AMD and Nvidia Will Make Arm-based Processors for PCs

[Donut417]

On 10/27/2023 at 4:27 PM, Monkey Dust said:

I'm not sure if it will be such a massive issue for many users. A lot of home users only use the browser, and maybe Zoom. Even in offices, some roles only require a browser and MS Office. As long as MS don't screw this up (place your bets ladies and gentlemen) it could meet the needs of most PC buyers. From a PC gaming perspective, I doubt we'll get native ARM games until the consoles go ARM.

 

I wonder if Intel & AMD could pull off hybrid ARM/x64 chips? I guess it would be more if Windows could be made to handle it.

There will be a lot of people who could use ARM just fine. Between half way decent emulation or The Cloud people would be able to adapt. Even gaming, I mean the Nintendo Switch does run on an old ass ARM chip at this point but it still an ARM chip. The biggest issue for PC gaming is going to be the death of DYI PC building. Because most ARM SOC's have everything built on them, from CPU, GPU and RAM. I know some of Qualcomm's chips have LTE and 5G modems built in. If all the things are integrated then upgradability would be dead at this point. 

 

But I think x86 is going to be staying for a while yet. The biggest thing this is going to change is with mobile devices. I could see maybe more Windows tablets and some ARM based laptops. If anything it will just mean a bigger selection of devices. 

[Forbidden Wafer]

4 hours ago, hishnash said:

(why this was ever supported I don't know). 

According to every single talk I've watched regarding bootloaders, bootstrapping, etc is that it is damned hard to get things right at that level, so everyone sticks to what already works. 

[captain_to_fire]

On 10/29/2023 at 3:59 AM, starsmine said:

apple was mad at Intel not improving power yes… because Intel was stuck at 14nm for far longer then expected, not because of the ISA.

Not just that. An ex-Intel engineer once said one of the reasons why Apple ditched Intel aside from the failure to shrink transistor size are the bugs that Apple engineers are finding

 

https://www.zdnet.com/article/ex-intel-engineer-apple-turned-away-from-intel-over-skylake-cpu-bugs/

[hishnash]

1 hour ago, captain_to_fire said:

An ex-Intel engineer once said one of the reasons why Apple ditched Intel aside from the failure to shrink transistor size are the bugs that Apple engineers are finding

The other reasons are around everything else within the SOC.. once apple got to the point were they were replacing large parts of the system arc with thier own silicon (for sec reasons) (T2) it was clear were they were going.   If they could not trust Intels UEFI secure boot and needed to put other chips in place the hand hold the cpu during boot to patch out big security issues they were clearly on the path away from intel. 

[LAwLz]

On 10/29/2023 at 2:16 AM, leadeater said:

That's nice and all but this won't actually stop the problem of just a few big companies being the mainstay i.e. silicon fabs. Anyone can be a silicon fab however the actual number of leading edge ones is exceedingly small. I actually don't see it being much different, we'll probably just end up with a few clear pack leaders that will dominate giving us a situation not really any better. Maybe we will go from 2 companies to 3 or 4 but it's also likely 1 or 2 will only be relevant to consumers each product generation as the lead in XYZ of consumer importance changes between companies.

 

And the other fact is very few, just Apple?, actually make their own full designs and the rest are direct implementation of ARM designs so I don't see how 1, 2 or 10 companies making the exact same microarchitectures all on TSMC for example is going to help anything. That's like going to the supermarket and saying you have "brand choice" when the reality is it's all made in the same factory with the same ingredients and recipe and only the brand label put on it is different. A bag of frozen peas all source from the same suppliers so yay choice  heh.

We would have:

1) Intel

2) AMD

3) Qualcomm

4) Samsung

5) Nvidia

6) Arm

 

All of those have their own CPU engineering teams that can make CPU architectures.

That's a lot better than today when we only have Intel and AMD. It's not just Apple making arm CPU designs. We have at least 4 companies doing it. If all companies are just making chips with the same cores on the same fab then you have a point, but that does not seem to be the situation we're heading towards.

 

If we're strictly speaking about top-tier performance then chances are things will play out as you said, that we might still end up with a few companies (hopefully more than 2) that are above everyone else, but hopefully who is on top might change from year to year. Even if it doesn't, it would still give room for competition in the lower segments. Instead of only having 2 companies trying to push prices at the ~100 dollar market for example we might have 4, which should result in lower prices and higher performance.

