benchmark results

Yep. Check your inbox for a reply.

- Eddy

message

Everything interacts with everything. If you run the benchmark on a couple of processors and a couple of graphics cards, you'll find that a faster processor will improve all of the scores. A faster graphics card will also increase all the scores, though usually not as much. I never ran it with different disc drives or faster RAM, so I'm only guessing that any effect would also show up on all three scores.

As I recall, the way they get the three subsystem scores is by checking the time for actions that lean particularly heavy on that subsystem. So how fast a part spins shows up in the graphics score, how fast a part regenerates shows up in the CPU score, and how fast a file loads shows up in the I/O score.

Jerry Steiger Tripod Data Systems

Here are the rest of my test results:

convert iges file #1: HP 68 Xi 47 Diff 21 (31%) save file #1: HP 7 Xi 6 Diff 1 (14%) open file #1: HP 3 Xi 3 Diff 0 (0%)

convert iges file #2: HP 434 Xi 318 Diff 116 (27%) save file #2: both hung (after saving) open file #2: HP 28 Xi 18 Diff 10 (36%)

convert UG file: HP 194 Xi 120 Diff 74 (38%) save file: HP 21 Xi 21 Diff 0 (0%) unsupress features: HP 78 Xi 59 Diff 19 (24%) save file: HP 77 Xi 63 Diff 14 (18%) open file: HP 84 Xi 67 Diff 17 (20%)

open top level assy: HP 83 Xi 56 Diff 27 (33%) rebuild: HP 2 Xi 2 Diff 0 (0%) save file: HP 22 Xi 17 Diff 5 (23%) open sub-assy drawing: HP 49 Xi 32 Diff 17 (35%) update sheet 2: HP 102 Xi 80 Diff 22 (22%) activate sheet 1: HP 8 Xi 5 Diff 3 (38%) update sheet 1: HP 66 Xi 52 Diff 14 (21%) save file: HP 14 Xi 11 Diff 3 (21%)

TOTAL HP 1340 Xi 977 Diff 363 (27%)

ship in bottle (50) rebuilds average 4 times HP 37.8 Xi 30.1 Diff 7.7 (20%)

If anyone has any other tests that you think might be good to run, let me know. These numbers are not scientific - just me and a stopwatch. I think that they are good enough to prove to us that the 3400+ is the way to go. We are going to swap our FX3000 card for a FX1100 because I don't think we are going to see enough speed increase with the 3000 to justify the cost. I will re-post to let you know what the FX1100 does to these tests. I couldn't get the SPEC benchmark to run on 2004 but I assume that the speed difference shown on 2003 should be similar if run on 2004.

Bob

You are right Bob, the speed difference with 2004 will be similar. But 2004 will be 5-10% slower than 2003. At this point, there are enough advantages to me with SP2.1 to justify going with 2004 but that's a judgement call. We do surfaces, cross sections, etc. and 2004 will help.

Also, I agree that the FX1100 over the FX3000 might be the way to go. I went with FX1000 and they rock! Anything more would have been a waste for us but maybe the 3000 for people who work with obscene assy's. Here they usually top out at under a couple thousand parts but because we use surfaces and "organic" shapes they are pretty challenging. I can spin an entire fully resolved assy around like it was a washer using the 1000. That is a cool feeling :)

BTW - out of jealousy I ran all the benches again and my scores are nearly equal to yours. I consistently get into the 28's with the "ship" using SW2003 and consistently 30's with SW2004. That means the higher price of the 3400+ may not be justified. Save a couple hundred and go 3200+ (when I built these a few weeks ago the 3400 wasn't available so I paid the same price for the 3200 as what the 3400 now sells for - Doh). Just make sure you get the larger cache and not the dumbed down "economy" chip.

- Eddy

going back to my youth :)