Question What will the CPU industry marketshare look like in the future? The Rise of ARM?

[Saylick]

I'd expect it to use the same DLA as the Tegra family.

Yep. Basically this. https://developer.nvidia.com/deep-learning-accelerator



If the DLA can't run it, there's the GPU to fall back on, but the DLA will be much faster in general, presumably because it's simply a much bigger tensor core.

 

[soresu]

You also need Qualcomm to have good drivers.

Precisely, I'd be interested to know if their DX12 driver even matches TURNIP for optimisation.

 

[StinkyPinky]

Semiaccurate also recently put out two articles.

Microsoft Absolutely Screws Intel and AMD Over AI PCs

It looks like Microsoft is turning on Intel and AMD over AI PCs.

semiaccurate.com

Microsoft Throws Qualcomm Under The Bus Too

Microsoft seems to be on a roll of late stabbing all of their ‘partners’ in the back.

semiaccurate.com

The TLDR is that Microsoft threw Intel and AMD under the bus, to promote their valued partner Qualcomm's ARM-based Snapdragon for their recent Copilot+ PC debut. But apparently Microsoft has already thrown Qualcomm under the bus too, to welcome a new valued partner on board:

NVIDIA.

MS can do what they want, they have all these companies over a barrel. Apple have their own hardware, and it's forever the "year of linux" so if these guys want to sell chips they need to work with Microsoft.

Once again proving the real money is in software and services.

 

[Mahboi]

My guess is that ARM will have a striking entrance in the market, climb higher and harder than anyone for 2-3 years, then slowly simmer down back to nothing.

 

[FlameTail]

My guess is that ARM will have a striking entrance in the market, climb higher and harder than anyone for 2-3 years, then slowly simmer down back to nothing.

That doesn't even make sense. Why would it?

 

[FlameTail]

Coming back to the topic of the post, who do you think will have the most success in the ARM PC World? (In terms if marketshare)

Qualcomm vs Nvidia vs Mediatek vs AMD vs Samsung vs Huawei

Edit: Added Huawei

 

[gdansk]

Coming back to the topic of the post, who do you think will have the most success in the ARM PC World? (In terms if marketshare)

Qualcomm vs Nvidia vs Mediatek vs AMD vs Samsung

Huawei.

 

[FlameTail]

Nvidia PC SoC leak:

https://twitter.com/x/status/1793469879334687122

TSMC N3P, not Intel 3- as some of you believed?

That last statement lines up with Semiaccurate's reporting;

Microsoft Throws Qualcomm Under The Bus Too

Microsoft seems to be on a roll of late stabbing all of their ‘partners’ in the back.

semiaccurate.com

 

[Mahboi]

That doesn't even make sense. Why would it?

Cause it'll be driven by obscene amounts of manufactured RISC Hype for a few years before reality sets in that it's not really better than x86 and has a hard time reaching their scale/high performance.
The ones who can do more can do less, including in power draw, and the ones who do less have a much harder road going up.
So I expect 1-3 years of hype driven market while x64 stays on top perf wise, before the reality sets in that ARM is just mediocre performance and roughly equal perf/battery life. And then they'll be seen as the poor man's option.

 

[moinmoin]

I think it's less about hypes but more about several markets converging and directly competing in the future.

1. At the starting point Intel with x86 was in a perfect bubble, with the x86 market still dominated by Intel and Intel's advantage in foundry tech still being undisputed. Even though the mobile market ensure an astronomic rise of pure play foundries.
2. Then came the revival of AMD, rivalling Intel's processor designs.
3. Then came AMD's move to TSMC for Zen 2, suddenly letting flood previously mobile foundry tech into the x86 market. Intel had no better response to that than to partake itself in ordering wafers from TSMC.
4. Then came Apple's move away from x86 to Apple Silicon, removing their products from the x86 market and showing that ARM based products are feasible for that audience as well.
5. This appears to have tickled Microsoft's pride which is since renewing its push for its own up to then halfhearted WoA initiative. That push may or may not have been a reason for Qualcomm to buy Apple staff offshot Nuvia.
6. The result may become a Windows PC market served not only with x86 by Intel and AMD but also ARM by Qualcomm, Arm through MediaTek and other standard core licensees, and possibly future custom core manufacturers like Huawei and Nvidia.

In a way that's an exciting future to look forward to since while introducing questionable hardware compatibility (still don't know how it's planed to solve that under WoA) it will make plenty different core designs competing with each other readily available in the mainstream market.

