Author: Robert Hyatt
Date: 20:22:33 03/18/03
Go up one level in this thread
On March 18, 2003 at 20:07:13, Aaron Gordon wrote: >Here are some tests that I've run on one of my machines. It's an AthlonXP 1900+ >@ 1.6GHz (non-overclocked), 133fsb on an Abit KT7a (KT133A) motherboard. The ram >type is regular SDRAM (non-DDR). 3 Dimms used, one 256mb and two 128mb's. So 3 >slots in use, 6 banks in use total (0-5). > >4-way interleave >Sisoft memory test: >ALU w/ SSE : 1002mb/s >ALU w/o SSE or MMX : 577mb/s > >CraftyK7 19.3: 1,046,116 nodes/sec > > >2-way interleave >Sisoft memory test: >ALU w/ SSE : 1002mb/s >ALU w/o SSE or MMX : 547mb/s > >CraftyK7 19.3: 1,046,116 nodes/sec > > >No Interleaving >Sisoft memory test: >ALU w/ SSE : 993mb/s >ALU w/o SSE or MMX : 517mb/s > >CrafyK7 19.3: 1,046,116 nodes/sec > > >As you can see interleaving helped nothing for crafty. Doubling ram bandwidth >while keeping all hardware, memory timings, etc identical also proved no >increase (unless you want to count 0.14% as something out of the margin of error >of the benchmark). That is interesting, but what it means I have no idea. I haven't run the test in a long time, but it is easy to get the new pentium processors to count and report the number of cache line misses while a program runs. Last time I tried it the number was huge, as expected, since nearly every hash probe is a miss for certain, and all the large tables used for move generation, bit counting, and so forth also drive this number up... > >No, before you claim my interleaving is 'broken' I should mention my Quake3 >scores. By enabling 4-way over no interleaving I was able to increase my frames >per second on that system by around 20%. Quake3 is extremely memory >speed/bandwidth/latency bound. You modify ANYTHING memory related and you get >drastic framerate changes. > 20% seems very low for going from nothing to 4-way, with a memory-intensive program like quake. There are some good memory benchmarking programs around that might be even more revealing about how effective your interleaving is. >To give you some idea.. if you upgrade your cpu from an AthlonXP 1500+ (1.33GHz) >to a 1900+ (1.6GHz) you get about a 9-10% boost in framerate (for Quake3). 9-10% >from a 20% boost in clock speed. Again, just by enabling 4-way interleaving it >jumped up 20% without changing anything else. So, without a doubt, interleaving >DOES work on that board. > >Ah, almost forgot to mention.. even my old FIC 503+ socket-7 motherboard (VIA >MVP3 chipset) supports 4-way interleaving. > > >One thing I did find doing some testing on much older cpus was L1 & L2 cache >sizes vs Crafty NPS. I don't have the numbers on-hand but I'll find them if >you'd like. Anyway, a 486 with 8K was about half as fast in Crafty as a 486 with >16K L1.. this is at the same MHz. Adding L2 cache further increased performance >(something on the order of 60-80%). This was with extremely small L1/L2 cache >sizes. On todays modern chips I believe you'll see next to no difference between >the same type of cpu with a different cache configuration. Take the Duron (128K >L1, 96K L2) vs a Thunderbird (128K L1, 256K L2) for example. Gets identical NPS >in Crafty. > >Also, of course, I'm speaking about normal end-user systems. Not $60,000 >itaniums, multi-million dollar cray machines, just normal 32bit x86 computers. >So before stating you get some odd percentage increase with such & such cache, >yadda memory timings, it's best to tell people, "On such & such Itanium it got >x% increase in nodes/sec, YMMV on a 32bit x86 however" or something similar. >Point being... someone may go out and get an entirely new board, ram, keeping >their same cpu, etc trying to increase memory bandwidth drastically. To their >surprise when they run the crafty benchmark all they'll see are the same numbers >(or something extremely close). All this leads to is disappointment and >partially wasted money (partially because they may use something else other than >Crafty that does get a boost from the increased bandwidth).
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.