AB-testruns of Slow Chess 2.2 and Rubichess 1.7.3 finished, because I have a new machine: From now, I use an AMD Ryzen 3900 12-core notebook with 32GB RAM. Now, 20 games are played simultaneously (!), so from now, each testrun will have 6000 or 7000 games (instead of 5000 before) and will take only 2 days, not 6-7 days as before! From now, all engine-binaries are popcount/pext/avx2, of course, because bmi2-compiles are extremly slow on AMD. To keep the rating-list engine-names consistent, the "bmi2"-extension in the engine-name is still in use for older engines - otherwise ORDO will not calculate all played games by this engine as one engine...
https://www.sp-cc.de
(Perhaps you have to clear your browsercache or reload the website)
SPCC: Testruns of Slow Chess 2.2 and Rubichess 1.7.3 finished
Moderators: hgm, Rebel, chrisw
-
- Posts: 2435
- Joined: Sat Sep 03, 2011 7:25 am
- Location: Berlin, Germany
- Full name: Stefan Pohl
-
- Posts: 476
- Joined: Sun Mar 17, 2019 12:00 pm
- Full name: Henk Drost
Re: SPCC: Testruns of Slow Chess 2.2 and Rubichess 1.7.3 finished
Nice! You have a new notebook.
-
- Posts: 670
- Joined: Sun Jan 26, 2020 10:38 pm
- Location: Turkey
- Full name: Mehmet Karaman
Re: SPCC: Testruns of Slow Chess 2.2 and Rubichess 1.7.3 finished
Good tests with the world’s most powerful notebook.
https://www.techradar.com/news/the-worl ... -expensive
https://www.techradar.com/news/the-worl ... -expensive
-
- Posts: 6808
- Joined: Wed Nov 18, 2009 7:16 pm
- Location: Gutweiler, Germany
- Full name: Frank Quisinsky
Re: SPCC: Testruns of Slow Chess 2.2 and Rubichess 1.7.3 finished
Hi Stefan,
have fun with your new hardware and the test you like to do with it.
Best
Frank
have fun with your new hardware and the test you like to do with it.
Best
Frank
-
- Posts: 550
- Joined: Tue Nov 19, 2019 8:48 pm
- Full name: Alayan Feh
Re: SPCC: Testruns of Slow Chess 2.2 and Rubichess 1.7.3 finished
For your information, PEXT is an instruction that's part of the BMI2 instruction set. Calling a build "pext" or "bmi2" is equivalent, it uses the same instructions that are very slow on ryzen processors.pohl4711 wrote: ↑Tue Jul 21, 2020 1:15 pmFrom now, all engine-binaries are popcount/pext/avx2, of course, because bmi2-compiles are extremly slow on AMD. To keep the rating-list engine-names consistent, the "bmi2"-extension in the engine-name is still in use for older engines - otherwise ORDO will not calculate all played games by this engine as one engine...
For use with a ryzen CPU, you should use popcount compiles, not PEXT compiles.
-
- Posts: 1754
- Joined: Tue Apr 19, 2016 6:08 am
- Location: U.S.A
- Full name: Andrew Grant
Re: SPCC: Testruns of Slow Chess 2.2 and Rubichess 1.7.3 finished
+1Alayan wrote: ↑Mon Aug 03, 2020 2:13 pmFor your information, PEXT is an instruction that's part of the BMI2 instruction set. Calling a build "pext" or "bmi2" is equivalent, it uses the same instructions that are very slow on ryzen processors.pohl4711 wrote: ↑Tue Jul 21, 2020 1:15 pmFrom now, all engine-binaries are popcount/pext/avx2, of course, because bmi2-compiles are extremly slow on AMD. To keep the rating-list engine-names consistent, the "bmi2"-extension in the engine-name is still in use for older engines - otherwise ORDO will not calculate all played games by this engine as one engine...
For use with a ryzen CPU, you should use popcount compiles, not PEXT compiles.
#WeAreAllDraude #JusticeForDraude #RememberDraude #LeptirBigUltra
"Those who can't do, clone instead" - Eduard ( A real life friend, not this forum's Eduard )
"Those who can't do, clone instead" - Eduard ( A real life friend, not this forum's Eduard )
-
- Posts: 2435
- Joined: Sat Sep 03, 2011 7:25 am
- Location: Berlin, Germany
- Full name: Stefan Pohl
Re: SPCC: Testruns of Slow Chess 2.2 and Rubichess 1.7.3 finished
Very stupid mistake by me. From now, Ethereal will play with popcount-compile. Good news is, that the Ethereal 12.25-Testrun itself was earlier done on my old Intel-Notebook, were pext was the correct compile, so the distortion of Ethereals Elo and other Elonumbers should be very small.AndrewGrant wrote: ↑Mon Aug 03, 2020 3:04 pm+1Alayan wrote: ↑Mon Aug 03, 2020 2:13 pmFor your information, PEXT is an instruction that's part of the BMI2 instruction set. Calling a build "pext" or "bmi2" is equivalent, it uses the same instructions that are very slow on ryzen processors.pohl4711 wrote: ↑Tue Jul 21, 2020 1:15 pmFrom now, all engine-binaries are popcount/pext/avx2, of course, because bmi2-compiles are extremly slow on AMD. To keep the rating-list engine-names consistent, the "bmi2"-extension in the engine-name is still in use for older engines - otherwise ORDO will not calculate all played games by this engine as one engine...
