mwyoung wrote: ↑Tue Dec 29, 2020 11:13 pmAfter 100 games you should running it against StoofvleesChessqueen wrote: ↑Tue Dec 29, 2020 10:33 pmStockfish 2 cpu vs Komodo 32 threads. +0 =28 -3 -34 Elo....mwyoung wrote: ↑Tue Dec 29, 2020 6:50 pmI can NOT follow who is winning, but I believe that Komodo Dragon on 32 threads is way faster than Droidfish on 2 cores, probably Komodo Dragon on 16 threads would be more reasonable or an even match. I might be wrong but what is the score so far. And since you are NOT using tablebase on Droidfish, you should NOT use TB for Komodo Dragon as wellChessqueen wrote: ↑Tue Dec 29, 2020 6:43 pmLive Chess Match Stockfish 280920 (2 cores, No TB) vs Dragon(32 Threads) (TC=3m+2s)mwyoung wrote: ↑Tue Dec 29, 2020 6:29 pm"Who is keeping track of the Cell Phone engines ratings ?"mwyoung wrote: ↑Tue Dec 29, 2020 9:28 amThe test is running.mwyoung wrote: ↑Tue Dec 29, 2020 8:24 amThere is no reason you need a cell phone to play the games. And play the games manually.Think outside the box. Stockfish is Stockfish on a phone or pc. What matters is the speed.M ANSARI wrote: ↑Tue Dec 29, 2020 8:03 am I would be really interested to know what the strength of engines on mobile platforms are today, especially with NN now being available. I would think that NN is the equivalent of massive hardware boost, which should help the limited resources offered by mobile platform. If you consider that massive boost with the fact that mobile processing has also become very powerful, I think it would make a huge improvement and maybe the engines on mobile are already at or above 3200 ELO. I guess someone with an older setup (say quadcore) could play a manual game and test that. If I remember correctly Rybka 3 was around 3200 ELO on quadcore at that time ... would be an interesting matchup ... Rybka 3 on Quadcore vs. latest SF NNUE on Iphone 12! I have a feeling that this would be easily won by SF NNUE even on mobile platform.
And to accurately determain the speed of a cell phone cpu. So you can make a equivalent mobile equivalent on your pc. You would most diffently would not want to use nodes per second.
You would need to dial in the speed using time to depth.
I will run such a test. And would be shocked if Stockfish is only 3200 Elo.
And we have a better version of Rybka with a very good ccrl rating called Fritz 17. Rybka 3 would be much too weak.
Here is the setup.
TC=3m+2s
Fritz 17 on 4 CPU's with TB's 6 man. CCRL Rating = Fritz 17 64-bit 4CPU 3381
PC equivalent version of Droidfish running Stockfish 280920.
Using a Samsung S20 Ultra 8 cores = 2 PC cores on a AMD 2950x.
I have no table bases on my phone. So Stockfish has no table bases in this test.
Results will be posted......
CCRL is keeping track, or any testing site that tests with 1 core.
1 core test results will be a very good analog for many many smart phones, and tablets running at full power.
Correction
Fritz 17 4 cores vs SF 12 1 core.Code: Select all
Stockfish 12 64-bit 3639 +15 −14 (+258)
My results
CCRL results Fritz 17 (4 cores) vs SF 12 (1 core)Code: Select all
DESKTOP-CORSAIR, Blitz 3.0min+2.0sec 0 1 Stockfish 280920 +331 +40/=14/-0 87.04% 47.0/54 2 Fritz 17 Popcnt -331 +0/=14/-40 12.96% 7.0/54
Now lets get shocked again. Let us see how my smartphone S20 Ultra running Droidfish will perform against Dragon at full power (32 threads).Code: Select all
Stockfish 12 64-bit 8CPU 3692 +15 −14 (+311) 7.5 − 83.5 (+0−76=15)
Live Stream:
Who is keeping track of the Cell Phone engines ratings ?
Moderators: hgm, Rebel, chrisw
-
- Posts: 5589
- Joined: Wed Sep 05, 2018 2:16 am
- Location: Moving
- Full name: Jorge Picado
Re: Who is keeping track of the Cell Phone engines ratings ?
Who is 17 years old GM Gukesh 2nd at the Candidate in Toronto?
https://indianexpress.com/article/sport ... t-9281394/
https://indianexpress.com/article/sport ... t-9281394/
-
- Posts: 4190
- Joined: Wed Nov 25, 2009 1:47 am
Re: Who is keeping track of the Cell Phone engines ratings ?