 

 

 

 

On 10/29/2023 at 9:59 AM, porina said:

I'm sure they could get into x86 if they wanted to

They can't. AMD and Intel are essentially the only ones allowed to make x86 processors (yes I know about Via and IBM, and a new one I learned about called DM&P which got their license from Rise Technology). 

In fact, Nvidia actually sued Intel back in ~2009 or something like that, and one of the things that came up during the hearings was whether or not Nvidia should be granted an x86 license. The end result was no, Nvidia are not allowed to make x86 CPUs and Intel didn't have to grant them permission to do so.

This was around the same time they tried to get an x86 license through Transmeta, but as you can read in the article above the court denied Nvidia even the right to make an x86 emulator.

 

Hell, even if let's say Nvidia bought AMD, Nvidia still wouldn't be allowed to make x86 processors because the cross-license agreement that AMD and Intel have is non-transferable. Although Intel would probably want to write a new agreement because otherwise, they would probably lose access to x86_64

 

 

 

On 10/29/2023 at 9:59 AM, porina said:

I see Starsmine covered it well already. It wasn't Intel didn't want to release better designs, they simply couldn't. AMD were fortunate in that as it gave them time to kick the ball rolling on Ryzen. Even then, it was only with Zen 2 they clearly passed Skylake in microarchitecture. Today there is very little to choose between them on performance, with AMD having a power advantage.

1) I don't agree with how starsmine covered what happened during those years.

2) I don't think AMD has the power advantage across the board. It really depends on which application you run and which chips you compare.

 

 

On 10/29/2023 at 5:54 PM, starsmine said:

Sandy to Broadwell had no slowdown

Sandy to die shrink Ivy
new arch Haswell to die shrink Broadwell

Ivy and Broadwell had negligible IPC improvements as they were shrinks. 

Sandy Bridge - Let's use this as the base.

Ivy Bridge - die shrink with 3-6% IPC gain.

Haswell - Same node and only ~3% performance gain.

Broadwell - die shirk so as you said, very little performance change (about 4%).

 

Those were the ones launched before Intel started having issues, and we got pretty much no performance improvements. Almost 4 years where we got ~10% better performance in total

 

 

Actually, let's look even further by comparing Sandy Bridge to Skylake. Skylake was the last chip Intel released following their usual tick-tok cadence, and before they started having issues moving past 14nm.

 

Sandy Bridge to Skylake, 4,5 years, we got an average of 37% more performance (24% higher IPC). That includes Dolphin which was a major outlier.

That's very bad for 4,5 years. And again, that was before Intel started struggling.

I don't think it was for technical reasons Intel decided to not release more than quad-core processors for the mainstream until AMD decided to do so. It was a business decision that could be justified by saying "we don't have to".

 

 

On 10/29/2023 at 6:52 PM, porina said:

Architecture updates typically give the biggest peak IPC uplifts.

 

My personal favourite Prime95 saw significant uplifts in the "Core i" era. 

Nehalem > Sandy Bridge was about 2x IPC due to addition of AVX

Sandy > Haswell was about 1.5X IPC due to updating to AVX2

Haswell > Skylake was +14% IPC due to general improvements.

Skylake > Rocket was about +40% IPC due to consumer AVX-512

Skylake > Skylake-X was about +80% IPC due to full fat AVX-512

Alder Lake would be a regression in IPC since they lost AVX-512 but I don't have numbers on that

 

If anyone wonders where AMD fits in, Zen and Zen+ were comparable to Sandy Bridge. Zen 2 is about +4% over Skylake. Zen 3 is about +10% over Zen 2. Zen 4 fits somewhere between Rocket Lake and Skylake-X but I don't have first hand data to be more exact. 

 

Don't like Prime95, how about Cinebench R15. It's an older version not using AVX at all.

Sandy > Haswell was about +16% IPC

Haswell > Skylake was about +12% IPC

I'd say those are bad benchmarks because it's basically just AVX throughput. Some people might look for that, but it's not really what I would say is the be-all and end-all way of making generalized statements about a processor's performance.

Those numbers are only relevant for a very specific use case and I am trying to talk more broadly than that.

 

As for those Cinebench R15 numbers, they seem to line up pretty well with the numbers I posted from Anandtech, but that means in 4,5 years we only got a ~30% performance increase from Intel, despite multiple node shrinks and several new microarchitectures. That's bad.

 

 

For comparison, Alder Lake gave a ~18% single-core performance increase and ~75% multi-core performance increase compared to Rocket Lake. That's what I call a good generation-to-generation improvement. Intel made a bigger jump going from 11th gen to 12th gen than they did going from 2nd gen to 6th gen.

