Author: Robert Hyatt
Date: 15:12:05 02/13/03
Go up one level in this thread
On February 13, 2003 at 15:09:28, Tom Kerrigan wrote: >On February 12, 2003 at 21:07:57, Robert Hyatt wrote: > >>>>>>The Cray T932 was the last 64 bit machine they built that I used. And it >>>>>How many NPS does Crafty get on it? >>>>about 7M. >>>>And that was Cray Blitz, not Crafty. I have not tried to run Crafty on a Cray, >>> >>>7M per processor? How many processors did those Crays come with? 32? So Crazy >>>Blitz was searching 224M NPS? I stand corrected, the T932 is several times >>>faster than any other processor ever made. >>No. 7M total as I said, spread over 32 processors (actually we only used 31 >>on that machine so that one could be dedicated to handling interrupts and the >>like to avoid unnecessary context switches). > >I guess I can count on you to cite MP numbers when discussing UP performance. I gave both. You can surely take the number I gave, which is the only number I have (7M nps) and divide by 31, to convert to the NPS on a given cpu? However, the topic at the moment was a 486 vs the cray and then a bunch of 486's vs the cray. Neither have a chance on any reasonable application. > >It's also apparently too much to ask to run the _same program_ on two platforms >to compare their performance. Yes it is, when one of the platforms sells for $60,000,000 and is _very_ difficult to get time on. I suspect that Cray Blitz and Crafty could be "close" in speed, but "suspect" is not "proving". > >That said, 7M/31 = 225k NPS. So at 500MHz, it runs about as fast as a 300MHz >Athlon (Crafty NPS)... Hmmm... Well, probably a slower Athlon, given the >relative memory latencies... Yeah, somehow you're not making the case that Cray >processors are super awesome and their compilers eliminate the need for OOOE... If you had ever followed discussions on Cray Blitz, you would know that we did a lot of things in it that I can't do in crafty due to the cost. For example, "qualitative mobility". Not just the number of squares a piece can move to, but the number of squares it can move to safely _and_ how useful the square is (ie e4 is much better than a8). Or computing all the pieces attacking squares around the king, even pieces through pieces, and so forth. All nice for vector stuff, all not so nice for a PC. What about multiple hash table probes? Not cheap on the PC, free on the cray, for example. I don't remember the specifics any longer, but I believe a 386/25mhz processor ran Cray Blitz (pure fortran version) at well under 100 nodes per second, for comparison. Run that up to 250mhz and it would not be at 1000. 500mhz would get near 2000 with some luck, so the cray did a lot to sizzle right past that number by yet another factor of 100... or more. > >-Tom
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.