Question x86 license value quickly dropping

Page 3 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

SpudLobby

Senior member
May 18, 2022
961
655
106
What is this "Zen 5's mess" you are referring to ?
AMD does not have the kind of ST efficiency or battery life Qualcomm has, and now Intel with Lunar Lake will add more to the party on the latter and to an extent the former ish.

The one redeeming thing AMD had was going to at least be an IPC lead or parity on Qualcomm (who is still slightly worse than Firestorm but close) and Intel, which in tandem with 20-25% better clocks and still having parts with 8-12 cores would give them much more ST than QC in mobile (if not matching Apple too) and also more MT via IPC. It was at least an arguable thing. I think overrated, because we know how AMD and Intel are to get there and I doubt this changes *too* drastically with Strix:




But now AMD doesn’t even have much of an ST lead. 5% GB6 ST lead over Qualcomm’s top part and that’s using the 4GHz X Elite on a Surface, not a 4.2GHz one, and vs a 5.1/5.2GHz AMD Strix part. Doubtless will zap more power too.



This is pathetic. I already don’t think the tradeoffs for AMD/Intel core and SoC designs are worthwhile where they blow on things that really matter to mobile users and target high clocks, but now they’re not even doing the one thing people here thought was possible and generally mentally jerk off to — an extra 15-30% ST for a serious power payment.
 

mpumalanga

Junior Member
Feb 18, 2022
6
14
51
AMD does not have the kind of ST efficiency or battery life Qualcomm has, and now Intel with Lunar Lake will add more to the party on the latter and to an extent the former ish.

The one redeeming thing AMD had was going to at least be an IPC lead or parity on Qualcomm (who is still slightly worse than Firestorm but close) and Intel, which in tandem with 20-25% better clocks and still having parts with 8-12 cores would give them much more ST than QC in mobile (if not matching Apple too) and also more MT via IPC. It was at least an arguable thing. I think overrated, because we know how AMD and Intel are to get there and I doubt this changes *too* drastically with Strix:
View attachment 100349



But now AMD doesn’t even have much of an ST lead. 5% GB6 ST lead over Qualcomm’s top part and that’s using the 4GHz X Elite on a Surface, not a 4.2GHz one, and vs a 5.1/5.2GHz AMD Strix part. Doubtless will zap more power too.



This is pathetic. I already don’t think the tradeoffs for AMD/Intel core and SoC designs are worthwhile where they blow on things that really matter to mobile users and target high clocks, but now they’re not even doing the one thing people here thought was possible and generally mentally jerk off to — an extra 15-30% ST for a serious power payment.

Thank you for your detailed answer!
It will be interesting to see which laptops will take on zen 5 then.
 
Reactions: SpudLobby

Doug S

Platinum Member
Feb 8, 2020
2,479
4,036
136
x86 is not mythical because of the hardware in and of itself, it is here for the long haul because of software and backwards compatibility.

If you are arguing purpose built hardware for specific workloads fine, but general purpose computing, personal computing and high performance computing will always lean to x86 based hardware (because software / technical debt).

Too much software and hardware would have to be changed for ARM to become dominant over x86.

I have an M1 Mac Mini, neat and capable hardware, but sure as the sky is blue it cannot run a VM of full featured x86 Windows.

And I highly doubt we will ever see the day where there is ARM based CPU personal computer giving the same gaming experience as a high end gaming PC.

You clearly have absolutely zero idea how the high performance computing market works if you think there is software/technical debt there. That's one field with almost no legacy, they are happy to throw out not only the baby and the bathwater, but remodel the entire bathroom whenever it comes time to make a new purchase. Over the decades that market has gone from CDC to multiple generations of Cray to RISC to x86 and is trending more and more to GPU (leaving out AI, just for number crunching)

There is no such thing as legacy binaries in that market. Legacy software, sure. But they write portable software because they KNOW they will have to run it on different hardware in the future, and recompile it as needed. Heck if a new release of the compiler could get them 0.5% faster performance you better believe they'll be recompiling it!
 