 

I think this comment on Anandtech's Alder Lake review sums it up pretty well:

Quote

Biggest improvements since sandybridge. If you look at the timeline, this wouldve been the first CPU designed since they saw Zen 1. This is their Zen 1 moment and they already took the performance crown back basically across the board and at a lower price.

 

The timeline lines up very well if we assume that Intel didn't feel the need to push forward that much with CPU performance after they dominated that segment.

They still made massive strides with their iGPUs though, which was the area where they still were behind.

While CPU performance only increased by like 30%, iGPU performance increased by 300%. That should tell you something about where Intel's focus was during those years. It wasn't on their CPUs.

[leadeater]

31 minutes ago, LAwLz said:

We would have:

1) Intel

2) AMD

3) Qualcomm

4) Samsung

5) Nvidia

6) Arm

 

All of those have their own CPU engineering teams that can make CPU architectures.

Except Arm doesn't actually make products so it's not really correct to count them. There's already lots of Arm server CPU/SoC makers and NIC DPUs etc that use Arm IP and license, not that it matters a whole lot since we are talking consumer devices like laptops here.

 

31 minutes ago, LAwLz said:

That's a lot better than today when we only have Intel and AMD. It's not just Apple making arm CPU designs.

I think you misunderstand the point, Apple does not literally use Arm core designs and actually designs microarchitectures. That list is exceedingly small. We have Apple, Fujitsu, Qualcomm (ish currently, excluding the future this news is showcasing, are derivative designs of core Arm IP, I'll list but iffy), Nvidia (custom and uses Arm core IP). I have probably missed one or two, Altra is just Arm core IP, Thunder X I think is custom. But as you can see the list for consumer devices is basically 2, maybe 3, and 1 for laptops (laptops that matter, Apple).

 

The rest are playing Lego with literal Arm core IP which is not what I would call innovative in a sense of pushing forward CPU/SoC microarchitecture and stimulating innovation and competition.

 

Product wise yes it's a little better than what we have now but not by much and we will still likely be subjected to the bigger players leading and being the only realistic choice in higher end devices.

 

So back to my point, we only have Apple making Arm ISA designs realistically in this context. Intel and AMD are not sharing microarchitecture designs, they are both making custom x86 ISA designs, that is what matters here for what I said, not a knock off company like Cyrix dependent on Intel for example. No I'm not saying companies using Arm core IP to make SoCs are knock off companies but Cyrix is a good illustration between the difference of Intel vs Cyrix and Intel vs AMD.

 

31 minutes ago, LAwLz said:

If all companies are just making chips with the same cores on the same fab then you have a point, but that does not seem to be the situation we're heading towards.

That literally is the point and this is what is happening now and will mostly happen. Qualcomm here in this story is the first to actually be delivering a custom Arm microarchitecture that will be used in consumer laptops. So I'll point out one does not make a trend. If this spurs on more companies to make Arm Windows laptops they will, and this I will guarantee, just be Arm core IP Lego of absolutely no innovation that matters (to me). Maybe 5 years after that happens more custom microarchitecture will get made but until that happens I don't see much changing other than getting one actually new and decent option.

 

The Nvidia & AMD rumors while interesting need a lot more details since assuming they are targeting consumer devices would be a misguided assumption. And we are still at the mercy of Microsoft when it comes to Windows as well, it has to actually support any innovations these companies want to do and that's actually not a safe bet sadly.

[porina]

10 minutes ago, LAwLz said:

They can't. AMD and Intel are essentially the only ones allowed to make x86 processors (yes I know about Via and IBM, and a new one I learned about called DM&P which got their license from Rise Technology). 

If there was sufficient value in making it happen, nvidia just had to wave its bank account in the direction of Intel to get a license. At this moment in time, nvidia is 6.7x the market cap of Intel. It's not a great measure of company size or worth, but it goes a long way to getting business deals done. That it hasn't happened, means the cost of doing so probably isn't worth it.

 

10 minutes ago, LAwLz said:

1) I don't agree with how starsmine covered what happened during those years.

I think the Intel part is accurate to my memory and understanding. I'll pass on the Apple part since I don't care.

 

10 minutes ago, LAwLz said:

2) I don't think AMD has the power advantage across the board. It really depends on which application you run and which chips you compare.

For sure there is a lot of "it depends" in everything, but if you do perf/power curves then more often than not AMD had an advantage in recent gens. It does depend on operating point, workload and which way the wind is blowing. The gap closes at the higher end of the curve.