For use with a ryzen CPU, you should use popcount compiles, not PEXT compiles.
For Houdini 6 and Komodo 14, I chose the right versions (popc for Houdini and Komodo 14 64-bit for Komodo). They are only named as pext/bmi2, because otherwise ORDO will not recognize that two different engine-names are the same engine. Same for Xiphos 0.6 and rofChade 2.3. Named still as bmi2, but sse-compile(Xiphos) and popcount (rofChade) is used on my new machine. Sorry, Andrew!
-
- Posts: 6442
- Joined: Tue Jan 09, 2007 12:31 am
- Location: PA USA
- Full name: Louis Zulli
Re: SPCC: Testruns of Slow Chess 2.2 and Rubichess 1.7.3 finished
On the website itself I see "From now, all engine-binaries are popcount/avx2, of course, because bmi2-compiles are extremly slow on AMD."Alayan wrote: ↑Mon Aug 03, 2020 2:13 pmFor your information, PEXT is an instruction that's part of the BMI2 instruction set. Calling a build "pext" or "bmi2" is equivalent, it uses the same instructions that are very slow on ryzen processors.pohl4711 wrote: ↑Tue Jul 21, 2020 1:15 pmFrom now, all engine-binaries are popcount/pext/avx2, of course, because bmi2-compiles are extremly slow on AMD. To keep the rating-list engine-names consistent, the "bmi2"-extension in the engine-name is still in use for older engines - otherwise ORDO will not calculate all played games by this engine as one engine...
For use with a ryzen CPU, you should use popcount compiles, not PEXT compiles.
So the binaries can contain popcnt and avx2 instructions, but not the pext instruction?
-
- Posts: 2435
- Joined: Sat Sep 03, 2011 7:25 am
- Location: Berlin, Germany
- Full name: Stefan Pohl
Re: SPCC: Testruns of Slow Chess 2.2 and Rubichess 1.7.3 finished
Yes. avx runs fast on my AMD Ryzen 9. And so does popcount. But bmi2/pext not - on AMD these compiles run measureable slower. My mistake was, to use the pext-compile of Ethereal 12.25 on my new AMD-notebook, because I did not mention bmi2=pext. Sorry again, but nobody is perfect.zullil wrote: ↑Mon Aug 03, 2020 6:46 pmOn the website itself I see "From now, all engine-binaries are popcount/avx2, of course, because bmi2-compiles are extremly slow on AMD."Alayan wrote: ↑Mon Aug 03, 2020 2:13 pmFor your information, PEXT is an instruction that's part of the BMI2 instruction set. Calling a build "pext" or "bmi2" is equivalent, it uses the same instructions that are very slow on ryzen processors.pohl4711 wrote: ↑Tue Jul 21, 2020 1:15 pmFrom now, all engine-binaries are popcount/pext/avx2, of course, because bmi2-compiles are extremly slow on AMD. To keep the rating-list engine-names consistent, the "bmi2"-extension in the engine-name is still in use for older engines - otherwise ORDO will not calculate all played games by this engine as one engine...
For use with a ryzen CPU, you should use popcount compiles, not PEXT compiles.
So the binaries can contain popcnt and avx2 instructions, but not the pext instruction?
-
- Posts: 6442
- Joined: Tue Jan 09, 2007 12:31 am
- Location: PA USA
- Full name: Louis Zulli
Re: SPCC: Testruns of Slow Chess 2.2 and Rubichess 1.7.3 finished
You need to be more careful. There's AVX and AVX2. Does any engine other than Stockfish+nnue use AVX2?pohl4711 wrote: ↑Mon Aug 03, 2020 7:03 pmYes. avx runs fast on my AMD Ryzen 9. And so does popcount. But bmi2/pext not - on AMD these compiles run measureable slower. My mistake was, to use the pext-compile of Ethereal 12.25 on my new AMD-notebook, because I did not mention bmi2=pext. Sorry again, but nobody is perfect.zullil wrote: ↑Mon Aug 03, 2020 6:46 pmOn the website itself I see "From now, all engine-binaries are popcount/avx2, of course, because bmi2-compiles are extremly slow on AMD."Alayan wrote: ↑Mon Aug 03, 2020 2:13 pmFor your information, PEXT is an instruction that's part of the BMI2 instruction set. Calling a build "pext" or "bmi2" is equivalent, it uses the same instructions that are very slow on ryzen processors.pohl4711 wrote: ↑Tue Jul 21, 2020 1:15 pmFrom now, all engine-binaries are popcount/pext/avx2, of course, because bmi2-compiles are extremly slow on AMD. To keep the rating-list engine-names consistent, the "bmi2"-extension in the engine-name is still in use for older engines - otherwise ORDO will not calculate all played games by this engine as one engine...
For use with a ryzen CPU, you should use popcount compiles, not PEXT compiles.
So the binaries can contain popcnt and avx2 instructions, but not the pext instruction?