SpudLobby

Senior member
May 18, 2022
961
655
106
You clearly have absolutely zero idea how the high performance computing market works if you think there is software/technical debt there. That's one field with almost no legacy, they are happy to throw out not only the baby and the bathwater, but remodel the entire bathroom whenever it comes time to make a new purchase. Over the decades that market has gone from CDC to multiple generations of Cray to RISC to x86 and is trending more and more to GPU (leaving out AI, just for number crunching)
Doug thanks for this, wasn’t even going to bother. That’s exactly why GPUs have taken off for HPC, why stuff like Grace exists independently even of AI waves, etc. It’s all highly specialized but bespoke stacks built ad-hoc.
There is no such thing as legacy binaries in that market. Legacy software, sure.
Yep. He also misses the part where MS is easing the transition phase with Arm64x and Arm64EC, and Linux frankly doesn’t even have the same legacy CRUD issues.
But they write portable software because they KNOW they will have to run it on different hardware in the future, and recompile it as needed. Heck if a new release of the compiler could get them 0.5% faster performance you better believe they'll be recompiling it!
 

B-Riz

Golden Member
Feb 15, 2011
1,530
676
136
Doug thanks for this, wasn’t even going to bother. That’s exactly why GPUs have taken off for HPC, why stuff like Grace exists independently even of AI waves, etc. It’s all highly specialized but bespoke stacks built ad-hoc.

Yep. He also misses the part where MS is easing the transition phase with Arm64x and Arm64EC, and Linux frankly doesn’t even have the same legacy CRUD issues.

I already qualified in an earlier comment non-AI workload HPC, super computers still use a lot of x86 chips and GPU compute is very much a thing, but I never brought that up, this is a FUD thread about the death of x86 ISA and ARM taking over.

Just a merry go round of ARM is great and winning now with no real concept of why x86 ISA is not really going anywhere, despite how "impressive" ARM gains are, that doesn't matter if all the software people really want to use is tied to x86.
 

scannall

Golden Member
Jan 1, 2012
1,947
1,638
136
I already qualified in an earlier comment non-AI workload HPC, super computers still use a lot of x86 chips and GPU compute is very much a thing, but I never brought that up, this is a FUD thread about the death of x86 ISA and ARM taking over.

Just a merry go round of ARM is great and winning now with no real concept of why x86 ISA is not really going anywhere, despite how "impressive" ARM gains are, that doesn't matter if all the software people really want to use is tied to x86.
I am glad to see ARM finally moving out of phones and mobile devices. It's long past due. But x86 isn't going to die anytime soon. *CAN* it die? Sure, and probably will at some point, long after I die. With almost 50 years of inertia and momentum it'll take a while. IBM has been 'doomed' for how long now? Yet they are still making money hand over fist.
 

B-Riz

Golden Member
Feb 15, 2011
1,530
676
136
I am glad to see ARM finally moving out of phones and mobile devices. It's long past due. But x86 isn't going to die anytime soon. *CAN* it die? Sure, and probably will at some point, long after I die. With almost 50 years of inertia and momentum it'll take a while. IBM has been 'doomed' for how long now? Yet they are still making money hand over fist.

Hot dang, IBM Mainframes have AI now!


It is good for the CPU space to have competition, not sure if ARM would ever overtake x86 completely, I can see it having a spot in the PC landscape; dunno if it could ever split the server market with x86.

But I am a bit surprised no one has mentioned one of the most successful ARM using companies of the last decade, Raspberry Pi; they make good software and good hardware, with a very well supported ecosystem of documentation, reference and OS updates.

WoA seems gimmicky; are we on a light-speed ARM hype train right now???
 
Reactions: soresu and gdansk

soresu

Platinum Member
Dec 19, 2014
2,941
2,164
136
WoA seems gimmicky; are we on a light-speed ARM hype train right now???
It's only a gimmick if it fails.