 

10 minutes ago, LAwLz said:

Sandy Bridge to Skylake, 4,5 years, we got an average of 37% more performance (24% higher IPC). That includes Dolphin which was a major outlier.

That's very bad for 4,5 years. And again, that was before Intel started struggling.

Interesting Anandtech got 22.3% IPC for Cinebench R15, where I got about 30%. Prime95 isn't one of their tests but would be +74% IPC. That's pretty close to Dolphin actually. While I haven't used it, I vaguely recall hearing it benefits from AVX, which Prime95 does too.

 

10 minutes ago, LAwLz said:

I'd say those are bad benchmarks because it's basically just AVX throughput. Some people might look for that, but it's not really what I would say is the be-all and end-all way of making generalized statements about a processor's performance.

Those numbers are only relevant for a very specific use case and I am trying to talk more broadly than that.

It is one data point, but still a valid one needed to form the bigger picture. Nowadays when manufacturer talk IPC gains, they use a wide variety of workloads. The Anandtech article had 7, bulked out by ST/MT and/or config. If you think Prime95 is a bit niche, then you might want to ignore 3DPM which is code Ian wrote as part of his doctorate in computational chemistry. To me it is interesting as it was the first piece of software that saw >50% IPC gains through enabling HT, and it had insane AVX-512 acceleration rate. You'd be hard pressed to find any production software that had those characteristics. At least Prime95 shares roots with scientific and engineering workloads.

 

This also reminds me a bit of Turing when it was released. The value add was the RTX side. You can't raster your way to decent RT. The direction improvements have taken for a while is by making specific things faster, but there will be some lag effect before it gets adopted if applicable.

 

Now that AMD finally has reasonable AVX performance since Zen 2, it does feel like more software is making use of it than before.

 

10 minutes ago, LAwLz said:

For comparison, Alder Lake gave a ~18% single-core performance increase and ~75% multi-core performance increase compared to Rocket Lake. That's what I call a good generation-to-generation improvement. Intel made a bigger jump going from 11th gen to 12th gen than they did going from 2nd gen to 6th gen.

Can we be clear here, are you talking product performance, or architecture performance? I've tended to stick to architecture, often boiled down to IPC, as it makes it clear how that improves. Product level means you have much wider cores and clocks to consider, and may not be very like for like. Zen 1 had great Cinebench scores but it generally sucked at everything else at the time.

[Donut417]

13 minutes ago, leadeater said:

And we are still at the mercy of Microsoft when it comes to Windows as well, it has to actually support any innovations these companies want to do and that's actually not a safe bet sadly

Microsoft has been trying to get Windows on ARM to be a thing forever. I think they would welcome any kind of innovation that leads to that.

 

 

14 minutes ago, leadeater said:

Apple does not literally use Arm core designs and actually designs microarchitectures.

I do recall reading that Apple uses a very small amount of ARM's IP in their  Apple Silicone SOC, but most of it just their engineering.

[igormp]

39 minutes ago, LAwLz said:

We would have:

1) Intel

2) AMD

3) Qualcomm

4) Samsung

5) Nvidia

6) Arm

 

All of those have their own CPU engineering teams that can make CPU architectures.

Let's not forget that Samsung also used to make their own designs, and I wouldn't doubt of a comeback in case WoA made a turn for the better. They could achieve total vertical integration this way.

 

Marvell and Ampere also have custom designs meant for servers, but maybe they could scale it down to desktops/HEDT/WS in case WoA got enough traction.

16 minutes ago, leadeater said:

Altra is just Arm core IP

Altra is, One is custom.

[Donut417]

1 minute ago, igormp said:

et's not forget that Samsung also used t

Samsung gave up on the Exynos?

[leadeater]

7 minutes ago, Donut417 said:

Microsoft has been trying to get Windows on ARM to be a thing forever. I think they would welcome any kind of innovation that leads to that.

Which would stop pretty fast once someone comes to them with 4 different layers of cores and asks Microsoft to update their scheduler to support that 

 

But I disagree, Microsoft hasn't really wanted to. If they did it would be a thing now already. Dipping toes and making it a strategic company direction are way different things.

[leadeater]

4 minutes ago, igormp said:

Let's not forget that Samsung also used to make their own designs, and I wouldn't doubt of a comeback in case WoA made a turn for the better. They could achieve total vertical integration this way.