 

[FlameTail]

It looks Huawei is going to follow Qualcomm's steps and make PC class chips called "Kirin X"

Qualcomm's brand is Snapdragon, which became Snapdragon X. Huawei's brand is Kirin, and it's said to become Kirin X. Get it?

As you all know, Intel/Qualcomm's license to sell chips to Huawei was revoked recently. This means Huawei has no option but to resort to using their own in-house processors for their smartphones, and PCs.

 

[Mahboi]

@moinmoin but the end result of that timeline you described is that everyone competes in the same market, which is where again, my standpoint is "the ones with the biggest/heaviest are the most ahead". That's still x86 and particularly AMD.

I think marketing hype is going to be the main driving force for people's interest in ARM, but after awhile, everyone will just buy what's best, and that'll still be x86 and still be AMD.
The only way it could go differently is if ARM somehow damages Intel's market share enough that they seriously go down (even more than they currently are), and then we might see a severe rebalancing, but again, I think it'll be in favour of AMD in x86Land.

We'll see very soon, Zen 5 vs Snapdragon X Elite is going to be a live test of all the ARM hype vs all the x86 glory.

 

[moinmoin]

@Mahboi I don't disagree, in tech logically at the current rate (that is Zen 5 keeping all the promises) AMD with x86 should be the clear winner even with all the added competition.

But we are mainly talking about the mainstream laptop market, which is dominated by Intel up to now and AMD even with all its clearly superior tech had a rather hard time getting into. Microsoft (mostly through its Surface brand) is now heavily pushing ARM which will mix that up some more. Intel is bound to be the biggest loser from this, and while I don't expect ARM to endanger x86 it's likely to take a significant share.

 

[poke01]

I woudn't underestimate Nuvia. ARM themselves is meh but Nuvia shouldn't be underestimated. Their v2 is going to be good.

Also depends on where this AI PC is going..

 

[Mahboi]

But we are mainly talking about the mainstream laptop market, which is dominated by Intel up to now and AMD even with all its clearly superior tech had a rather hard time getting into. Microsoft (mostly through its Surface brand) is now heavily pushing ARM which will mix that up some more. Intel is bound to be the biggest loser from this, and while I don't expect ARM to endanger x86 it's likely to take a significant share.

Fully agree, but that's why I'm expecting a massive entrance and slow disappearance. MS in my eyes is a financial company, not a tech one. It putting all its weight behind ARM and AI means little in the long run, it'll last if the tech is good. For now, I'm very skeptical on Qualcomm's ability to really shine. Come very close, sure, but really take wins? We'll know in a few weeks.

I woudn't underestimate Nuvia. ARM themselves is meh but Nuvia shouldn't be underestimated. Their v2 is going to be good.

I'm all for good competition from ARM, but I feel like there's going to be a glass ceiling that x86 sits on top of.

 

[SpudLobby]

Nvidia PC SoC leak:

https://twitter.com/x/status/1793469879334687122

TSMC N3P, not Intel 3- as some of you believed?

That last statement lines up with Semiaccurate's reporting;

Microsoft Throws Qualcomm Under The Bus Too

Microsoft seems to be on a roll of late stabbing all of their ‘partners’ in the back.

semiaccurate.com

No, the Intel 3 part is a separate report. This seems more like the MediaTek collaboration. Either way we have reason to believe there are credible rumors for both.

 

[FlameTail]

I'm all for good competition from ARM, but I feel like there's going to be a glass ceiling that x86 sits on top of

And what is that glass ceiling? Are you implying that ARM processors are incapable of matching/exceeding x86 levels of performance?

Tell me, which CPU has the fastest single thread in the world right now?

(Hint: It's an ARM CPU).

 

[Mahboi]

And what is that glass ceiling? Are you implying that ARM processors are incapable of matching/exceeding x86 levels of performance?

In a word, weight.
x64 just has a very long history of beefy server CPUs, high performance high wattage DT CPUs, and all kinds of fat things.

ARM rose through phones. I just don't expect the gap between Xeon/EPYC makers and phone chipmakers to be as easy to cross as some expect. Especially since on the performance/complexity curve, ARM could more or less stop at "good for phones". Admittedly M1 changed the game, and Nuvia is meant to do the same thing again, but that's just promises for now, and frankly if on ARM Windows we end up with 95% of the perf of x86 with the same power draw, I won't be surprised at all.

Tell me, which CPU has the fastest single thread in the world right now?

Irrelevant. Zen 5 comes out in 2 weeks.