That being said, the likelihood of failure is not zero, something that is certainly not true of ARM Mac.
 

gdansk

Platinum Member
Feb 8, 2011
2,488
3,377
136
It's only a gimmick if it fails.
Based on the claims of some Microsoft programmers: since xtajit64 launched in 2020 they have added numerous optimizations that Rosetta2 doesn't even do and in some select code sequences it can dynamically optimize the recompiled code to better than native. I'm sure those are rare cases but still a neat possibility.

I'm not sure it's a gimmick even if it fails. They actually tried with the 64 bit version but no one noticed them improving it because, well, there was no good/compelling ARM chip for WoA until about two weeks from now.
 

FlameTail

Diamond Member
Dec 15, 2021
3,147
1,796
106
WoA seems gimmicky; are we on a light-speed ARM hype train right now???
Maybe, but I hope the ARM hype train doesn't reach velocities like Zen5 did. We know how that turned out. ARM admirers should be cautious.
 

soresu

Platinum Member
Dec 19, 2014
2,941
2,164
136
Based on the claims of some Microsoft programmers: since xtajit64 launched in 2020 they have added numerous optimizations that Rosetta2 doesn't even do and in some select code sequences it can dynamically optimize the recompiled code to better than native. I'm sure those are rare cases but still a neat possibility.
My reference to ARM Mac had nothing to do with its performance with legacy x86 code.

It was about Apple's ability to sway people to buy its products coupled with the fact that their change up is still a singular CPU ISA ecosystem.

Just like the POWER -> x86 shift back in the 00s.

In WoA's case though you have 2 competing ecosystems, hence the non zero chance of failure.

MS could do an Apple and make Windows ARM only going forward, I just don't see that happening.
 

FlameTail

Diamond Member
Dec 15, 2021
3,147
1,796
106
Some sage advice from the guy posting ARM will have 50% of the PC market by 2029 articles
Haha. I am just posting what the CEOs are saying. Doesn't mean that I am buying it. I am indeed quite skeptical that 50% marketshare is possible in 5 years. We'll have to see.

'Cautious optimism' ~Flametail.
 

soresu

Platinum Member
Dec 19, 2014
2,941
2,164
136
but I hope the ARM hype train doesn't reach velocities like Zen5 did
Zen5 hype train has been running for much longer, since before Zen4 touched the ground.

Cortex X5/925 hype train only started in January, and even Oryon/Phoenix hype only started in earnest last year.

Mike Clark went and did a Raja Koduri on us 🤦‍♂️
 

soresu

Platinum Member
Dec 19, 2014
2,941
2,164
136
Based on the claims of some Microsoft programmers: since xtajit64 launched in 2020 they have added numerous optimizations that Rosetta2 doesn't even do and in some select code sequences it can dynamically optimize the recompiled code to better than native. I'm sure those are rare cases but still a neat possibility.
Also AVX and the lack of SVE2 implementations to directly map it, restricting emulated SIMD to SSE4 basically.

AVX512 may not be in everything by any means, but there's a helluva lot of AVX2 code paths in modern Wintel software.

If it can map 1x AVX1/2 instruction to multiple NEON instructions that might be good enough for the time being.
 

FlameTail

Diamond Member
Dec 15, 2021
3,147
1,796
106
Also AVX and the lack of SVE2 implementations to directly map it, restricting emulated SIMD to SSE4 basically.

AVX512 may not be in everything by any means, but there's a helluva lot of AVX2 code paths in modern Wintel software.

If it can map 1x AVX1/2 instruction to multiple NEON instructions that might be good enough for the time being.
Isn't the lack of AVX emulation in Windows on ARM due to Intel's x86 patents for AVX?
 

Doug S

Platinum Member
Feb 8, 2020
2,479
4,036
136
My reference to ARM Mac had nothing to do with its performance with legacy x86 code.