I didn't include them because everything they have done recently under Exynos is just a mixture of Cortex cores, we have to go back to like 2016 for something actual custom. Didn't think it was worth it since they aren't currently doing it, unless I missed something.

[Donut417]

1 minute ago, leadeater said:

But I disagree, Microsoft hasn't really wanted to. If they did it would be a thing now already. Dipping toes and making it a strategic company direction are way different thinfs.

It is a thing. The Surface Pro 9 with the Microsoft SQ 3, which is an ARM based SOC with 5G support. And its the ONLY way to get 5G on a Surface. Previous they had the Surface X. Then lets not forget Windows RT. So its technically a thing. They might not be selling like hot cakes but Microsoft keeps pushing ARM devices.

 

The only part of this is no other OEM is doing any ARM based Windows devices from what I have seen. But at this point I just figured Microsoft wants people to buy their hardware rather than from an OEM.

[leadeater]

1 minute ago, Donut417 said:

It is a thing. The Surface Pro 9 with the Microsoft SQ 3, which is an ARM based SOC with 5G support. And its the ONLY way to get 5G on a Surface. Previous they had the Surface X. Then lets not forget Windows RT. So its technically a thing. They might not be selling like hot cakes but Microsoft keeps pushing ARM devices.

 

The only part of this is no other OEM is doing any ARM based Windows devices from what I have seen. But at this point I just figured Microsoft wants people to buy their hardware rather than from an OEM.

I'm very aware of the Arm Surface and the custom SoC for it, my statement stands.

[pound591]

This is pretty crazy news, I liked the videos you have put out on the news so far but I hope Emily does the write up when the cpu's launch since Em did the really good video on windows switching to it from x86 last year. Also I don't think a full change from x86 is in the near future just think it's cool to see new tech like this and how it progresses. https://youtu.be/LFQ3LkVF5sM?si=nICaDztSkGYs0kuG

[LAwLz]

11 hours ago, leadeater said:

Except Arm doesn't actually make products so it's not really correct to count them. There's already lots of Arm server CPU/SoC makers and NIC DPUs etc that use Arm IP and license, not that it matters a whole lot since we are talking consumer devices like laptops here.

I was listing companies that design microarchitectures. 

The reference cores will probably end up in SoCs from MediaTek and possibly a few others.

 

My hope is that in 10 years or so we will have maybe 4-5 different core architectures to choose from for PCs.

That might be AMD, Intel, Qualcomm, arm reference (MediaTek) and Nvidia, for example.

 

 

 

11 hours ago, leadeater said:

I think you misunderstand the point, Apple does not literally use Arm core designs and actually designs microarchitectures. That list is exceedingly small. We have Apple, Fujitsu, Qualcomm (ish currently, excluding the future this news is showcasing, are derivative designs of core Arm IP, I'll list but iffy), Nvidia (custom and uses Arm core IP). I have probably missed one or two, Altra is just Arm core IP, Thunder X I think is custom. But as you can see the list for consumer devices is basically 2, maybe 3, and 1 for laptops (laptops that matter, Apple).

I think you misunderstood me.

The list I gave you is for companies that have their own microarchitecture design teams. All 6 of them have (and could continue to) design their own core architectures in the same way Apple does.

I understand how the Arm ecosystem works. I understand that a lot of players just license the same reference cores. That is not what I was talking about.

There are more companies with CPU engineers, who are or have been designing custom CPU cores that are compatible with the Arm ISA. It's not just Apple and Qualcomm, and I would also include the reference cores as potentially being a good buy for a lot of people if WoA takes off.

 

It might not be for you or me, but it would still benefit the majority of consumers if things got more competitive on the mid and low end.

 

 

 

11 hours ago, leadeater said:

Product wise yes it's a little better than what we have now but not by much and we will still likely be subjected to the bigger players leading and being the only realistic choice in higher end devices.

High-end is not the only product category that matters if you ask me.

 

 

10 hours ago, igormp said:

Let's not forget that Samsung also used to make their own designs, and I wouldn't doubt of a comeback in case WoA made a turn for the better. They could achieve total vertical integration this way.

 

Marvell and Ampere also have custom designs meant for servers, but maybe they could scale it down to desktops/HEDT/WS in case WoA got enough traction.

Altra is, One is custom.

Samsung has been gobbeling up CPU architecture engineers for quite some time now. There are several rumors about them building a "dream team" to once again create their own custom core designs. The current rumors and speculations point towards it launching in 2025.

I wouldn't be surprised to see that architecture make its way into Windows machines if Windows on Arm takes off.