What are you talking about? This whole thread is a full phone (whichever number of threads it has) vs a single core CPU. And the high-end phones (with usually 4 slow and 4 fast threads) are on par if not better than a single core of a high-end CPU.MikeB wrote: ↑Tue Dec 29, 2020 3:41 amThat would surprise me — are you using one thread/core as indicated? Please double check your settings. Using one thread with NNUE enabled.Milos wrote: ↑Tue Dec 29, 2020 2:54 amI have a 180$ Chinese phone Xiaomi Redme Note 8 Pro with a mediocre CPU for current standards and I get 1565knps from a starting position with latest SF NNUE Droidfish. I'd expect most of high-end phones would actually be well over 2Mnps.MikeB wrote: ↑Mon Dec 28, 2020 7:18 pmMost certainly, most high end phones will see 800 to 1000k/nps with SF NNUE based engine, that would be around 30 to 40% of a recent single core CPU. Clearly less than 200 Elo difference.Chessqueen wrote: ↑Mon Dec 28, 2020 3:41 pm Is there any place that Cell Phone latest Engines rating is kept Updated, and how far are they from computer Engines ratings, Are they within 200 ratings points from Single chip or one core computer
There is absolutely no phone that can do 800knps on a single thread with SF-NNUE.
-
- Posts: 5589
- Joined: Wed Sep 05, 2018 2:16 am
- Location: Moving
- Full name: Jorge Picado
Re: Who is keeping track of the Cell Phone engines ratings ?
Anyway this match of Stockfish on Samsung Galaxy S20 vs Komodo Dragon on 32 threads is more interesting than even TCEC, who would ever thought that Cell phone Droidfish was so powerful, therefore this Samsung Galaxy S20 with droidfish can give Magnus Carlsen either the f7 pawn or f2 pawn on standard time control or long time control and destroy the championMilos wrote: ↑Wed Dec 30, 2020 2:53 amWhat are you talking about? This whole thread is a full phone (whichever number of threads it has) vs a single core CPU. And the high-end phones (with usually 4 slow and 4 fast threads) are on par if not better than a single core of a high-end CPU.MikeB wrote: ↑Tue Dec 29, 2020 3:41 amThat would surprise me — are you using one thread/core as indicated? Please double check your settings. Using one thread with NNUE enabled.Milos wrote: ↑Tue Dec 29, 2020 2:54 amI have a 180$ Chinese phone Xiaomi Redme Note 8 Pro with a mediocre CPU for current standards and I get 1565knps from a starting position with latest SF NNUE Droidfish. I'd expect most of high-end phones would actually be well over 2Mnps.MikeB wrote: ↑Mon Dec 28, 2020 7:18 pmMost certainly, most high end phones will see 800 to 1000k/nps with SF NNUE based engine, that would be around 30 to 40% of a recent single core CPU. Clearly less than 200 Elo difference.Chessqueen wrote: ↑Mon Dec 28, 2020 3:41 pm Is there any place that Cell Phone latest Engines rating is kept Updated, and how far are they from computer Engines ratings, Are they within 200 ratings points from Single chip or one core computer
There is absolutely no phone that can do 800knps on a single thread with SF-NNUE.
Who is 17 years old GM Gukesh 2nd at the Candidate in Toronto?
https://indianexpress.com/article/sport ... t-9281394/
https://indianexpress.com/article/sport ... t-9281394/
-
- Posts: 4190
- Joined: Wed Nov 25, 2009 1:47 am
Re: Who is keeping track of the Cell Phone engines ratings ?
What's the fascination and surprise?Chessqueen wrote: ↑Wed Dec 30, 2020 3:39 amAnyway this match of Stockfish on Samsung Galaxy S20 vs Komodo Dragon on 32 threads is more interesting than even TCEC, who would ever thought that Cell phone Droidfish was so powerfulMilos wrote: ↑Wed Dec 30, 2020 2:53 amWhat are you talking about? This whole thread is a full phone (whichever number of threads it has) vs a single core CPU. And the high-end phones (with usually 4 slow and 4 fast threads) are on par if not better than a single core of a high-end CPU.MikeB wrote: ↑Tue Dec 29, 2020 3:41 amThat would surprise me — are you using one thread/core as indicated? Please double check your settings. Using one thread with NNUE enabled.Milos wrote: ↑Tue Dec 29, 2020 2:54 amI have a 180$ Chinese phone Xiaomi Redme Note 8 Pro with a mediocre CPU for current standards and I get 1565knps from a starting position with latest SF NNUE Droidfish. I'd expect most of high-end phones would actually be well over 2Mnps.MikeB wrote: ↑Mon Dec 28, 2020 7:18 pmMost certainly, most high end phones will see 800 to 1000k/nps with SF NNUE based engine, that would be around 30 to 40% of a recent single core CPU. Clearly less than 200 Elo difference.Chessqueen wrote: ↑Mon Dec 28, 2020 3:41 pm Is there any place that Cell Phone latest Engines rating is kept Updated, and how far are they from computer Engines ratings, Are they within 200 ratings points from Single chip or one core computer
There is absolutely no phone that can do 800knps on a single thread with SF-NNUE.
