harware vs software advances

[Don]

I move this to a new thread because the subject matter of this discussion does not belong in the move count pruning thread.

I have been basically running some chess genius 3 games against Robbolitto on the same hardware in order to ascertain how much software has progressed. I started out with some 24 to 1 time odds but I'm now up to 1000 to 1 odds, as Robbolitto has not yet lost a game. I'm running a time control of 100 moves in 30 seconds for robbolitto and 100 moves in 8 hours 10 minutes for genius.

I am assuming that Genius 3 represents the state of the art in 1994, the year it came out. This may not be the case, but I think it's probably close - but at any rate I don't have any other programs that run on my computer from 1994 or earlier.

I read on-line that in March of 1994 200 MHZ Pentiums were available. Bob Hyatt estimates that the i7 on one core is 1000 times faster. That number seems high to me but I don't have a way to check it out with timing of chess programs. At any rate I think it's probably within a factor of 2 to 1. Clock for clock this is about 15 to 1 difference, but computers do much more in one clock cycles than they used to.

If you assume a doubling in speed every 2 years you would get about a 200 times increase in CPU power. I upgrade my computers every 2 years and rarely have gotten a 2 to 1 increase and I'm always a little bit disappointed.

But if anyone has one of those 200 MHZ machines, we can probably test the speed with a DOS chess program or anything that runs on it and still runs on todays hardware. My best guess is that we achieve 500X speedup since then.

The last 2 games were played with 500 to 1 time handicap and Robbolitto won both of those games.

Now I'm playing a 1000 to 1 time handicap (while I'm doing other things) and both programs believe they have an advantage, a typical situation with Chess programs. In reality I believe Genius has the advantage (as well as the white pieces) but so far it looks like it will not be so one-sided. One cannot tell from just a few games however.

There is no possible way to get a significant sampling of games. But if Robbo keeps winning I am satisfied to say that software advances is at least on the same general scale as hardware advances (since 1994). For reasons we have discussed on the other thread it's impossible to make this a perfectly "fair" test because it's difficult to define fair.

[Don]

I'm looking on another web site and it appears that in March of 1994 Inel released the 90 MHZ pentium. Another web site says they released 200 MHZ processors in the same month, but I see nothing else that verifies this.

However I'm inclined to believe it was the P90 that was state of the art. In Hong Kong 1995 I think Fritz was running on a P90 (I was there) and it won the world championship. But what stands out to me was that it running on the slowest CPU at the tournament - everyone commented on that and the fact that even the other micros were running on faster hardware. Wchess might have been also running on a P90, I don't remember for sure.

I move this to a new thread because the subject matter of this discussion does not belong in the move count pruning thread.

I have been basically running some chess genius 3 games against Robbolitto on the same hardware in order to ascertain how much software has progressed. I started out with some 24 to 1 time odds but I'm now up to 1000 to 1 odds, as Robbolitto has not yet lost a game. I'm running a time control of 100 moves in 30 seconds for robbolitto and 100 moves in 8 hours 10 minutes for genius.

I am assuming that Genius 3 represents the state of the art in 1994, the year it came out. This may not be the case, but I think it's probably close - but at any rate I don't have any other programs that run on my computer from 1994 or earlier.

I read on-line that in March of 1994 200 MHZ Pentiums were available. Bob Hyatt estimates that the i7 on one core is 1000 times faster. That number seems high to me but I don't have a way to check it out with timing of chess programs. At any rate I think it's probably within a factor of 2 to 1. Clock for clock this is about 15 to 1 difference, but computers do much more in one clock cycles than they used to.

If you assume a doubling in speed every 2 years you would get about a 200 times increase in CPU power. I upgrade my computers every 2 years and rarely have gotten a 2 to 1 increase and I'm always a little bit disappointed.

But if anyone has one of those 200 MHZ machines, we can probably test the speed with a DOS chess program or anything that runs on it and still runs on todays hardware. My best guess is that we achieve 500X speedup since then.

The last 2 games were played with 500 to 1 time handicap and Robbolitto won both of those games.

Now I'm playing a 1000 to 1 time handicap (while I'm doing other things) and both programs believe they have an advantage, a typical situation with Chess programs. In reality I believe Genius has the advantage (as well as the white pieces) but so far it looks like it will not be so one-sided. One cannot tell from just a few games however.