 

 

  

10 hours ago, Donut417 said:

Samsung gave up on the Exynos?

Let's not get Exynos confused with their core architectures.

Exynos is Samsung's name for their SoCs, which include CPU cores, GPU, and so on.

 

Samsung has not given up on that. They didn't ship any phone with a high-end Exynos processor this generation, but rumors have it that they will do so next year. They also released a lower-end Exynos SoC, the Exynos 1380, this year. And also some stuff like the Exynos W930 for smartwatches. But the Google Tensor chip is designed and manufactured by Samsung's Exynos team too. It's basically a custom version of an Exynos chip.

 

So Samsung has not given up on Exnos.

What they did give up on a few years ago was their custom CPU architectures, which were called "M#" for example "M1", "M2". They used to put those CPU cores in their Exynos chips, but because of lacking performance and efficiency, they switched over to cores designed by arm instead. 

But the current rumors and hirings indicate that they are working on more custom cores that will probably show up in future Exynos chips.

 

 

  

10 hours ago, leadeater said:

I didn't include them because everything they have done recently under Exynos is just a mixture of Cortex cores, we have to go back to like 2016 for something actual custom. Didn't think it was worth it since they aren't currently doing it, unless I missed something.

2018 was the last time they shipped a custom CPU core.

But recent rumors and hirings heavily point towards them designing custom CPU cores again. See links above for news articles about them hiring ex-Apple, AMD, and Intel CPU architects. It seems like it's just not ready to launch.

I expect to see custom CPU cores from Samsung in the coming 1-3 years.

 

 

The list of companies we know have CPU architecture teams that are or are ready to design CPU cores compatible with the Arm instruction set are:

Apple (no examples needed hopefully but won't show up in PCs)

AMD (AMD K12, which Jim Keller said AMD "stupidly canceled")

Ampere Computing (Siryn)

Fujitsu (A64FX)

Nvidia (Denver and Carmel)

Marvell (through their acquisition of Cavium which made the ThunderX cores)

Samsung (their M series)

Qualcomm (Oryon)

Huawei (HiSilicon TaiShan)

 

 

Apparently, Microsoft is also hiring SoC architects, but I am not sure what that job entails. Might not be CPU core architectures. They did have a job listing for "CPU design verification engineer" and the job description included being part of their "custom CPU development team".

I wouldn't bet on Microsoft making a custom CPU architecture though.

[leadeater]

Should be interesting, watching it now 

 

 

[leadeater]

7 hours ago, LAwLz said:

There are more companies with CPU engineers, who are or have been designing custom CPU cores that are compatible with the Arm ISA. It's not just Apple and Qualcomm, and I would also include the reference cores as potentially being a good buy for a lot of people if WoA takes off.

Don't get me wrong the Arm reference cores are really good, often the best or more recent ones just don't get used or take a very long time to since they often aren't what is required for products at the time. It's just I don't find it particularly interesting to get multiple rebranded offerings of the same thing, 1 is the same as 4 in that situation for me.

 

The other problem is while the list you made includes companies that could that doesn't mean the are or will. Nvidia for example their Grace CPU is actually Arm Neoverse V2 core with most of the custom SoC design portions being NVLink. Currently it does not make sense for Nvidia to design a custom microarchitecture since their business and product focus is empowering their GPUs and by using what Arm has to offer gives them the quickest path to market with good future development plans. This extends to their Tegra product family and everything else which have all moved to stock Arm cores, nothing past 2018 is custom from Nvidia. Also the more that use Arm core designs the better these will actually be meaning going custom to have the best possible increasingly become less required.

 

I can also see AMD doing similar, I also see them also sticking mostly to servers rather than consumer as even their past toe dip in to Arm was server only. Intel loves always having their own stuff so them I see doing it from the very start.

 

Samsung might but it's also just as likely they will stick to Arm cores and design other aspects like GPU (AMD licensed) and NPUs etc.

 

The only literal and announced consumer device custom Arm cores coming any time soon are from Qualcomm and they do sound pretty damn good, consumer being what I specified in my first comment about this. Apple has currently been the only with Qualcomm soon/now, everything else is rumor and conjecture without good supporting evidence and most evidence for many being counter to custom or consumer at all even.

