Author: Robert Hyatt
Date: 17:25:37 12/09/03
Go up one level in this thread
On December 09, 2003 at 15:05:29, Peter Kasinski wrote: >On December 08, 2003 at 15:31:51, Robert Hyatt wrote: > >>On December 08, 2003 at 14:32:15, Aaron Gordon wrote: >> >>>On December 08, 2003 at 13:08:41, Slater Wold wrote: >>> >>>>On December 08, 2003 at 12:55:36, Leen Ammeraal wrote: >>>> >>>>>I saw the Athlon 64 based >>>>>"PC Vobis Power 64 3200+ XD" (euro 1299). >>>>> >>>>>How does this compare with an Intel P4 3.2 EE ? >>>>>Which would you prefer for chess? >>>>>Leen >>>> >>>>According to http://www.hothardware.com/hh_files/CCAM/a64fx_51_launch.shtml, the >>>>3.2EE would be faster. >>> >>>Shame they only tested (except for UT2k3) applications where the P4 does decent. >>>As I said in a previous post.. where is DVD2AVI (where an Athlon does best for >>>Mpeg encoding). It is much faster than XMpeg for me (on my XP). Also, in Povray >>>(a renderer a normal person would use, rather than drop $4000 for one of the >>>ones they use to test) the Athlon is a good 50%+ faster than the P4. In 64bit >>>mode it is something like 3 times faster. The Athlon FX is even faster than the >>>XP. Expecially for chess, which I believe what he was asking about in the first >>>place. >>> >>>Go test Fritz, Shredder, Crafty, etc. on a 2.2GHz FX. You'll see in 32bit mode >>>you'll gain 20-30% over an Athlon XP (which is already faster than a P4 for >>>chess). >>> >>>About PCMark and 3DMark (all made by madonion/futuremark).. they're all bunk. >>>PCMark is biased towards the P4 and now that ATI slapped them with some cash >>>they're biased towards ATI. If you don't believe me, try this. Get a Pentium 2 >>>400MHz (yes, 400MHz) and a Radeon 9700 pro, 9800, whatever. Now, put a Geforce4 >>>ti4600 in a P4-3.2EE, Athlon FX 2.8GHz, you name it.. doesn't matter. The >>>Pentium 2 400MHz will get a higher 3DMark. Why? Well, thats what happens when >>>you dump a ton of cash on a company.. they do what you want. >>> >>>The semi-technical reason why it is like this was something nvidia found out. >>>They found that the way futuremark did the pixel shaders was ridiculous. Adding >>>a specific loop (or something like that, you can search for it on google) that >>>only the ATI chips could do.. and the entire test setup was this crap. In real >>>life, and in any other 3D program the Geforce4 on the faster CPU would >>>absolutely kill the P2-400MHz with the 9700 Pro. This is an example of the crap >>>companies do to fool customers. >>> >>>I've done the testing (P4-2.53 @ 3.32ghz and limited testing at 3.5ghz) for >>>chess and other programs, my Athlon XP at 2.5GHz beat it in 95% of the tests and >>>ALL of the chess programs. An Athlon FX 2.2GHz is 20-30% faster than an XP at >>>2.2GHz, so you can figure it'd be equal to an XP 2.64-2.86GHz.. which is >>>definitely faster than any P4 (even if they clocked it up to 4GHz and more) to >>>date. If you want the fastest, go for this: >>>http://www.aceshardware.com/read.jsp?id=60000268 >>> >>>Plus, if you get a P4-3.2EE what do you think you'll be doing when lots of the >>>chess programs go to 64bit, and most of everything else? You'll end up buying an >>>Athlon FX.. the P4 will get smoked even more when 64bit stuff hits the scene. >>>You can get one now and when everything switches over you won't have to upgrade >>>at all. If you do go with a P4-EE you'll just be one of many who wonder why it >>>isn't as fast as the 'review' pages say. >>> >>>If in doubt, test it yourself.. I did. >> >> >>I agree about the amd64 speeds. IE a 1.8ghz opteron is faster than my >>2.8ghz xeon by a significant margin. One cpu test puts the 1.8ghz opteron >>at a bench speed of 1,639.241 nps. My 2.8ghz xeon weighs in at 1,079,714 >>nps. Add .4ghz to reach 3.2, and add .4 to reach amd's 2.2, and the >>amd is going to be 50-60% _faster_... > >Bob, could you please clarify if there is any special recompiling involved, or >do these numbers represent Crafty's performance in the version as is - from your >site. > >Thanks, >PK Version 19.5/19.6, exactly from my web site. 19.7 will have a linux-amd64 makefile target so that you can use _exactly_ what I use to compile as well, if you have an opteron with the Suse 9 release. BTW, I finally got a PGO compile to work. It was slower. Not by much, but it was slower. Oddball result for sure. (-fprofile-arcs and later -fbranch-probabilities). You can _not_ PGO on multithreaded code, however, it will wreck the .da files horribly. I can post my amd makefile target if you want...
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.