Discussion AMD's Soundwave ARM APU: The Beginning of Transformation !!!

Page 24 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Jul 27, 2020
28,162
19,187
146
So they should hurt the market where they are leading the duopoly in favour of the market where they will be one of many vendors for little boost to revenue that they will have to share with ARM anyway?;)
But if they sit back, the other ARM vendors will hurt the x86 market anyway. Better get in there early and grab as many cents as they can! :)
 

DrMrLordX

Lifer
Apr 27, 2000
22,925
12,995
136
So they should hurt the market where they are leading the duopoly in favour of the market where they will be one of many vendors for little boost to revenue that they will have to share with ARM anyway?;)
Nobody should willingly become a thrall to ARM that can avoid doing so. And AMD can certainly avoid that.
 

Tigerick

Senior member
Apr 1, 2022
851
802
106
Even though A19 Pro's FP16 shown 85% boost, Geekerwan's benchmarking of 3 real games shown average 54% faster FPS. Thus, I am assuming new RDNA architecture will boost average 40% in FPS, not as high as SNL but more realistic estimation.

AMD is definitely working on new shader architecture, namely GFX125x. As for any new design which boost FPS, AMD will try to hide the technique especially with new ARM platform. Therefore, I am creating tables below trying to demystify the ISA codenames. If anyone has any new info, feel free to pitch in. :D

GFX11.5 - RDNA 3.5CUShader ISA
Strix Point16gfx1150
Strix Halo40gfx1151
Kracken Point8gfx1152
GFX12.0 - RDNA 4
N4432gfx1200
N4864gfx1201



GFX12.5 - RDNA 4.5?PlatformCU (New ISA)CU (~Equal to Old ISA)Shader ISA
Exynos 2600ARM8 CU12
SoundwaveARM16 CU?24
Medusa Point MDS1x868 CU12
Medusa Point MDS2x864 CU6
Medusa Point MDS3x862 CU3



GFX13.0 - RDNA 5?PlatformCU (New ISA)CU (~Equal to Old ISA)Shader ISA
Medusa Premiumx862436
ARM + AT4 ?ARM2436
Medusa Halox864872
ARM + AT3 ?ARM4872
AT2x866496
AT0x86154231

PS: Almost all the above new SoCs / GPUs are targeting to be released in 2027; well except E2600 and Soundwave....

Phoronix has updated the LLVM info about GFX125x: AMD has added one more APU model to list. That makes two APU with RDNA4.5 ISA: GFX1250 and GFX1251. The site specifically mentioned APU term; thus, it should be similar to STX, KRK and Halo. It might refer to upcoming Soundwave and E2600 because there are only 2 APUs which will be released soon.

In case people still don't get it, Soundwave is considered ARM version of Strix Point that is fabbed by N3P node. It won't support SMT cause AMD is using off the shelf ARM IP, namely C1 series as part of Niva CSS. However AMD will use new generation of RDNA4.5? as speculated above. Even though C1 won't support SMT and STX have 24 threads, the final benchmark scores might show difference results:

X2 Elite ExtremeIntel Ultra 7 268VAMD Ryzen 9 9950XAMD Ryzen 9 HX375iPad Pro M5M5 Compare to STXMediatek D9500D9500 Compare to STX
CodenameLunar LakeZen 5Strix PointM5D9500 5G
Core Count12 + 6 without SMT4P + 4LPe without SMT16P with SMT4P + 8E with SMT3P + 6E without SMT1 + 3 + 4 without SMT
Threads188322498
Max Clock Speed5 GHz5 GHz5.7 GHz5.1 GHz4.42 GHz4.21 GHz
TDP~ 40 W17 W170 W28 - 54 W~14 W2 - 3.9X Power Efficiency~14W2 - 3.9X Power Efficiency
GB6 ST40832729338528564133+ 45%3700+ 30%
Performance Per Clock817546594560935+ 67%879+ 57%
GB6 MT2334910025214301436715437+ 7.5%11000- 30%
Performance Per Thread129712536705991715+ 286%1375+ 230%
M4 (10-core)
C2024 MT196761523401130962
Performance Per Thread1097773.147.196.2
SNL56483555NA3600+~ 5000+3100+

Soundwave will be competing with M5 MacBook Pro and Air; even though C1 might not score as high as M5. And that's why AMD is switching to ARM platform; I really hope members here won't be so senseless about the coming changes. Cheers !!!
 
Last edited:
  • Like
Reactions: perry mason

Kepler_L2

Golden Member
Sep 6, 2020
1,001
4,287
136

Phoronix has updated the LLVM info about GFX125x: AMD has added one more APU model to list. That makes two APU with RDNA4.5 ISA: GFX1250 and GFX1251. The site specifically mentioned APU term; thus, it should be similar to STX, KRK and Halo. It might refer to upcoming Soundwave and E2600 because there are only 2 APUs which will be released soon.