 

7 hours ago, LAwLz said:

The list of companies we know have CPU architecture teams that are or are ready to design CPU cores compatible with the Arm instruction set are:

Apple (no examples needed hopefully but won't show up in PCs)

As I said and listed most of those, they are primarily server product not consumer and won't come to consumer. I did however forget Ampere are going custom and I didn't think about Huawei. The rest I did mention and excluded since they won't and aren't applicable to consumer as well as excluded for being so old it's not relevant,

 

I urge caution that just because a company did something say 10 years ago that that is what they will do today or even 5 years from now. A presumption of will because they can/could isn't such a great idea. You can be designing custom silicon requiring silicon engineers while still actually using Arm cores. Above all if you go custom you have to actually achieve better than what Arm offers which many have failed to do in the past.

[Forbidden Wafer]

57 minutes ago, leadeater said:

Should be interesting, watching it now 

 

 

Good interview. The problem with Qualcomm is that their support sucks (most android phones get stuck in old versions due to them refusing to produce new blobs required for newer android versions).

[Kisai]

On 10/29/2023 at 3:08 PM, Donut417 said:

There will be a lot of people who could use ARM just fine.

Name some. Don't put "Children in school" in there though, we've already seen the abject failure of chromebooks.

 

On 10/29/2023 at 3:08 PM, Donut417 said:

Between half way decent emulation

 

You mean pirates.  Pirates don't buy things, they steal them. No pirate is buying an ARM chip device to use for themselves. No, you see these "10,000 in one" game consoles on ebay that are just 10,000 pirated games on a weak device that can't actually play most of them.

 

On 10/29/2023 at 3:08 PM, Donut417 said:

or The Cloud people would be able to adapt. Even gaming, I mean the Nintendo Switch does run on an old ass ARM chip at this point but it still an ARM chip. The biggest issue for PC gaming is going to be the death of DYI PC building. Because most ARM SOC's have everything built on them, from CPU, GPU and RAM. I know some of Qualcomm's chips have LTE and 5G modems built in. If all the things are integrated then upgradability would be dead at this point. 

The Switch is an ARM chip that is a Nvidia Tegra X1, it's not a smartphone chip, it's the same SoC in the Nvidia shield, that was also popular with "Kodi box" pirates.

 

On 10/29/2023 at 3:08 PM, Donut417 said:

But I think x86 is going to be staying for a while yet. The biggest thing this is going to change is with mobile devices. I could see maybe more Windows tablets and some ARM based laptops. If anything it will just mean a bigger selection of devices. 

 

I think what needs to shake up the landscape is AMD and Nvidia producing ARM SoC's themselves that are patterned after Apple's offerings (forget Qualcomm) and intended for iMac/iPad Air type of solutions, not trying to re-invent the iPhone (which is what Intel tried to do) or re-invent the laptop (which is what "Netbooks were originally.)

 

Yes, Netbooks existed for a while, so did PDA's. Neither of these caught on beyond a few niches unless they could run desktop Windows. Nobody wanted them when they ran Linux, and nobody wanted them they ran WindowsCE. 

 

If there is one thing Apple seems to always get right, is despite a weaker computer offering, everything they've sold since the PPC era, has always had much better battery life, when it was a laptop, and much better use-case-specific performance (eg photoshop, final cut, logic pro, etc) that was a direct consequence of it simply engineering the OS to better match the hardware offering. Until the switch to ARM, Apple mostly had to make do with the anemic Intel iGPU's, or back in the PPC era, try unsuccessfully to convince IBM and Motorola to make parts they wanted. I guess finally since they had an excuse to make their own SoC's for the iPod, iPhone, iPad, they could apply that to the Mac lineup at the expense of the Mac Pro.

 

Call it what you want, but the Mac Pro seems to be where Apple either doesn't know, or doesn't care about the customer use case. If you can't add additional GPU's to it, then what is the point of it existing? Yes other PCIe stuff exists, but we are scraping the bottom of the barrel when it comes to "stuff that still works on a mac" because Apple has ignored the customer segment that would buy the Mac Pro, and catch-22 thus manufacturers don't bother building pro equipment and drivers for it.

 

Maybe it will take someone building a ARM Nvidia SoC "mac Pro killer (*laugh*)" before Apple does. Right now, there are no "ARM PC"'s for which manufactures can target except the Mac Pro. 

 

 

[igormp]

31 minutes ago, Kisai said:

Name some. Don't put "Children in school" in there though, we've already seen the abject failure of chromebooks.

 

I dare to say over 60% of PC users. As the person you replied to said, all they need is a browser and maybe office, this can be achieved with any decent ARM soc as long as the software does its part just well. Modern Mac users are pretty happy with their ARM devices.