There is no possible way to get a significant sampling of games. But if Robbo keeps winning I am satisfied to say that software advances is at least on the same general scale as hardware advances (since 1994). For reasons we have discussed on the other thread it's impossible to make this a perfectly "fair" test because it's difficult to define fair.

[uaf]

I'm pretty sure the Pentium 90 was the fastest processor at that time. My uncles bought two of them for their company in 1994 and they were the fastest processors available. Intel introduced the Pentium Pro family (to which the P200 belongs) in late 1995.

Interesting tests, let us know how it end with the 1:1000 time handicap

[rbarreira]

According to cpu-world the Pentium 200 only came out in June of 1996:

http://www.cpu-world.com/CPUs/Pentium/I ... 02200.html

[Don]

I'm pretty sure the Pentium 90 was the fastest processor at that time. Intel introduced the Pentium Pro family (to which the P200 belongs) in late 1995.

Yes, I'll assume that circa 1994 CG3 was running on the P90.

Any idea of the speed? I don't suppose anybody has a Pentium 90 laying around do they?

Don

[rbarreira]

I'm pretty sure the Pentium 90 was the fastest processor at that time. Intel introduced the Pentium Pro family (to which the P200 belongs) in late 1995.

Yes, I'll assume that circa 1994 CG3 was running on the P90.

Any idea of the speed? I don't suppose anybody has a Pentium 90 laying around do they?

Don

This could be fun... I got REBEL 10 from this page and I'm now trying to get it to run in VMWare in order to compare with these benchmarks.

[bob]

I'm pretty sure the Pentium 90 was the fastest processor at that time. My uncles bought two of them for their company in 1994 and they were the fastest processors available. Intel introduced the Pentium Pro family (to which the P200 belongs) in late 1995.

Interesting tests, let us know how it end with the 1:1000 time handicap

Your last comment is the wrong architecture. Intel released both a pentium and pentium pro in 200mhz versions. The pentium-200 was an in-order dual-pipe box just like all original pentiums. The pentium-pro 200 and beyond introduced out of order execution, register renaming, reorder buffer, all the other clever stuff to greatly increase IPC. In fact, I think they called the 200mhz pentium "pentium-mmx 200mhz" as they had started adding the mmx stuff prior to the P6 core with its OOE stuff.

[bob]

I'm pretty sure the Pentium 90 was the fastest processor at that time. Intel introduced the Pentium Pro family (to which the P200 belongs) in late 1995.

Yes, I'll assume that circa 1994 CG3 was running on the P90.

Any idea of the speed? I don't suppose anybody has a Pentium 90 laying around do they?

Don

Best number I have for Crafty is 30K nps on a P5/133mhz box. Went to 75K on the first P6/200 I had. One year later about 300K on the quad P6/200 box I bought.

Bruce Moreland has a P5/66 and then a p5/90. I remember running Crafty on his 90mhz, but I only vaguely recall 18-20K as the speed. I'm not very confident since I didn't have the box on my desk to watch. But that seems to sort of track the mhz reasonably so it has a chance of being right, for Crafty of course. Back then, Fritz was quite a bit faster than Crafty, however... So that number has nothing to do with fritz back then nor today.

[bob]

According to cpu-world the Pentium 200 only came out in June of 1996:

http://www.cpu-world.com/CPUs/Pentium/I ... 02200.html

That was just a normal pentium, using the new manufacturing facility they had built. The P5/200 was way cheaper to build than the P6 architecture, and they were still addressing the el-cheapo end of the market as well as the server market where the P6/200 was king.

[uaf]

I'm pretty sure the Pentium 90 was the fastest processor at that time. My uncles bought two of them for their company in 1994 and they were the fastest processors available. Intel introduced the Pentium Pro family (to which the P200 belongs) in late 1995.

Interesting tests, let us know how it end with the 1:1000 time handicap

Your last comment is the wrong architecture. Intel released both a pentium and pentium pro in 200mhz versions. The pentium-200 was an in-order dual-pipe box just like all original pentiums. The pentium-pro 200 and beyond introduced out of order execution, register renaming, reorder buffer, all the other clever stuff to greatly increase IPC. In fact, I think they called the 200mhz pentium "pentium-mmx 200mhz" as they had started adding the mmx stuff prior to the P6 core with its OOE stuff.

You are perfectly right. I forgot the 200 MMX.