Nps is nps it doesn't matter whether on the phone or on the PC. SF on S20 is equivalent to two average CPU cores (of 2950x in this case). Actually it's a bit weaker coz on 8 cores you lose more than on 2 real cores due to SMP. But for the sake of argument it's easy to calculate Komodo Dragon gets around 80Elo going from 2 cores to 32 cores. And on 2 cores current SF is about 50Elo stronger than Komodo Dragon on 2 cores. So net result around 30Elo for Komodo Dragon.
Last edited by Milos on Wed Dec 30, 2020 3:55 am, edited 1 time in total.
-
- Posts: 2727
- Joined: Wed May 12, 2010 10:00 pm
Re: Who is keeping track of the Cell Phone engines ratings ?
Cell phones are very good today. But the truth is NNUE is the Magic here. NNUE may scale like crap, but it finds really good moves with minimal CPU power.Chessqueen wrote: ↑Wed Dec 30, 2020 3:39 amAnyway this match of Stockfish on Samsung Galaxy S20 vs Komodo Dragon on 32 threads is more interesting than even TCEC, who would ever thought that Cell phone Droidfish was so powerful, therefore this Samsung Galaxy S20 with droidfish can give Magnus Carlsen either the f7 pawn or f2 pawn on standard time control or long time control and destroy the championMilos wrote: ↑Wed Dec 30, 2020 2:53 amWhat are you talking about? This whole thread is a full phone (whichever number of threads it has) vs a single core CPU. And the high-end phones (with usually 4 slow and 4 fast threads) are on par if not better than a single core of a high-end CPU.MikeB wrote: ↑Tue Dec 29, 2020 3:41 amThat would surprise me — are you using one thread/core as indicated? Please double check your settings. Using one thread with NNUE enabled.Milos wrote: ↑Tue Dec 29, 2020 2:54 amI have a 180$ Chinese phone Xiaomi Redme Note 8 Pro with a mediocre CPU for current standards and I get 1565knps from a starting position with latest SF NNUE Droidfish. I'd expect most of high-end phones would actually be well over 2Mnps.MikeB wrote: ↑Mon Dec 28, 2020 7:18 pmMost certainly, most high end phones will see 800 to 1000k/nps with SF NNUE based engine, that would be around 30 to 40% of a recent single core CPU. Clearly less than 200 Elo difference.Chessqueen wrote: ↑Mon Dec 28, 2020 3:41 pm Is there any place that Cell Phone latest Engines rating is kept Updated, and how far are they from computer Engines ratings, Are they within 200 ratings points from Single chip or one core computer
There is absolutely no phone that can do 800knps on a single thread with SF-NNUE.
And no A/B engine can tough it, even when given massive hardware advantages.
"The worst thing that can happen to a forum is a running wild attacking moderator(HGM) who is not corrected by the community." - Ed Schröder
But my words like silent raindrops fell. And echoed in the wells of silence.
But my words like silent raindrops fell. And echoed in the wells of silence.
-
- Posts: 2727
- Joined: Wed May 12, 2010 10:00 pm
Re: Who is keeping track of the Cell Phone engines ratings ?
