Last Update: 15.08.2022


Dear Chess Friends,

As usually, I am pleased to announce that,Another new Multi-CPUs Testing is organized!Btw, for previous Multi -CPU results: Ranking

1st of all, and to be honest:To run this testing is born after reading Samurai's postingWhere he criticized that via 1 CPU,1m+1s etc testings areUseless ridiculous...plus he said that even simple book willPerform same, if not better... hehe..)) if so...Why between1st and last rank, there is so much difference (in points )?It seems this guy has no clue what's going on, and if same...I invite Samurai to submit...to: BNV let's see Shogun's power?)But I highly doubt this ever will happen, because these kind ofBrains are as Winners only via Papers...no more...no less....
Anyhow, I realized to run this new testing, to check again...nowWe have more data about the reality:e.g 1 CPU is so weak... ?Meanwhile, even in Analyzing,-mode, when I tested as MP+NNUE,The influence of many engines were not so BIG, for more DetailsSure, 1 CPU is weaker ..but not so much....e.g via 1 CPU, 1m+1sI am able to test a lot of books and to produce so many games...And all in very short of time...this is another very important factor!Actually there are such weak books, even 100 games is too muchAnd I'm not talking about as MP, 60min+15sec..etc. I mean aboutTestings played even 1CPU,1m+1s ..really waste of time, but I test..And I'm afraid that Samurai will never understand...what I mean...
However, here are the latest played results and GAMES

1st Test: 16 CPUs vs 1 CPU, Elo difference: 28


 1 SF 120822 16 CPU +16/-0/=184 54.00% 108.0/2002 SF 120822 1 CPU +0/-16/=184 46.00% 92.0/200
Note: The above testing is played under Blitz 2min+1sec, 16 CPUs = 1024 MB Hash / 1 CPU = 128 MB Hash
------------------------------------------------2nd Test : 64 CPUs vs 1 CPU, Elo difference: 70
 1 SF 120822 64 CPU +20/-0/=80 60.00% 60.0/1002 SF 120822 1 CPU +0/-20/=80 40.00% 40.0/100
Note: The above test is played under Bullet 1min+1sec 64 CPUs = 1024 MB Hash / 1 CPU = 64 MB Hash
2nd, for a better conclusion, we need more games,However I think that the current data can be useful forComparing.,normally I could run more games, but myTournament PC will be busy with other tour: ResultsLater I may run more...but I don't want to promise...
Third, more details about NNUE testings, in shortI was wondering again what will be Elo influences..As we see, in case of testing the latest SF as NNUE:There is no much Elo difference again..on other hand,Any experienced in chess, probably is aware of thatMore cores = more points, and this is also true that:Stronger chess = more draws, I mean in case of testsSuch as 16 CPUs vs 16 CPUs, 64 CPUs vs 64 CPUs...Actually if giving more time...even 1 CPU vs 1 CPU,We may see more draws..e.g close to 99% for sure...In short, it all depends on our used test conditions..
In other words (since NNUE appeared on scene..),As I stated earlier: the strength is increased a lot...Less blunder moves, better quality chess for sure!But in same time, another BIG problem is appeared:Sure it all depends, but under optimal test conditions,The overall draws are increased a lot...e.g mainly viaHandicapped, critical, weak openings or with ultra fastTime controls, 1CPU, Ponder OFF etc. then we'll haveChances to reduce the overall draws, for more Details
Just to be more clear, (In case of serious testings)What I'm trying to tell (if NNUE 3700+ vs 3700+),There is no much alternatives to reduce the draws..And I think that this is the real major BIG problem...Otherwise, if almost all game results will be draw, thenWhat is the point of running these kind of testings?)
Sure, there are some other ways to reduce the DRAWS,And in the same time, to increase the FUN of course e.g:The Top engines to be played as NON-NNUE, especiallyCfish 030820 is Great..not so Drawish as many StockfishBased ones, this is already proved in many competitions!
Of course, we should not forget to mention also that,Raubfish SC is another Great engine (even as NNUE)Which is not so Drawish too, proved also in many tours!Yes..not the strongest (just 18 Elo weaker...) but performsApprox. 85%-90% draws, I mean close to 5%-10% less..!That means a lot... .if nothing else, at least for my side!Not sure about Super League, but probably the drawsWould be approx. 90%- 95%...after all, time will tell ...!
That's all for now and thanks for your interest...)

Greetings,Sedat