what is the oldest hardware when engines at GM level?

[lkaufman]

Unfortunately the opening book used has more to do with which engine avoids the occasional draw with superGM more often (as Black) than the engine itself.

Can someone create a book that beats the strongest human 100 games out of 100?

The only somewhat meaningful test of which engine is better at beating humans would be chess960

Still not chess, not any better than knight odds! It's like we can't answer the question and we create unrelated scenarios that answer other questions. The only way to answer chess questions is by playing chess (in the same conditions that were used to rank the human player.)

I ran some quick tests which indicate that on one thread on a fairly fast machine, the break-even point between latest Komodo and Stockfish 11 is about 15 ms movetime (after making a change to make Komodo observe the time limit more accurately, to be fair). At this time limit Komodo averaged 11.4 ply, Stockfish 11.6 ply. This is around the level that I think would give strong grandmasters (perhaps not the very top ones) a fair battle at 15' + 10" Rapid.

Again, the strength of engines are being used to make guesses about the strength of humans. I don't think any of this has been really tested, like, get a human to play a bunch of 3000 elo engines and a bunch of 3100 engines and see if any difference in strength can be measured. Perhaps the 3000 elo engines are 3000 elo, but the 3400 elo engines are really 3040 against humans (perhaps the 3000 elo engines perform better than the 3100 ones!), and no amount of engine v engine or human with odds/960 stuff would get us closer to the truth.

Since Komodo already went 10 out of 10 vs. Nakamura a couple weeks ago at ten minute chess, and anything slower or more games would probably not be practical without a sponsor, testing on fully even terms (including time) vs. humans is not plausible. But testing to see which engine can make a plus score against GMs in Rapid or blitz games with less time per move on one thread is a valid competition that could be done without an exorbitant cost. I think Komodo would be competitive with Stockfish in this.

[Chessqueen]

Unfortunately the opening book used has more to do with which engine avoids the occasional draw with superGM more often (as Black) than the engine itself.

Can someone create a book that beats the strongest human 100 games out of 100?

The only somewhat meaningful test of which engine is better at beating humans would be chess960

Still not chess, not any better than knight odds! It's like we can't answer the question and we create unrelated scenarios that answer other questions. The only way to answer chess questions is by playing chess (in the same conditions that were used to rank the human player.)

I ran some quick tests which indicate that on one thread on a fairly fast machine, the break-even point between latest Komodo and Stockfish 11 is about 15 ms movetime (after making a change to make Komodo observe the time limit more accurately, to be fair). At this time limit Komodo averaged 11.4 ply, Stockfish 11.6 ply. This is around the level that I think would give strong grandmasters (perhaps not the very top ones) a fair battle at 15' + 10" Rapid.

Again, the strength of engines are being used to make guesses about the strength of humans. I don't think any of this has been really tested, like, get a human to play a bunch of 3000 elo engines and a bunch of 3100 engines and see if any difference in strength can be measured. Perhaps the 3000 elo engines are 3000 elo, but the 3400 elo engines are really 3040 against humans (perhaps the 3000 elo engines perform better than the 3100 ones!), and no amount of engine v engine or human with odds/960 stuff would get us closer to the truth.

Since Komodo already went 10 out of 10 vs. Nakamura a couple weeks ago at ten minute chess, and anything slower or more games would probably not be practical without a sponsor, testing on fully even terms (including time) vs. humans is not plausible. But testing to see which engine can make a plus score against GMs in Rapid or blitz games with less time per move on one thread is a valid competition that could be done without an exorbitant cost. I think Komodo would be competitive with Stockfish in this.

The question that I have is how strong is your latest Komodo on an old 386 or 486DX2 intel processo from 2650 to 2750 compared to a FIDE human player ? I do have an old 486DX2 on my baseman collecting dust, that I have not used in more than 18 years

[mig2004]

Undust your old 486dx, get all the connections cleaned up and a good keyboard and monitor. If they work, then you will be surprised how much more you can get from old hardware.

