Can somebody compare the AMD Ryzen processors to the intel processors

[Chessqueen]

Can somebody compare the AMD Ryzen processors to the intel processors engine benchmarks, I do NOT want to see any other processors on the list.

For instance:
Ryzen 5 3600 and Next to it the intel i5
AMD Ryzen 7 5800X ‎and Next to it the intel i7
AMD Ryzen 9 5950X Next to it the intel i9

[Ras]

http://ipmanchess.yolasite.com/amd---in ... -bench.php

Do your part of the work yourself.

[Chessqueen]

http://ipmanchess.yolasite.com/amd---in ... -bench.php

Do your part of the work yourself.

Thanks for providing that link, I see that AMD Ryzen is superior in comparison to the intel

78.574.335 AMD Ryzen 9 5950X @4.4Ghz 32threads
76.363.079 Intel Core i9 7980XE @4.7Ghz 4000Mhz 36threads
------------------------------------------------------------------------------------------
40.526.615 AMD Ryzen 7 5800X 16threads
31.954.375 Intel Core i7 10700K @4.6Ghz 12threads
------------------------------------------------------------------------------------------

[Modern Times]

There is also the Ryzen 9 5900X, 12 cores and 24 threads. Much more reasonably priced I suspect than their flagship 5950X.

[pohl4711]

http://ipmanchess.yolasite.com/amd---in ... -bench.php

Do your part of the work yourself.

These speed measurements are outdated, because they are done with an old asmFish engine: The nnue-engines run much faster on AMD Ryzen/Threadripper with an avx2-compile (+25%-30% more nodes!) compared to all Intel CPUs. So, for nnue-engines the Intel CPUs are much worse than they appear in that list.

[Modern Times]

These speed measurements are outdated, because they are done with an old asmFish engine: The nnue-engines run much faster on AMD Ryzen/Threadripper with an avx2-compile (+25%-30% more nodes!) compared to all Intel CPUs. So, for nnue-engines the Intel CPUs are much worse than they appear in that list.

That is interesting. You have both Intel and AMD so you would know. So that I understand what you're saying:

- Run a non-NNUE engine on the Intel and AMD and calculate a performance ratio
- Run an NNUE engine on the Intel and AMD and calculate a performance ratio

And the performance ratio is much higher for AMD when using NNUE engines ?
(I guess you could run this sort of test with current Stockfish with NNUE, and a previous Stockfish without NNUE)

[pohl4711]

Her my measurements on my notebooks. All singlethread:

All with latest Stockfish from abrok (March, 24), go depth 28

Intel Haswell i7-6700HQ (bmi2 binary):

SF nnue off: 1071209 nps
SF nnue on: 932672 nps
= using nnue is 87% speed compared to nnue off


AMD Ryzen 9 3900 (avx2 binary):

SF nnue off: 1399478 nps
SF nnue on: 1830004 nps (!!!)
= using nnue is 131% speed compared to nnue off

[MMarco]

AMD Ryzen 9 3900 (avx2 binary):

SF nnue off: 1399478 nps
SF nnue on: 1830004 nps (!!!)
= using nnue is 131% speed compared to nnue off

That looks too good to be true.

Here is what I get with my Ryzen 4900H and SF-13 avx2 (NNUE at 91% of classical speed):

Code: Select all

setoption name Hash value 1024
go movetime 60000
...
info depth 34 seldepth 42 multipv 1 score cp 37 nodes 91479854 nps 1524638 hashfull 627 tbhits 0 time 60001 pv e2e4 c7c5
bestmove e2e4 ponder c7c5

setoption name Hash value 1024
setoption name Use NNUE value false
go movetime 60000
...
info depth 31 seldepth 38 multipv 1 score cp 59 nodes 100402432 nps 1673345 hashfull 693 tbhits 0 time 60001 pv d2d4
bestmove d2d4 ponder e7e6

Intel 11th gen has avx 512 and vnni instructions. This is supposed to help with Stockfish (but apparently the processor gets hot).

[MikeB]

abrok compile 3/24/21 - go depth 28 , hash 1024 , single thread ,3970x, nnue == true, single thread

info depth 28 seldepth 38 multipv 1 score cp 19 nodes 31835290 nps 1756914 hashfull 249 tbhits 0 time 18120 pv e2e4 e7e5 g1f3 b8c6 f1b5 a7a6 b5a4 g8f6 e1g1 f6e4 d2d4 e5d4 f1e1 f7f5 b1d2 f8e7 d2e4 f5e4 f3d4 e8g8 d4c6 d7c6 a4b3 g8h8 d1d8 e7d8 e1e4 c8f5 e4f4 d8g5
bestmove e2e4 ponder e7e5

nnue=off

info depth 28 seldepth 40 multipv 1 score cp 55 nodes 65714777 nps 1949068 hashfull 514 tbhits 0 time 33716 pv d2d4 d7d5 c2c4 e7e6 b1c3 g8f6 g1f3 c7c5 c4d5 c5d4 f3d4 e6d5 c1g5 f8e7 e2e3 e8g8 f1e2 b8c6 e1g1 c8e6 g5f4 a7a6 a1c1 a8c8 d4e6 f7e6 f4g3 e7d6
bestmove d2d4 ponder d7d5

when I compile it myself everything is a little bit faster:

nnue==true
info depth 28 seldepth 38 multipv 1 score cp 19 nodes 31835290 nps 1811911 hashfull 249 tbhits 0 time 17570 pv e2e4 e7e5 g1f3 b8c6 f1b5 a7a6 b5a4 g8f6 e1g1 f6e4 d2d4 e5d4 f1e1 f7f5 b1d2 f8e7 d2e4 f5e4 f3d4 e8g8 d4c6 d7c6 a4b3 g8h8 d1d8 e7d8 e1e4 c8f5 e4f4 d8g5
bestmove e2e4 ponder e7e5

nnue==false
info depth 28 seldepth 40 multipv 1 score cp 55 nodes 65714777 nps 1966448 hashfull 514 tbhits 0 time 33418 pv d2d4 d7d5 c2c4 e7e6 b1c3 g8f6 g1f3 c7c5 c4d5 c5d4 f3d4 e6d5 c1g5 f8e7 e2e3 e8g8 f1e2 b8c6 e1g1 c8e6 g5f4 a7a6 a1c1 a8c8 d4e6 f7e6 f4g3 e7d6
bestmove d2d4 ponder d7d5
setoption name Use NNUE value true
setoption name Hash value 1024

[pohl4711]

AMD Ryzen 9 3900 (avx2 binary):

SF nnue off: 1399478 nps
SF nnue on: 1830004 nps (!!!)
= using nnue is 131% speed compared to nnue off

That looks too good to be true.

This is what I get on my Ryzen 9 3900. Fresh booted system, offline.