 

[FlameTail]

In a word, weight.
x64 just has a very long history of beefy server CPUs, high performance high wattage DT CPUs, and all kinds of fat things.

I will counter that with the fact that ARM also has more weight, in a different sense. Consider how many companies are backing ARM: Apple, Qualcomm, Nvidia, Mediatek, Samsung, Huawei, the hyperscalers (Google, Meta, Amazon, Microsoft).... They are all leveraging ARM CPUs in some form or the other. Meanwhile, the x86 camp only has Intel and AMD.

ARM rose through phones. I just don't expect the gap between Xeon/EPYC makers and phone chipmakers to be as easy to cross as some expect. Especially since on the performance/complexity curve, ARM could more or less stop at "good for phones". Admittedly M1 changed the game, and Nuvia is meant to do the same thing again, but that's just promises for now, and frankly

ARM already has a presence in servers. See Nvidia Grace, Ampere Computing, Amazon Graviton, Google Axion and Microsoft Cobalt.

ARM isn't just phones. It is everywhere.

if on ARM Windows we end up with 5% of the perf of x86 with the same power draw, I won't be surprised at all.

what.

Irrelevant. Zen 5 comes out in 2 weeks.

IT IS relevant. It is the reigning king. Besides, I don't think Zen5 mobile is going to top it. It will take Zen5 desktop to do so, and of course use more power to do so.

Some of your statements clearly show an AMD/x86 bias. There's nothing wrong with having biases, as we all do. But it is important to be open minded and not let ourselves be blinded from the facts.

 

[Mahboi]

I will counter that with the fact that ARM also has more weight, in a different sense. Consider how many companies are backing ARM:
Apple, Qualcomm, Nvidia, Mediatek


Samsung, Huawei, the hyperscalers (Google, Meta, Amazon, Microsoft).... They are all leveraging ARM CPUs in some form or the other. Meanwhile, the x86 camp only has Intel and AMD.

Ridiculous.
Apple has already seemingly lost most of its CPU people to Nuvia/Qualcomm.
Nvidia basically rides ARM just to hamper AMD. Their "Grace" CPUs are so unwanted and pointless that they basically have to force sell them in bundles with Hopper GPUs. Nobody cares about NVIDIA CPUs, they literally HAVE to buy them.

Mediatek is not exactly top tier silicon either.
Which leaves Qualcomm as the only big contender for now, and until we get the real benchmarks, it's all wild promises.

As for all the others, they're clients. You're shilling ARM's TAM, not ARM's quality. And I never meant financial weight in the first place, I meant the weight of effort/experience towards real top level stuff.

ARM already has a presence in servers. See Nvidia Grace, Ampere Computing, Amazon Graviton, Google Axion and Microsoft Cobalt.

ARM isn't just phones. It is everywhere.

Memes, all of them.
Graviton is the poster child for these chips: done in-house for the dozen or so services that AWS could either put on a Xeon/EPYC for full price, or design at home, buy at TSMC, and get 75% of an EPYC/Xeon's perf for 50% of the price. When you control the entire software/HW stack that the chip is meant to serve, it's just a smart call. None of those sell as high performance generic chips in the server market, except Ampere. The rest is all in-house-for-house, or a literal "we taped this awesome Grace CPU to your GPU and you can't buy the GPU without the CPU bruh".

what.

I must've deleted the 9 in 95%. 95% of x86 perf.

IT IS relevant. It is the reigning king. Besides, I don't think Zen5 mobile is going to top it. It will take Zen5 desktop to do so, and of course use more power to do so.

Some of your statements clearly show an AMD/x86 bias. There's nothing wrong with having biases, as we all do. But it is important to be open minded and not let ourselves be blinded from the facts.

The entirety of your statements are soaked with pro-ARM bias, so please don't meme yourself by calling out people's biases now.
And it is not relevant. Zen 5 comes out in 2 weeks and that's that. Apple's fancy fat cores won because they were fat and always a node ahead. If Zen 5, on N4P rather than N3E, comes even close to M4, that's it, it's GG for "ARM domination". The "Apple is ARM and Apple is Best thus ARM is best" is a ridiculous fallacy when you look at how wide, how low-frequency, and how expensive they are to make, on a more advanced node almost every time.

 

[eek2121]

I will counter that with the fact that ARM also has more weight, in a different sense. Consider how many companies are backing ARM: Apple, Qualcomm, Nvidia, Mediatek, Samsung, Huawei, the hyperscalers (Google, Meta, Amazon, Microsoft).... They are all leveraging ARM CPUs in some form or the other. Meanwhile, the x86 camp only has Intel and AMD.