Im still running my intel dual core e8400... up and running since 2006. I have used windows xp, win 7, lubuntu, suse, windows 8, 8.1, and soon will be moving back to the linux system, probably lubuntu gain, althought im not decided on that.

It has been a great decade for this old workhorse. it's running 24x7 with just 2 coolling fans: the cpu and one external fan. (the power supply fan went down about 3 yeas ago).

For your kind of hardware you can easily install lubuntu. It is easier than windows as of today. And there are other linux distros just as easy now days.

https://lubuntu.net/downloads/

[lkaufman]

Unfortunately the opening book used has more to do with which engine avoids the occasional draw with superGM more often (as Black) than the engine itself.

Can someone create a book that beats the strongest human 100 games out of 100?

The only somewhat meaningful test of which engine is better at beating humans would be chess960

Still not chess, not any better than knight odds! It's like we can't answer the question and we create unrelated scenarios that answer other questions. The only way to answer chess questions is by playing chess (in the same conditions that were used to rank the human player.)

I ran some quick tests which indicate that on one thread on a fairly fast machine, the break-even point between latest Komodo and Stockfish 11 is about 15 ms movetime (after making a change to make Komodo observe the time limit more accurately, to be fair). At this time limit Komodo averaged 11.4 ply, Stockfish 11.6 ply. This is around the level that I think would give strong grandmasters (perhaps not the very top ones) a fair battle at 15' + 10" Rapid.

Again, the strength of engines are being used to make guesses about the strength of humans. I don't think any of this has been really tested, like, get a human to play a bunch of 3000 elo engines and a bunch of 3100 engines and see if any difference in strength can be measured. Perhaps the 3000 elo engines are 3000 elo, but the 3400 elo engines are really 3040 against humans (perhaps the 3000 elo engines perform better than the 3100 ones!), and no amount of engine v engine or human with odds/960 stuff would get us closer to the truth.

Since Komodo already went 10 out of 10 vs. Nakamura a couple weeks ago at ten minute chess, and anything slower or more games would probably not be practical without a sponsor, testing on fully even terms (including time) vs. humans is not plausible. But testing to see which engine can make a plus score against GMs in Rapid or blitz games with less time per move on one thread is a valid competition that could be done without an exorbitant cost. I think Komodo would be competitive with Stockfish in this.

The question that I have is how strong is your latest Komodo on an old 386 or 486DX2 intel processo from 2650 to 2750 compared to a FIDE human player ? I do have an old 486DX2 on my baseman collecting dust, that I have not used in more than 18 years

I have no idea whether Komodo will actually run on that hardware, but if it does and you can see how long it takes to reach various depths in typical middlegames, we have a basis for making an estimate. I am reasonably confident that Komodo searching 12 ply on one thread with a good opening book would be competitive with Carlsen or Nakamura playing 15' + 10". A time control of 3" + .03" inc on a modern i7 is about equivalent to this. So that's 300 to 1 time odds. So if your old hardware running 300 times faster would be as fast as one thread of a modern i7, then it should be able to play on even terms in rapid with the best players in the world. I'll leave it to hardware experts to say whether your particular hardware is at that speed.

[Milos]

Too bad you didn't say say 'slowest' instead of 'oldest' - it would be easier to answer. You mentioned stockfish, based on all the testing I have done with stockfish as fewer nps, my estimate is that SF needs somewhere between 80K/nps to 150K/nps to be legitimate GM strength and that is probably only attainable with using at least 5 man EGTB as well.

I'm a sorry but your estimation is a complete joke.
SF on 100knps is only 100 times slower than SF on 4 cores top of the line CPU that is equivalent of at least 3400Elo. 100 times slower is less than 7 doublings and assuming even 80 Elo per doubling (which is probably a big exaggeration), SF on 100knps would still be around 2850 Elo, i.e. stronger than Carlsen on his best.
GM level starts from 2450. That is additional 5 doublings or ~30x speed factor. So it is quite safe to assume that SF on 3knps would be on GM level, and most probably even on 1knps.