In case people still don't get it, Soundwave is considered ARM version of Strix Point that is fabbed by N3P node. It won't support SMT cause AMD is using off the shelf ARM IP, namely C1 series as part of Niva CSS. However AMD will use new generation of RDNA4.5? as speculated above. Even though C1 won't support SMT and STX have 24 threads, the final benchmark scores might show difference results:

AMD Ryzen 9 HX375iPad Pro M5
CodenameStrix PointM5
Core Count4P + 8E with SMT3P + 6E without SMT
Threads249
GB6 ST28564133
GB6 MT1436715437

Soundwave will be competing with M5 MacBook Pro and Air; even though C1 might not score as high as M5. And that's why AMD is switching to ARM platform; I really hope members here won't be so senseless about the coming changes. Cheers !!!
gfx1251 is MI430X, HPC version of MI400
 

Tigerick

Senior member
Apr 1, 2022
851
802
106

Phoronix has updated the LLVM info about GFX125x: AMD has added one more APU model to list. That makes two APU with RDNA4.5 ISA: GFX1250 and GFX1251. The site specifically mentioned APU term; thus, it should be similar to STX, KRK and Halo. It might refer to upcoming Soundwave and E2600 because there are only 2 APUs which will be released soon.

In case people still don't get it, Soundwave is considered ARM version of Strix Point that is fabbed by N3P node. It won't support SMT cause AMD is using off the shelf ARM IP, namely C1 series as part of Niva CSS. However AMD will use new generation of RDNA4.5? as speculated above. Even though C1 won't support SMT and STX have 24 threads, the final benchmark scores might show difference results:

AMD Ryzen 9 HX375iPad Pro M5M5 Compare to STXMediatek D9500D9500 Compare to STX
CodenameStrix PointM5D9500 5G
Core Count4P + 8E with SMT3P + 6E without SMT1 + 3 + 4 without SMT
Threads2498
Max Clock Speed5.1GHz4.42 GHz4.21 GHz
TDP28 - 54 W~14 W2 - 3.9X Power Efficiency~14W2 - 3.9X Power Efficiency
GB6 ST28564133+ 45%3700+ 30%
Performance Per Clock560935+ 67%879+ 57%
GB6 MT1436715437+ 7.5%11000- 30%
Performance Per Thread5991715+ 286%1375+ 230%

Soundwave will be competing with M5 MacBook Pro and Air; even though C1 might not score as high as M5. And that's why AMD is switching to ARM platform; I really hope members here won't be so senseless about the coming changes. Cheers !!!
Hoho, as expected, there are few short-sighted members mocking my thread even though I show the table with numbers. I have updated the table to include more numbers; I will let you all have an own judgement.

Remember Lisa told us to expect 100X power efficiency. And no, Venice fabbed by N2 won't save that, they still require up to 700W per socket to power 256-core. The only way to achieve 100X power efficiency is switching to ARM as shown in the table above. ARM SoC have become so powerful (ST and MT) and yet still maintaining same power envelope. x86 has to juice up to 54W in order to reach 5.1GHz and yet the performance is subpar. And no Zen6 APU like Medusa Point won't change that; the high clock speed commitment undermine the potential of GPU and NPU. CPU cores need to add more extension not clock speed; it has been cursed by Intel since long time ago. That's why Intel has to disable HT and AVX-512 to achieve that with PTL. Of course, poor 18A node is part of the contribution factor.

Most of the leakers has access to the same specs of Soundwave: 2P+4E with 4CU. Again, don't believe blindly in the roadmap. If you believe AMD is naive enough to offer such APU, well you shouldn't be here. :p

If anyone wants to mock my thread, here is your last chance cause AMD is about to offer new roadmap about switching to ARM platform after Zen6. :cool:
 
Last edited:

eek2121

Diamond Member
Aug 2, 2005
3,415
5,056
136
Hoho, as expected, there are few short-sighted members mocking my thread even though I show the table with numbers. I have updated the table to include more numbers; I will let you all have an own judgement.

Remember Lisa told us to expect 100X power efficiency. And no, Venice fabbed by N2 won't save that, they still require up to 700W per socket to power 256-core. The only way to achieve 100X power efficiency is switching to ARM as shown in the table above. ARM SoC have become so powerful (ST and MT) and yet still maintaining same power envelope. x86 has to juice up to 54W in order to reach 5.1GHz and yet the performance is subpar. And no Zen6 APU like Medusa Point won't change that; the high clock speed commitment undermine the potential of GPU and NPU. CPU cores need to add more extension not clock speed; it has been cursed by Intel since long time ago. That's why Intel has to disable HT and AVX-512 to achieve that with PTL. Of course, poor 18A node is part of the contribution factor.