I am not sure about that. I am getting on 1 core with NNUE on a Samsung S20 Ultra almost 700 KN/s. And I know the Snapdragon CPU is not the fastest single core performer. Apple phones can be even faster on a single core.Milos wrote: ↑Wed Dec 30, 2020 2:53 amWhat are you talking about? This whole thread is a full phone (whichever number of threads it has) vs a single core CPU. And the high-end phones (with usually 4 slow and 4 fast threads) are on par if not better than a single core of a high-end CPU.MikeB wrote: ↑Tue Dec 29, 2020 3:41 amThat would surprise me — are you using one thread/core as indicated? Please double check your settings. Using one thread with NNUE enabled.Milos wrote: ↑Tue Dec 29, 2020 2:54 amI have a 180$ Chinese phone Xiaomi Redme Note 8 Pro with a mediocre CPU for current standards and I get 1565knps from a starting position with latest SF NNUE Droidfish. I'd expect most of high-end phones would actually be well over 2Mnps.MikeB wrote: ↑Mon Dec 28, 2020 7:18 pmMost certainly, most high end phones will see 800 to 1000k/nps with SF NNUE based engine, that would be around 30 to 40% of a recent single core CPU. Clearly less than 200 Elo difference.Chessqueen wrote: ↑Mon Dec 28, 2020 3:41 pm Is there any place that Cell Phone latest Engines rating is kept Updated, and how far are they from computer Engines ratings, Are they within 200 ratings points from Single chip or one core computer
There is absolutely no phone that can do 800knps on a single thread with SF-NNUE.
"The worst thing that can happen to a forum is a running wild attacking moderator(HGM) who is not corrected by the community." - Ed Schröder
But my words like silent raindrops fell. And echoed in the wells of silence.
But my words like silent raindrops fell. And echoed in the wells of silence.
-
- Posts: 2727
- Joined: Wed May 12, 2010 10:00 pm
Re: Who is keeping track of the Cell Phone engines ratings ?
End of match.mwyoung wrote: ↑Tue Dec 29, 2020 6:29 pm"Who is keeping track of the Cell Phone engines ratings ?"mwyoung wrote: ↑Tue Dec 29, 2020 9:28 amThe test is running.mwyoung wrote: ↑Tue Dec 29, 2020 8:24 amThere is no reason you need a cell phone to play the games. And play the games manually.Think outside the box. Stockfish is Stockfish on a phone or pc. What matters is the speed.M ANSARI wrote: ↑Tue Dec 29, 2020 8:03 am I would be really interested to know what the strength of engines on mobile platforms are today, especially with NN now being available. I would think that NN is the equivalent of massive hardware boost, which should help the limited resources offered by mobile platform. If you consider that massive boost with the fact that mobile processing has also become very powerful, I think it would make a huge improvement and maybe the engines on mobile are already at or above 3200 ELO. I guess someone with an older setup (say quadcore) could play a manual game and test that. If I remember correctly Rybka 3 was around 3200 ELO on quadcore at that time ... would be an interesting matchup ... Rybka 3 on Quadcore vs. latest SF NNUE on Iphone 12! I have a feeling that this would be easily won by SF NNUE even on mobile platform.
And to accurately determain the speed of a cell phone cpu. So you can make a equivalent mobile equivalent on your pc. You would most diffently would not want to use nodes per second.
You would need to dial in the speed using time to depth.
I will run such a test. And would be shocked if Stockfish is only 3200 Elo.
And we have a better version of Rybka with a very good ccrl rating called Fritz 17. Rybka 3 would be much too weak.
Here is the setup.
TC=3m+2s
Fritz 17 on 4 CPU's with TB's 6 man. CCRL Rating = Fritz 17 64-bit 4CPU 3381
PC equivalent version of Droidfish running Stockfish 280920.
Using a Samsung S20 Ultra 8 cores = 2 PC cores on a AMD 2950x.
I have no table bases on my phone. So Stockfish has no table bases in this test.
Results will be posted......
CCRL is keeping track, or any testing site that tests with 1 core.
1 core test results will be a very good analog for many many smart phones, and tablets running at full power.
My results
CCRL results Fritz 17 (4 cores) vs SF 12 (1 core)Code: Select all
DESKTOP-CORSAIR, Blitz 3.0min+2.0sec 0 1 Stockfish 280920 +331 +40/=14/-0 87.04% 47.0/54 2 Fritz 17 Popcnt -331 +0/=14/-40 12.96% 7.0/54
Now lets get shocked again. Let us see how my smartphone S20 Ultra running Droidfish will perform against Dragon at full power (32 threads).Code: Select all
Stockfish 12 64-bit 8CPU 3692 +15 −14 (+311) 7.5 − 83.5 (+0−76=15)
Chess Match Stockfish 280920 (2 cores, No TB) vs Dragon(32 Threads) (TC=3m+2s)
Code: Select all
DESKTOP-CORSAIR, Blitz 3.0min+2.0sec 0
1 Dragon by Komodo Chess 64-bit +10 +5/=64/-3 51.39% 37.0/72
2 Stockfish 280920 -10 +3/=64/-5 48.61% 35.0/72
"The worst thing that can happen to a forum is a running wild attacking moderator(HGM) who is not corrected by the community." - Ed Schröder
But my words like silent raindrops fell. And echoed in the wells of silence.
