- Mar 3, 2017
- 1,684
- 6,227
- 136
That's a 28W part and I'm pretty sure M3m juices a whole lot more than that in cinememe.But still not matching M3 Max's MT as @Ghostsonplanets estimated.
MLID published slides that AMD themselves presented. Hypetrain conductors spammed some 3rd party (at best) estimations or even hallucinations. RDNA3 anyone?Apparently I woke up on the darkest timeline side of the bed today since I've learned that MLID was apparently correct.
Tech bros, I'm not feeling so well right now.
RDNA3 slides from AMD itself said 50% moar PPW.RDNA3 anyone?
its not the same as M3 Max. Different class. Need to compare with M3 pro 5 or 6 P core versionThis would put the Ryzen AI HX 170's MT performance at 1250 points (130% × 950).
That's a solid +40% improvement over Ryzen 7840HS.
But still not matching M3 Max's MT as @Ghostsonplanets estimated.
On the right side it says +47% vs Core Ultra 185H. According to CPU monkey that SKU scores 1070 pts in nT CB 2024, so 47% more is 1572pts (unless I fumbled something )Ladies and gentlemen, I detail to you the approximate Cinebench 2024 MT performance of Strix Point.
View attachment 100366
They say the Cinebench 2024 MT performance is 30% better than the X Elite.
But first, we need to to know which X Elite SKU it is. According to the footnotes (which somebody has previously posted in this thread), the X Elite SKU used here is the X1E-80-100. That thing has 4.0 GHz Dual Core Boost and 3.4 GHz all-core. It does about 950 points in Cinebench 2024 MT.
This would put the Ryzen AI HX 170's MT performance at 1250 points (130% × 950).
That's a solid +40% improvement over Ryzen 7840HS.
But still not matching M3 Max's MT as @Ghostsonplanets estimated.
I never said so. It was Ghost who was saying it was touching the nT of M3 Max.its not the same as M3 Max. Different class. Need to compare with M3 pro 5 or 6 P core version
Hmm🤔On the right side it says +47% vs Core Ultra 185H. According to CPU monkey that SKU scores 1070 pts in nT CB 2024, so 47% more is 1572pts (unless I fumbled something )
Sure, but Strix point has 50% more CPU cores than Phoenix.Ladies and gentlemen, I detail to you the approximate Cinebench 2024 MT performance of Strix Point.
View attachment 100366
They say the Cinebench 2024 MT performance is 30% better than the X Elite.
But first, we need to to know which X Elite SKU it is. According to the footnotes (which somebody has previously posted in this thread), the X Elite SKU used here is the X1E-80-100. That thing has 4.0 GHz Dual Core Boost and 3.4 GHz all-core. It does about 950 points in Cinebench 2024 MT.
View attachment 100368
This would put the Ryzen AI HX 170's MT performance at 1250 points (130% × 950).
That's a solid +40% improvement over Ryzen 7840HS.
But still not matching M3 Max's MT as @Ghostsonplanets estimated.
With half of them having a pretty ass v/f for a 30W part.Sure, but Strix point has 50% more CPU cores than Phoenix.
It’s accurate, it somewhat aligns with R23 result earlier. But at this point it’s using 80watts to reach that score.On the right side it says +47% vs Core Ultra 185H. According to CPU monkey that SKU scores 1070 pts in nT CB 2024, so 47% more is 1572pts (unless I fumbled something )
4+6, the SKU is literally on the AMD website.Btw is the cutdown strix point 4+6 or 2+8? I hope 4+6.
I dont know if anyone noticed that in the keynote AMD said that EPYC 5 ZEN 5 is build at 3nm for the Chiplets kai 6nm for the I/O , I was under the impression that ZEN 5 Ryzen is build at 4nm for the Chiplets.
If that is the case, it would be the first time in the ZEN history that Ryzen and EPYC are using different dies.
CPU monkey scores are usually in the best case range for all SKUs. So, that's there.On the right side it says +47% vs Core Ultra 185H. According to CPU monkey that SKU scores 1070 pts in nT CB 2024, so 47% more is 1572pts (unless I fumbled something )
I did come across it, and it was kind of hard to believe that it is nearly identical. So I was looking for 2nd opinion.You want it done from that picture or is this good enough?
Thanks, I appreciate it. You came very close to his numbers.I took a crack at scaling based on the image. The die size is pretty similar.
View attachment 100377
the core increased in area which will have an impact in notebook. Going by only the CCD die size can get misleading.So, in view of no change in die size, how does 16% IPC look? Pretty good IMO.
To clarify, the combination of small die size and improved power efficiency are fantastic news for notebook and server markets.
Lol, getting within 1mm2 is probably purely luck on my part since it depended on which pixel I measured from.Thanks, I appreciate it. You came very close to his numbers.
As far as official die size of Zen 4, it is alternatively listed as 70 mm2 and 71 mm2, so you are with 1 mm2, which is quite amazing.
Fwiw, TSMC N4P comes with a 6% reduction in area for logic in comparison to TSMC N5, but 16% higher IPC with a 26% larger core and presumably 34% more transistors (1.25*1.06 = 1.34) is right on the money with respect to the square law of scaling (square roof of 1.34 = 1.15). Not Zen 3 levels of improvement (19% higher IPC, 200 MHz higher max clocks, for 8% larger CCD) but at least Zen 5 isn't under par. I'm hoping that the 16% IPC uplift is true for general performance and not significantly skewed due to AVX512.So, in view of no change in die size, how does 16% IPC look? Pretty good IMO.
To clarify, the combination of small die size and improved power efficiency are fantastic news for notebook and server markets.
But less than fantastic for desktop / gaming market.