It was about Apple's ability to sway people to buy its products coupled with the fact that their change up is still a singular CPU ISA ecosystem.

Just like the POWER -> x86 shift back in the 00s.

In WoA's case though you have 2 competing ecosystems, hence the non zero chance of failure.

MS could do an Apple and make Windows ARM only going forward, I just don't see that happening.


This is why I'm skeptical about the long run survival of Windows/ARM. Maintaining two versions of the OS, and their stable of applications, is not without cost. For developers supporting two versions (as opposed to just shipping fat binaries and only really testing on x86 because that's the dominant platform) is not without cost.

Apple is able to successfully execute their migrations because they are migrations. If they had announced an intention to build ARM Macs and support two platforms from now on I think that would have been a strategic disaster for them, risking their already low marketshare in the PC space. Microsoft isn't going to lose market share by supporting two platforms, but focus will be split and some buyers may sit on the sidelines: should I pull the trigger on an ARM PC, or will Microsoft drop it like a hot potato like last time? Should I get one of those shiny new Intel or AMD "AI PCs", or should I wait and see in case ARM is the new trend and I'll feel left behind on x86?

Now sure if ARM's wet dreams come true and they are 50% of the Windows market by the end of the decade Microsoft is supporting both for the long haul. But if they aren't comfortably in the teens a few years from now I could easily see Microsoft losing interest and beginning to hint in various ways that ARM Windows will not stand the test of time.
 
Reactions: moinmoin

soresu

Platinum Member
Dec 19, 2014
2,941
2,164
136
Isn't the lack of AVX emulation in Windows on ARM due to Intel's x86 patents for AVX?
I had thought so myself based on previous conversations about it, but after perusing the FEX-emu github repo and finding the stems of an AVX implementation (mostly constrained by a lack of available SVE hardware) I'm starting to think it's got more to do with the lack of ARM CPU cores implementing SVE, which sounds like mostly QC's fault, albeit also you can't actually do much with it unless the implementation is 256 bit wide.
 

FlameTail

Diamond Member
Dec 15, 2021
3,147
1,796
106
I had thought so myself based on previous conversations about it, but after perusing the FEX-emu github repo and finding the stems of an AVX implementation (mostly constrained by a lack of available SVE hardware) I'm starting to think it's got more to do with the lack of ARM CPU cores implementing SVE, which sounds like mostly QC's fault, albeit also you can't actually do much with it unless the implementation is 256 bit wide.
Cortex X925 comes to save the day. Hail Nvidia/Mediatek.
 

soresu

Platinum Member
Dec 19, 2014
2,941
2,164
136
From the ARM technical reference manual:

Cortex®-X925 core features​

You can use the Cortex®-X925 core in a standalone DynamIQ™ configuration where your homogenous DSU-120 DynamIQ™ cluster includes one or more Cortex®-X925 cores. You can also use the Cortex®-X925 core as the high-performance core in a heterogenous cluster.
Regardless of the cluster configuration, the Cortex®-X925 core always has the same features as described in the following lists.

Core features​

  • Implementation of the Arm®v9.2-A A64 instruction set
  • AArch64 Execution state at all Exception levels, EL0 to EL3
  • Memory Management Unit (MMU)
  • 40-bit Physical Address (PA) and 48-bit Virtual Address (VA)
  • Generic Interrupt Controller (GIC) CPU interface to connect to an external interrupt Distributor
  • Generic Timers interface that supports 64-bit count input from an external system counter
  • Implementation of the Reliability, Availability, and Serviceability (RAS) Extension
  • Implementation of the Scalable Vector Extension (SVE) with a 128-bit vector length and Scalable Vector Extension 2 (SVE2)
  • Integrated execution unit with Advanced Single Instruction Multiple Data (SIMD) and floating-point support
  • Activity Monitoring Unit (AMU)
  • Support for the optional Cryptographic Extension
Seems like just 128 bit.
 
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |