Werewolf wrote: ↑Sat Jul 02, 2022 7:50 pm
Good choice.
This isn't a perfect comparison by any means, but here are some Cinebench R23 CPU scores for interest:
Threadripper 3990X - 75671
Ryzen 5950X - 28782
Apple M1 Ultra (Mac Studio only) - 21740
Intel 10940X - 21309
Apple M1 Max (MacBook Pro) - 12422
Apple M1 Pro (MacBook Pro) - 12170
I notice the score you show is 'Multi-Score'. If you chose 'Single Score' it shifts Intel PCU's to the top - well above Threadripper.
What is the difference between the two...especially as concerns chess analysis?
I ask because Ras said (in another thread about chess analysis), "NPS is good, core split is bad. If you can have the same NPS either with a single fast core or with two slower cores, the former is always better. That's pretty obvious when considering how multithreading works in chess engines."
Werewolf wrote: ↑Sat Jul 02, 2022 7:50 pm
Good choice.
This isn't a perfect comparison by any means, but here are some Cinebench R23 CPU scores for interest:
Threadripper 3990X - 75671
Ryzen 5950X - 28782
Apple M1 Ultra (Mac Studio only) - 21740
Intel 10940X - 21309
Apple M1 Max (MacBook Pro) - 12422
Apple M1 Pro (MacBook Pro) - 12170
You have forgotten to write the watt usage behind the numbers
The MacBook M1 MAX, when running all parts at the limit, will take much lesser watt than the CPU from Intel or AMD
If you are looking at M1 MAX CPU cores only, the watt difference compared to AMD or Intel is so big that you don’t even need to compare.
We also understand that some people can feel the cold in their room with only 104°F, so they need an Intel or AMD CPU to reach at least 122°F.
I’m not talking about the noise, because I have non
It’s better to look here: https://ipmanchess.yolasite.com/amd--in ... ckfish.php
Werewolf wrote: ↑Sat Jul 02, 2022 7:50 pm
Good choice.
This isn't a perfect comparison by any means, but here are some Cinebench R23 CPU scores for interest:
Threadripper 3990X - 75671
Ryzen 5950X - 28782
Apple M1 Ultra (Mac Studio only) - 21740
Intel 10940X - 21309
Apple M1 Max (MacBook Pro) - 12422
Apple M1 Pro (MacBook Pro) - 12170
I notice the score you show is 'Multi-Score'. If you chose 'Single Score' it shifts Intel PCU's to the top - well above Threadripper.
What is the difference between the two...especially as concerns chess analysis?
I ask because Ras said (in another thread about chess analysis), "NPS is good, core split is bad. If you can have the same NPS either with a single fast core or with two slower cores, the former is always better. That's pretty obvious when considering how multithreading works in chess engines."
The multi-thread score is far more important, because chess engines can use all available threads.
However, regarding NPS on chess engines: if two CPUs have a similar NPS but one CPU has fewer cores (so fewer available threads) that usually means it is a bit quicker as fewer threads incurs less search inefficiency.
But really, if you just want the simple, rough answer: more nps = better. And for that many threads is better than one.
The Cinebench scores have nothing to do with chess, but they do give an idea of how fast the CPU is on one thread and all its threads. We care about the latter.
What do I get with my Ryzen 9 5900X? Let's take a look.
My System uses 290W at full CPU load inkl. a 32" Monitor (29W). CPU running on 4,4 - 4,6Ghz, optimied with Hydra from Yuri Bubliy.
Stockfishbench gives me 54.726.000 nps -
so 54.726.000/290 = 188.710nps for 1 Watt
now the Mac M1 Pro Max
M1 Pro Max at full high power mode 110W (test on Mtech.news)
Stockfishbench 12.313.063nps (test on Ipman)
so 12.313.063/110 = 111.937nps for 1 Watt
Comparison for chess only...
regards William
Ryzen 5900x, 64GB Ram, RTX3080, 2x 2TB SSD, 1x 1TB SSD, 1x 2TB HDD, all custom watercooled, Win10 and Win7 Multiboot, Win98/Windows Millennium/WinXP virtual machines for 16bit
Retro Chess Chessmasterhttp://descent.at/files/
Wilhelm wrote: ↑Wed Jul 13, 2022 4:36 pm
What do I get with my Ryzen 9 5900X? Let's take a look.
My System uses 290W at full CPU load inkl. a 32" Monitor (29W). CPU running on 4,4 - 4,6Ghz, optimied with Hydra from Yuri Bubliy.
That 290W seems very high. The CPU itself uses 105W, maybe as much as 142W, where is the rest ?
Magnum wrote: ↑Sat Jul 02, 2022 11:16 pm
You have forgotten to write the watt usage behind the numbers
Please add your watt numbers then.. Only an idiot or a salesguy at apple would care about this (and only to whitewash that performance is sub par).. If you care about watt and not about chess, then buy a 4nm ARM phone and run SF on it FFS...
Like yea I gett 1 FPS in my game but I am so happy because my fans does not spin up... Wow how cool...
Wilhelm wrote: ↑Wed Jul 13, 2022 4:36 pm
What do I get with my Ryzen 9 5900X? Let's take a look.
My System uses 290W at full CPU load inkl. a 32" Monitor (29W). CPU running on 4,4 - 4,6Ghz, optimied with Hydra from Yuri Bubliy.
Stockfishbench gives me 54.726.000 nps -
so 54.726.000/290 = 188.710nps for 1 Watt
now the Mac M1 Pro Max
M1 Pro Max at full high power mode 110W (test on Mtech.news)
Stockfishbench 12.313.063nps (test on Ipman)
so 12.313.063/110 = 111.937nps for 1 Watt
Comparison for chess only...
regards William
Stockfish engine is far away from being optimized for Apple M1 devices.
Magnum wrote: ↑Wed Jul 13, 2022 8:30 pm
I’m so far away from 110W on my MacBook Pro M1 MAX
The MacBook is cool and quiet and it runs Stockfish 24 hours every day
Stockfish with profile build is more optimized for [Moderation] please mind your language M1 than for Dimensity 9000+ CPUs and still, the Dimensity uses only 4w. Why did you buy this power-consuming overpriced mobile chipset from Apple, when there are so much more efficient tiny computers available for chess for less.
The M1 is underperforming other ARM alternatives with its power consumption, applying your faulty logic.
Also, no one should consider (applying stupid watt-logic) the higher-end Ultras and M2s as they draw an extreme amount of performance/watt compared to an M1 Air...
Besides, you are not that far away from 110 W if you measure Watt on a [Moderation] please mind your language Pro...
Running aztek+Aztec+511MT benchjmark on a Pro 16 sucks 120W (92W package power)..
gordonr wrote: ↑Sat Jul 02, 2022 3:53 pm
Thanks everyone for your help - really appreciated. I think I will go for the AMD 5950X.
If I do go for this, should I keep hyperthreading on or off in the BIOS? Do people sometimes find a benefit to running more threads than physical cores?
Probably little real difference.
For what it's worth here is a screenshot from mine.