30 minutes ago, Kisai said:

You mean pirates.  Pirates don't buy things, they steal them. No pirate is buying an ARM chip device to use for themselves. No, you see these "10,000 in one" game consoles on ebay that are just 10,000 pirated games on a weak device that can't actually play most of them.

I believe he meant x86 emulation on arm.

33 minutes ago, Kisai said:

The Switch is an ARM chip that is a Nvidia Tegra X1, it's not a smartphone chip, it's the same SoC in the Nvidia shield, that was also popular with "Kodi box" pirates.

 

The X1 does use smartphone cores lol

It's a cortex a57 with a custom GPU, same CPU as the SD810 and other devices.

[Forbidden Wafer]

12 minutes ago, igormp said:

I dare to say over 60% of PC users. As the person you replied to said, all they need is a browser and maybe office, this can be achieved with any decent ARM soc as long as the software does its part just well. Modern Mac users are pretty happy with their ARM devices.

That is only the case because they don't use MSFT software.

[Kisai]

4 minutes ago, igormp said:

I dare to say over 60% of PC users. As the person you replied to said, all they need is a browser and maybe office,

and an internet connection with with to use it. Oh dear, that means that cloud company you have to pay a monthly subscription to, can say that they don't support your 2 year old laptop at any time.

 

It's one thing to say "I can use this POS ARM device, it's powerful enough for my business needs" like a store kiosk, ATM or a subway ticket vending machine, something that is essentially glued in place and has to be supported by some guy on the other side of the planet. It's another to say "this thing works, but only if you're paying for 5G internet"

 

4 minutes ago, igormp said:

this can be achieved with any decent ARM soc as long as the software does its part just well. Modern Mac users are pretty happy with their ARM devices.

I believe he meant x86 emulation on arm.

Which is still terrible.

 

4 minutes ago, igormp said:

The X1 does use smartphone cores lol

It's a cortex a57 with a custom GPU, same CPU as the SD810 and other devices.

And yet, the Switch is not a smartphone, and neither is the Nvidia shield. Both of these devices require internet access to be of any use what-so-ever, and even if you only bought the Switch games on carts, you'd be unable to play a bunch of them with friends, since everyone needs to be on the same version, which only happens with internet access.

 

The point I'm making here is that "ARM PC" is not generally a thing. ARM has been around since the 80's. It's been a CPU in PDA's long before it was in iPods and iPhones. Why didn't we see ARM PC's then huh?

 

Couldn't have been because battery life sucked. Couldn't have been because no two PDA's, despite running Windows CE could run the same software (ARM was not the only CPU, and different screen sizes had different versions)

 

I'm not even willing to give Apple credit on this, because they also abandoned their PDA, despite inventing the "PDA". 

 

No, even Intel had invested in ARM back in the 90's, and decided that wasn't what they wanted to do with it.

 

It seems like people want to pretend that stuff didn't get adopted before, because there wasn't any pressing need for it. The only reason any company is even trying  ARM laptops now is because they want to capture the customers Apple is leaving behind. Yet, you want to offer them WINDOWS?

 

 

[Donut417]

2 hours ago, Kisai said:

Name some. Don't put "Children in school" in there though, we've already seen the abject failure of chromebooks.

Any one who needs a basic machine. Any one who wants something that will work well for travel. Business who rely on the cloud. Bare in mind Im talking about having good support from Microsoft as well as good support from devs.

 

Also if devs were more on board you could probably do about anything with an ARM based machine. Just need the right hardware and software.

 

2 hours ago, Kisai said:

You mean pirates.  Pirates don't buy things, they steal them. No pirate is buying an ARM chip device to use for themselves. No, you see these "10,000 in one" game consoles on ebay that are just 10,000 pirated games on a weak device that can't actually play most of them.

Im talking x86 emulation. So you know can run x86 apps that might not have an ARM version. So no, Im not talking about piracy.

 

2 hours ago, Kisai said:

The Switch is an ARM chip that is a Nvidia Tegra X1, it's not a smartphone chip, it's the same SoC in the Nvidia shield, that was also popular with "Kodi box" pirates.

ARM is ARM, doesnt need to be a phone chip. Also why do you keep bringing up priacy? Are you invested in some company that deals with piracy or something?

 

[Paul Thexton]

1 hour ago, Kisai said:

and an internet connection with with to use it. Oh dear, that means that cloud company you have to pay a monthly subscription to, can say that they don't support your 2 year old laptop at any time.

Are you saying an ARM based computer is useless without an internet connection and cloud based services?