But my words like silent raindrops fell. And echoed in the wells of silence.
-
- Posts: 5589
- Joined: Wed Sep 05, 2018 2:16 am
- Location: Moving
- Full name: Jorge Picado
Re: Who is keeping track of the Cell Phone engines ratings ?
https://bgr.com/2020/10/21/iphone-12-pr ... 0-pixel-5/mwyoung wrote: ↑Wed Dec 30, 2020 4:14 amI am not sure about that. I am getting on 1 core with NNUE on a Samsung S20 Ultra almost 700 KN/s. And I know the Snapdragon CPU is not the fastest single core performer. Apple phones can be even faster on a single core.Milos wrote: ↑Wed Dec 30, 2020 2:53 amWhat are you talking about? This whole thread is a full phone (whichever number of threads it has) vs a single core CPU. And the high-end phones (with usually 4 slow and 4 fast threads) are on par if not better than a single core of a high-end CPU.MikeB wrote: ↑Tue Dec 29, 2020 3:41 amThat would surprise me — are you using one thread/core as indicated? Please double check your settings. Using one thread with NNUE enabled.Milos wrote: ↑Tue Dec 29, 2020 2:54 amI have a 180$ Chinese phone Xiaomi Redme Note 8 Pro with a mediocre CPU for current standards and I get 1565knps from a starting position with latest SF NNUE Droidfish. I'd expect most of high-end phones would actually be well over 2Mnps.MikeB wrote: ↑Mon Dec 28, 2020 7:18 pmMost certainly, most high end phones will see 800 to 1000k/nps with SF NNUE based engine, that would be around 30 to 40% of a recent single core CPU. Clearly less than 200 Elo difference.Chessqueen wrote: ↑Mon Dec 28, 2020 3:41 pm Is there any place that Cell Phone latest Engines rating is kept Updated, and how far are they from computer Engines ratings, Are they within 200 ratings points from Single chip or one core computer
There is absolutely no phone that can do 800knps on a single thread with SF-NNUE.
Who is 17 years old GM Gukesh 2nd at the Candidate in Toronto?
https://indianexpress.com/article/sport ... t-9281394/
https://indianexpress.com/article/sport ... t-9281394/
-
- Posts: 508
- Joined: Fri Jun 04, 2010 7:23 am
Re: Who is keeping track of the Cell Phone engines ratings ?
So is CFish 191020 the strongest engine for Droid fish?
If not then what NNUE engine is?
If not then what NNUE engine is?
-
- Posts: 4190
- Joined: Wed Nov 25, 2009 1:47 am
Re: Who is keeping track of the Cell Phone engines ratings ?
Geekbench .Chessqueen wrote: ↑Wed Dec 30, 2020 8:45 amhttps://bgr.com/2020/10/21/iphone-12-pr ... 0-pixel-5/mwyoung wrote: ↑Wed Dec 30, 2020 4:14 amI am not sure about that. I am getting on 1 core with NNUE on a Samsung S20 Ultra almost 700 KN/s. And I know the Snapdragon CPU is not the fastest single core performer. Apple phones can be even faster on a single core.Milos wrote: ↑Wed Dec 30, 2020 2:53 amWhat are you talking about? This whole thread is a full phone (whichever number of threads it has) vs a single core CPU. And the high-end phones (with usually 4 slow and 4 fast threads) are on par if not better than a single core of a high-end CPU.MikeB wrote: ↑Tue Dec 29, 2020 3:41 amThat would surprise me — are you using one thread/core as indicated? Please double check your settings. Using one thread with NNUE enabled.Milos wrote: ↑Tue Dec 29, 2020 2:54 amI have a 180$ Chinese phone Xiaomi Redme Note 8 Pro with a mediocre CPU for current standards and I get 1565knps from a starting position with latest SF NNUE Droidfish. I'd expect most of high-end phones would actually be well over 2Mnps.MikeB wrote: ↑Mon Dec 28, 2020 7:18 pmMost certainly, most high end phones will see 800 to 1000k/nps with SF NNUE based engine, that would be around 30 to 40% of a recent single core CPU. Clearly less than 200 Elo difference.Chessqueen wrote: ↑Mon Dec 28, 2020 3:41 pm Is there any place that Cell Phone latest Engines rating is kept Updated, and how far are they from computer Engines ratings, Are they within 200 ratings points from Single chip or one core computer
There is absolutely no phone that can do 800knps on a single thread with SF-NNUE.
As relevant to chess as is astrology to your life.