Author: Bruce Moreland
Date: 09:18:10 11/07/97
There has been interest in the 767 mhz overclocked Kryotech Alpha I used
in Paris. Between rounds I did some benchmarks, and I would like to
present these.
The 767 mhz Alpha is approximately 2.46 times faster than a 200 mhz
Pentium Pro, when running my chess program. I ran some suites and
checked three of them, and this was the factor achieved in all three.
The 533 mhz Alpha is about 1.6 times faster than a 200 mhz Pentium Pro.
This factor wasn't as consistent. At other times I have measured this
as 1.75, that may be a better number.
Here are some suite scores, if you like suite scores. The reason BT2630
was worse on the 767 is that the program changed its mind away from the
solution move in positions 27 and 28.
All of this should be taken to be somewhat random. I don't tune for
these suites, and the scores tend to skip around.
At the risk of saying something obvious, an implication is that if my
program went 100K nps on a P6/200, it would go 246K nps on the big
Alpha. In practice it typically went 250-500K nps, so apparently it
would go a little faster than 100K on the P6/200.
I didn't test on a PII/300 in Paris, this is unfortunate. I was
concerned that someone would break into the machine and steal software,
as has apparently happened numerous times at these events.
bruce
--------------
Encyclopedia of Chess Middlegames (879 positions)
200 533 767
--- --- ---
5sec 489 538 547
10sec 547 577 591
15sec 573 606 608
20sec 593 618 627
LCT I
Pos 84 88 88
Tac 116 116 116
End 112 116 120
---- ---- ----
"Elo" 2563 2578 2584
LCT II
Pos 55% 58% 52%
Tac 74% 79% 81%
End 67% 70% 76%
---- ---- ----
2575 2620 2615
BT2630
2470 2455 2436
BS2830
2628 2678 2702
Win at Chess (300 positions)
5sec 282 283 288
10sec 287 291 290
30sec 293 294 292
60sec 295 296 294
Covax
2 5 5
Reiter
13 13 12
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.