instruction set in common. Apple gets a lot of credit for the M-series performance and efficiency, not just Arm.
5. Client CPUs are more complex than most server CPUs, because they need additional hardware acceleration units, graphics support, and have much more restrictive power limitations. One could jokingly ask what Nvidia knows about low power.
So in the end I'm on the fence. I think there is an opportunity for an M-series like CPU in the Windows client space, or even a Linux-based client, but I'm having trouble seeing AMD or Nvidia stepping up to that much risk. Especially AMD, when they already have substantial client share with high margins.
Last edited:
I think it'll be interesting. I would agree is no inherent performance advantage of the ARM ISA. However I think it allows for changes in business model that are not possible in x86 ecosystem, and it's the business model differences that allow for lower cost.
This is also driven by Microsoft, who may want to to do this for marketing/market segmentation reasons.
This is also driven by Microsoft, who may want to to do this for marketing/market segmentation reasons.
hist78
Well-known member
Microsoft wants to sell more various Windows OS and related products and services. Microsoft is essentially derisking on Intel/x86 by diversifying to AMD, Nvidia, Qualcomm, and ARM. Intel's major problem is on its products, not on its foundry. All the major Intel competitors are fabless without a fab. Those fabless companies are eating into those market traditionally dominated by Intel while Intel can't take market share quick enough (if any) from those fabless companies.
I can see Intel will start using more outside foundries in order to push more competitive products to the market faster. Otherwise Intel is missing market opportunities (mobile, GPU, AI, etc.) one after another.
Qualcomm unveils new PC laptop chip with AI features for 2024
Qualcomm on Tuesday gave details about a chip for Microsoft Windows-based laptops that it claims will be faster at some tasks than Apple's chips for Mac computers. Qualcomm executives said the company's new Snapdragon Elite X chip will be available in laptops starting next year and has been...
Qualcomm's initial marketing is t-e-r-r-i-b-l-e. No specs, no factual information; the highlight is:Qualcomm unveils new PC laptop chip with AI features for 2024
Qualcomm on Tuesday gave details about a chip for Microsoft Windows-based laptops that it claims will be faster at some tasks than Apple's chips for Mac computers. Qualcomm executives said the company's new Snapdragon Elite X chip will be available in laptops starting next year and has been...www.yahoo.com
Introducing an all-new naming convention for our next-generation of intelligent PC platforms: The Snapdragon X Series
Snapdragon X compute platforms will deliver next-level performance, AI, connectivity and battery life, building on our years of experience engineering heterogeneous compute architectures across the CPU, GPU and NPU.www.qualcomm.com
New logos and platform badges... those are things to look for in a mobile computing platform?
"Faster than Apple's M2 at some tasks..." Too bad Apple is about announce the M3.
Qualcomm should be better than this.
delong.height
Active member
I didn't say anything about this. I said that AMD is a big chip designer that has its own architecture, which in some way has its competitive advantage compared to ARM. It's certainly not persuasive enough given Microsoft hasn't pivot an all-in Windows on Arm strategy. Does it make sense to you that AMD should persue some of these WoA market that are highly overlapped with existing x86 market dominated by Intel and AMD already?
Yes that’s why it’s an interesting decision on the part of AMD.
My guess is AMD starting to see themselves as a chip design company rather than x86 company.
Microsoft made similar trade offs themselves but supporting Linux on Azure and its proven to be the right decision.
Speedfriend
Member
There were plenty of benchmark numbers in the presentations, obviously, we can take these with a pinch of salt. But claiming that it has 15% better single-thread performance at 30% lower power than an M2 Max means it is very much in the M3 performance envelope. Even if it gets close to M"/M3 on performance per watt, that would be outstanding for the Windows ecosystem
On their live presentation I kept waiting for them to come alive. That was an almost amateur level of presentation, dull, low energy, weird alternation between flat boring graphs and tinsel concept videos flashing through artist projections too fast to grasp. It had a nervous energy but not the good kind. More like the nervous dread. It was weird.
I do believe Oryon will be a good chip and far better at energy management than AMD, never mind the Nvidia team with no mobile track record. I would bet on QC for mobile, absolutely. But that weird vibe makes me wonder ...
One could jokingly ask what Nvidia knows about low power.
Tegras did relatively well, though poor power efficiency because of VLIW inside was well known.
AAPL
New member
Ignorant question here but what's the status of Microsoft's 'Windows x86 emulation on ARM'? What Apple did with Rosetta 2 for running Intel compiled apps was simply amazing and won't the same be absolutely necessary for Nvidia, AMD and Qualcomm to be successful with ARM PC chips?
Well, it does exist and is fairly complete these days. https://learn.microsoft.com/en-us/windows/arm/apps-on-arm-x86-emulation
Whether it matches Rosetta will likely depend on what special case support is built into the ARM variants used for Win11. There are some issues of awkward differences in handling flags, as well as some deep differences in how locks work, which affect performand after cross compiling. If the hardware adds a few extended instructions (and ARM has supported some, though I forget the details) then the risk of critical code having weird overheads is much reduced.