Most of the leakers has access to the same specs of Soundwave: 2P+4E with 4CU. Again, don't believe blindly in the roadmap. If you believe AMD is naive enough to offer such APU, well you shouldn't be here. :p

If anyone wants to mock my thread, here is your last chance cause AMD is about to offer new roadmap about switching to ARM platform after Zen6. :cool:

Efficiency != watts consumed.
 

Covfefe

Member
Jul 23, 2025
51
76
51
Hoho, as expected, there are few short-sighted members mocking my thread even though I show the table with numbers. I have updated the table to include more numbers; I will let you all have an own judgement.

Remember Lisa told us to expect 100X power efficiency. And no, Venice fabbed by N2 won't save that, they still require up to 700W per socket to power 256-core. The only way to achieve 100X power efficiency is switching to ARM as shown in the table above. ARM SoC have become so powerful (ST and MT) and yet still maintaining same power envelope. x86 has to juice up to 54W in order to reach 5.1GHz and yet the performance is subpar. And no Zen6 APU like Medusa Point won't change that; the high clock speed commitment undermine the potential of GPU and NPU. CPU cores need to add more extension not clock speed; it has been cursed by Intel since long time ago. That's why Intel has to disable HT and AVX-512 to achieve that with PTL. Of course, poor 18A node is part of the contribution factor.

Most of the leakers has access to the same specs of Soundwave: 2P+4E with 4CU. Again, don't believe blindly in the roadmap. If you believe AMD is naive enough to offer such APU, well you shouldn't be here. :p

If anyone wants to mock my thread, here is your last chance cause AMD is about to offer new roadmap about switching to ARM platform after Zen6. :cool:

Lisa Su's 100x efficiency claim was about AI accelerators, not CPUs.

 
  • Like
Reactions: Tlh97 and marees

Tigerick

Senior member
Apr 1, 2022
851
802
106
AMD-SONY-RDNA5-3.jpg
Hoho, have you wondered why Sony and AMD are pre-annoucing new RDNA5 (filename itself) architectures with official terms?

Both RDNA5 GPUs and PS6 are about two years away. Why the rush?

Well, if you have watched my thread, then you should know Soundwave's iGPU should be based on new RDNA ISA, which should be RDNA4.5: subset/mobile version of RDNA5. Sony just want to be the first one to announce the architecture which will be common features for all future RDNA4.5/5 GPU. Yep, that's mean Soundwave and WoA12 launching should be imminent.... :cool: Stay tuned for more updates.
 
  • Wow
Reactions: igor_kavinski

reaperrr3

Member
May 31, 2024
133
377
96
If anyone wants to mock my thread, here is your last chance cause AMD is about to offer new roadmap about switching to ARM platform after Zen6.:cool:
🤣
...The date you joined the forum is very fitting 😅

RDNA4.5: subset/mobile version of RDNA5.
That's not how RDNA's versioning works.

Yep, that's mean Soundwave and WoA12 launching should be imminent.... :cool: Stay tuned for more updates.
:laughing:

All in all, you remind me of someone from beyond3d forums (was it juanrga?).
Never listening to facts presented by others, always spinning your own tale, no matter how little sense it makes.
 
  • Like
Reactions: Tlh97 and marees

itsmydamnation

Diamond Member
Feb 6, 2011
3,075
3,904
136
🤣
...The date you joined the forum is very fitting 😅


That's not how RDNA's versioning works.


:laughing:

All in all, you remind me of someone from beyond3d forums (was it juanrga?).
Never listening to facts presented by others, always spinning your own tale, no matter how little sense it makes.
wasnt juanrga mainly the old semi accurate forums...... i miss that forum :(
 

Doug S

Diamond Member
Feb 8, 2020
3,588
6,348
136
So they should hurt the market where they are leading the duopoly in favour of the market where they will be one of many vendors for little boost to revenue that they will have to share with ARM anyway?;)

💯

Unless/until Intel gets their act together and takes back the crown, AMD is now the "face" of x86. Just like old dominant Intel was never going to do anything to risk the dominance of x86 on Windows, AMD has to think the same way.

Sure if Windows on ARM starts gaining real market share they have to be prepared to jump on board the train. But since that is IMHO unlikely they should not want to do anything to help legitimize ARM in the eyes of the market by making it a first class citizen in their product lineup.
 

marees

Golden Member
Apr 28, 2024
1,777
2,397
96
💯

Unless/until Intel gets their act together and takes back the crown, AMD is now the "face" of x86. Just like old dominant Intel was never going to do anything to risk the dominance of x86 on Windows, AMD has to think the same way.

Sure if Windows on ARM starts gaining real market share they have to be prepared to jump on board the train. But since that is IMHO unlikely they should not want to do anything to help legitimize ARM in the eyes of the market by making it a first class citizen in their product lineup.
Apparently they were strong-armed by Microsoft into doing this !!

But with zen doing so well I don't see the need for ARM for AMD

Microsoft may not like being dependent on x86 but that's the legacy of windows now