ARM already has a presence in servers. See Nvidia Grace, Ampere Computing, Amazon Graviton, Google Axion and Microsoft Cobalt.

ARM isn't just phones. It is everywhere.

what.

IT IS relevant. It is the reigning king. Besides, I don't think Zen5 mobile is going to top it. It will take Zen5 desktop to do so, and of course use more power to do so.

Some of your statements clearly show an AMD/x86 bias. There's nothing wrong with having biases, as we all do. But it is important to be open minded and not let ourselves be blinded from the facts.

You are comparing 2024 uarch with 2022-2023 uarch.

Zen 5/Arrow Lake is the competition for the M4 and the snapdragon.

As for Qualcomm, I have seen enough that I (and others I have spoken with) am extremely skeptical of Qualcomm’s claims. My suspicion is perf/watt won’t match what has been claimed. The chip will be fast, sure, but it will struggle to compete with other offerings due out this year.

RISC based designs have come and gone before. x86 has survived.

I don’t think you realize that building chips that can perform as fast as Zen 4 or the M1 isn’t easy, and only a handful of engineers in the entire world are capable of doing it.

Microsoft has always chased things like they have ADHD. They will chase ARM and AI until the bubble pops. Once that happens, a market correction will occur and things will change.

I do need to clarify that I am not against ARM or RISC-V, I am just old enough to have followed the evolution of computing from the 1980s until now. People have been saying x86 was doomed since the early 90s, yet here we are.

 

[Tup3x]

Having the "best" core doesn't really mean much. If basically everyone can build their own CPU with off the shelf ARM design, it's going to make a difference. Also the cost is going to be a big factor. If you choose x86, you're stuck with Intel or AMD and I don't think companies want that now that there are actual alternatives.

 

[FlameTail]

You are comparing 2024 uarch with 2022-2023 uarch.

Where? I was alluding to a Zen5 vs M4 comparison, both of which are 2024 uarch.

Zen 5/Arrow Lake is the competition for the M4 and the snapdragon.

Sure

As for Qualcomm, I have seen enough that I (and others I have spoken with) am extremely skeptical of Qualcomm’s claims. My suspicion is perf/watt won’t match what has been claimed. The chip will be fast, sure, but it will struggle to compete with other offerings due out this year.

Indeed. If you have been following the Qualcomm thread, you would see I myself have noted this, and have been raising alarm over it, so much so that Spudlobby is telling me to calm down (XD).

Microsoft has always chased things like they have ADHD. They will chase ARM and AI until the bubble pops. Once that happens, a market correction will occur and things will change.

Do you have any ideas how that market correction will pan out?

I do need to clarify that I am not against ARM or RISC-V, I am just old enough to have followed the evolution of computing from the 1980s until now.

I salute you.

 

[branch_suggestion]

It is more accurate to compare the M4 P-core to Z5c as both are N3E and aim to offer a strong balance of performance, power and area.
Z5 classic is N4X and is built for pure performance.

 

[FlameTail]

The entirety of your statements are soaked with pro-ARM bias, so please don't meme yourself by calling out people's biases now.

I will not deny that I have a preference for ARM, but I strive to be factual and open-minded.

And it is not relevant. Zen 5 comes out in 2 weeks and that's that.

Last time I checked, as per the leakers, Zen5 is said to bring a 40-50% ST performance improvement.

Geekbench 6
Hawk Point: 2700
M3 : 3150
M4 : 3900
Strix Point : 3700-4000

So Zen5 in Strix is going to match M4 ST performance at best, but it will most certainly use significantly more power to do so.

Apple's fancy fat cores won because they were fat and always a node ahead.

Attributing Apple's success to them being on an advanced node is a ridiculous presumption.

The node is just the foundation. It's not magic. The microarchitecture is where the magic is.

https://www.reddit.com/r/hardware/comments/1cxq7em/comment/l54rzba

Case in point: Ryzen 7840HS vs Apple M2

Both score about 2600 points in GB6 ST, but the Ryzen uses about 2x the power to do so. That 2x difference isn't coming from the node, and the irony is that Apple doesn't even have a node advantage here (N4 vs N5P). You can't use the newer uarch excuse, because both Zen4 and M2 are 2022 uarches.

If Zen 5, on N4P rather than N3E, comes even close to M4, that's it

Unless Zen5 is bringing like a 2x ST performance improvement or even better- 2x the performance-per-watt (both of which are unlikely), I don't see any issue for Apple. They'll be fine.