Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Crafty and Alpha processors

Author: Robert Hyatt

Date: 17:34:09 01/20/01

Go up one level in this thread


On January 20, 2001 at 18:04:54, Joshua Lee wrote:

>On January 20, 2001 at 16:56:17, Joshua Lee wrote:
>
>>On January 20, 2001 at 16:12:25, Eugene Nalimov wrote:
>>
>>>On January 20, 2001 at 16:01:46, Robert Hyatt wrote:
>>>
>>>>On January 20, 2001 at 15:39:02, Joshua Lee wrote:
>>>>
>>>>>How fast is crafty on a 800Mhz PIII? AMD Athlon at 800Mhz? I don't know about
>>>>>the Alpha too much which CPU speeds do they offer for comparison? Would a Dual
>>>>>or Quad Xenon or Athlon compare with a Single Alpha ? Thanks
>>>>
>>>>Here is the output for running WAC at 60 seconds per move on a single xeon cpu,
>>>>700mhz, 1M L2 cache:
>>>>test results summary:
>>>>
>>>>total positions searched..........         300
>>>>number right......................         299
>>>>number wrong......................           1
>>>>percentage right..................          99
>>>>percentage wrong..................           0
>>>>total nodes searched.............. 132559227.0
>>>>average search depth..............         4.7
>>>>nodes per second..................      412161
>>>>
>>>>
>>>>
>>>>Here is the same test run on a 21264 alpha at 600mhz:
>>>>
>>>>test results summary:
>>>>
>>>>total positions searched..........         300
>>>>number right......................         300
>>>>number wrong......................           0
>>>>percentage right..................         100
>>>>percentage wrong..................           0
>>>>total nodes searched.............. 236973211.0
>>>>average search depth..............         4.5
>>>>nodes per second..................      783641
>>>>
>>>>As you can see, the alpha has a 16% slower clock speed, but is nearly 2x
>>>>faster...
>>>>
>>>>Bob
>>>
>>>I believe that part of the explanation is the compiler. You are using obsolete
>>>GNU compiler on the x86, but state-of-the-art compiler on Alpha. Good x86 C
>>>compiler will give you at least additional 20%.
>>>
>>>Eugene
>>
>>I have decided to run the same test on my 800Mhz Athlon which should get
>>destroyed compared with your results, and again on my PIII 500 the drawback will
>>be lack of ram on the laptop 32MB tops for hashtables should make for a huge
>>disadvantage compared with an extra 300Mhz and an extra 104MB for Hash Tables
>>(136) I am not sure if this is better with just 128MB but i can check.
>>
>>I am seeing between 340 and 480Knps so i hope under chessbase's process test set
>>it will let me know an exact average. So far 31 wasn't solved so....we'll see.
>
>
>
>Ok when my computer was finished it ended up saying 4 weren't solved but i went
>back through and see where Crafty 17.14 doesn't solve 8 of them so it's best to
>get 1 specific position in which to test NPS and compare the results from that
>position so any thoughts as to what position to use ? It may be a big difference
>also if we are using different verison which is what it sounds like to me.


For your test this doesn't matter...  just run the _same_ version on the two
different machines.  Best would be to simply type "bench" and let that run.
default hash and everything.



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.