[Gary Internet]

This discussion on how strong is the latest software when it is made to run on the oldest/weakest/slowest/worst hardware that it's capable of running on is far more interesting than how the latest software performs when it's running on hardware that costs at least as much as a decent second hand car.

I know there have been threads about Komodo giving knight odds etc to a GM when running on 30+ cores. I'm not interested in that. I'd much rather see how good it is on 1 core, on something like my Dell Windows 10 dual core laptop, which I purchased in May 2011. All the software on it is new, but as far as hardware is concerned it's nearly a decade old and was probably classed as mediocre-at-best when I got it.

I have a 5 or 6 year old Huawei Y635 mobile phone. It's Android. I have whatever the latest version of Droidfish is. The other day, watching about 25 moves of a TCEC game, I was entering the moves into my phone and Droidfish running at no more than 153 kn/s (often 140 or 145 kn/s) was picking the same moves as Ethereal and Komodo on 176 threads within about 30 seconds. I'm not saying my phone is as strong as the TCEC hardware, but the fact we live in a world where out of date and very cheap hardware, coupled with free cutting edge software can pick the same moves as an incredibly powerful setup like the one at TCEC is pretty cool.

[Ovyron]

I know there have been threads about Komodo giving knight odds etc to a GM when running on 30+ cores. I'm not interested in that. I'd much rather see how good it is on 1 core, on something like my Dell Windows 10 dual core laptop, which I purchased in May 2011. All the software on it is new, but as far as hardware is concerned it's nearly a decade old and was probably classed as mediocre-at-best when I got it.

+1!

This idea of getting the fastest possible hardware and then taking away material to compensate makes no sense! You want it to do better so you use faster hardware but you want the human to do better so you take material away from the engine?? Make up your mind!

So I hope slower hardware gets tested with full material. Apparently the biggest obstacle is how much humans charge, but Nakamura has been playing weak humans (and teaching them) for free.

We're at a point where all this is being sponsored by Nakamura's audience on Twitch, so he'd do anything that his fans want him to do (like getting to 3000 elo from a new chess.com account by playing the Bong Cloud and similar unsound openings...)

[Vinvin]

This discussion on how strong is the latest software when it is made to run on the oldest/weakest/slowest/worst hardware that it's capable of running on is far more interesting than how the latest software performs when it's running on hardware that costs at least as much as a decent second hand car.

I agree with you.
It would be a great fun to watch SF running at 10 Kn/s (or 1/100 time handicap on a phone) versus a 2700 human.

But I also would like to see star engines more than 20 year old on modern machine (full speed with VirtualBox) vs current IMs and GMs.
All that engines that made the Computer chess history around 1999 : Chess Tiger 12.0, Fritz 5.32, Hiarcs 7.32, Junior 5.0, Crafty 18, Chessmaster 6000, MChess Pro 8.0, Rebel 8.0, Virtual Chess and even Chess Genius 3.

May be one day ...

[Ovyron]

Maybe some people could stop investing in hardware and start investing into paying the humans to play under those conditions, hehe.

[Gary Internet]

https://www.chess.com/news/view/stockfi ... an-foreest

There's another example of what I'm talking about. Stockfish running on 72 threads. Even if that was "only" Stockfish 11, instead of the latest development version, it's a total joke. How many nodes per second will it be examining? 70 to 80 million per second in the opening? 90+ million in the middle game and 130 million in the endgame? This is a similar level of hardware that we see Stockfish using to play against other engines at CCC. Other engines which are also far stronger than any human past, present or future. Now it's using that hardware against a human. Sure, two pawn down is a big deal, but SF is playing as white and it will be ruthless. Time control is 30+10 as well. A 10 second increment is huge for Stockfish. I bet it has 6 man tablebases as well. What a joke.

How about running Stockfish 8 on a knackered old laptop with no tablebases and just letting them play 20 standard games at 90+10 over the course of a month, with the human playing as white the whole time. Much more interesting, but it's less glamorous. I get it. It's all about publicity and money.

Instead of seeing if a man can outrun a dog over 100 metres, we're now seeing if a man can outrun a Formula One racing car over 400 metres when the man has a 300 metre head start.