results -> AMD Phenom 2 945 -vs- Intel Core2 Quad 9400

Looking for new hardware to run WRF? Intel or AMD? Check this forum.
Post Reply
cubemonkey
Posts: 13
Joined: Wed Aug 26, 2009 1:48 am

results -> AMD Phenom 2 945 -vs- Intel Core2 Quad 9400

Post by cubemonkey » Mon Aug 31, 2009 3:46 am

I recently purchased both a Intel Core 2 Quad 9400, and an AMD Phenom 2 945 for the purpose of modeling with WRF. I was initially going to go with two Intels, as they scored higher in various benchmarks online. But one of these systems was going to replace the guts of my micro ATX media PC, and there's a much larger selection of micro ATX boards for AMD processors.

Plus, I read a couple of places that the Phenom 2 was a bit better at not missing a beat playing videos or music while under major load. So I figured it might be a little slower, but if my wife is watching something on the media PC, there's a better chance I won't hear any yelling when I kick off a WRF run. A worthwhile tradeoff :)

So they both arrived, and I got them both up and running, and decided to do an identical run on each and see how they compared.

So, the setup: The master is a quad core Xeon sitting in another room. Both the 945 and the 9400 systems are clients sitting on gigabit, on the same switch, have the same disk configuration, and run the exact same type and amount of memory. Only difference really is the motherboard (the AMD is a $60 micro ATX).

I did the same run, one on each processor and noted the rsl.log output ..

Intel Core 2 Quad 9400:

Code: Select all

Timing for main: time 2009-08-30_20:30:33 on domain   3:    1.72550 elapsed seconds.
Timing for main: time 2009-08-30_20:31:06 on domain   3:    1.06192 elapsed seconds.
Timing for main: time 2009-08-30_20:31:40 on domain   3:    1.05789 elapsed seconds.
Timing for main: time 2009-08-30_20:31:40 on domain   2:    7.24999 elapsed seconds.
Timing for main: time 2009-08-30_20:32:13 on domain   3:    1.05318 elapsed seconds.
Timing for main: time 2009-08-30_20:32:46 on domain   3:    1.05384 elapsed seconds.
Timing for main: time 2009-08-30_20:33:20 on domain   3:    1.06264 elapsed seconds.
Timing for main: time 2009-08-30_20:33:20 on domain   2:    5.41951 elapsed seconds.
Timing for main: time 2009-08-30_20:33:53 on domain   3:    1.05781 elapsed seconds.
Timing for main: time 2009-08-30_20:34:26 on domain   3:    1.05526 elapsed seconds.
Timing for main: time 2009-08-30_20:35:00 on domain   3:    1.05903 elapsed seconds.
Timing for main: time 2009-08-30_20:35:00 on domain   2:    5.42381 elapsed seconds.
Timing for main: time 2009-08-30_20:35:00 on domain   1:  18.04014 elapsed seconds.
Timing for main: time 2009-08-30_20:35:33 on domain   3:    1.05769 elapsed seconds.
Timing for main: time 2009-08-30_20:36:06 on domain   3:    1.05874 elapsed seconds.
Timing for main: time 2009-08-30_20:36:40 on domain   3:    1.05607 elapsed seconds.
Timing for main: time 2009-08-30_20:36:40 on domain   2:    5.49184 elapsed seconds.
Timing for main: time 2009-08-30_20:37:13 on domain   3:    1.07111 elapsed seconds.
Timing for main: time 2009-08-30_20:37:46 on domain   3:    1.05897 elapsed seconds.
Timing for main: time 2009-08-30_20:38:20 on domain   3:    1.05737 elapsed seconds.
Timing for main: time 2009-08-30_20:38:20 on domain   2:    5.44671 elapsed seconds.
Timing for main: time 2009-08-30_20:38:53 on domain   3:    1.05412 elapsed seconds.
Timing for main: time 2009-08-30_20:39:26 on domain   3:    1.05585 elapsed seconds.
Timing for main: time 2009-08-30_20:40:00 on domain   3:    1.05819 elapsed seconds.
Timing for main: time 2009-08-30_20:40:00 on domain   2:    5.41011 elapsed seconds.
Timing for main: time 2009-08-30_20:40:00 on domain   1:   17.73353 elapsed seconds.

AMD Phenom 2 945:

Code: Select all

Timing for main: time 2009-08-30_18:50:33 on domain   3:    0.91235 elapsed seconds.
Timing for main: time 2009-08-30_18:51:06 on domain   3:    0.78286 elapsed seconds.
Timing for main: time 2009-08-30_18:51:40 on domain   3:    0.78552 elapsed seconds.
Timing for main: time 2009-08-30_18:51:40 on domain   2:    4.30799 elapsed seconds.
Timing for main: time 2009-08-30_18:52:13 on domain   3:    0.78371 elapsed seconds.
Timing for main: time 2009-08-30_18:52:46 on domain   3:    0.78609 elapsed seconds.
Timing for main: time 2009-08-30_18:53:20 on domain   3:    0.78049 elapsed seconds.
Timing for main: time 2009-08-30_18:53:20 on domain   2:    4.15188 elapsed seconds.
Timing for main: time 2009-08-30_18:53:53 on domain   3:    0.78400 elapsed seconds.
Timing for main: time 2009-08-30_18:54:26 on domain   3:    0.78428 elapsed seconds.
Timing for main: time 2009-08-30_18:55:00 on domain   3:    0.78546 elapsed seconds.
Timing for main: time 2009-08-30_18:55:00 on domain   2:    4.14905 elapsed seconds.
Timing for main: time 2009-08-30_18:55:00 on domain   1:   13.66494 elapsed seconds.
Timing for main: time 2009-08-30_18:55:33 on domain   3:    0.78544 elapsed seconds.
Timing for main: time 2009-08-30_18:56:06 on domain   3:    0.78582 elapsed seconds.
Timing for main: time 2009-08-30_18:56:40 on domain   3:    0.80785 elapsed seconds.
Timing for main: time 2009-08-30_18:56:40 on domain   2:    4.10314 elapsed seconds.
Timing for main: time 2009-08-30_18:57:13 on domain   3:    0.78760 elapsed seconds.
Timing for main: time 2009-08-30_18:57:46 on domain   3:    0.78541 elapsed seconds.
Timing for main: time 2009-08-30_18:58:20 on domain   3:    0.78586 elapsed seconds.
Timing for main: time 2009-08-30_18:58:20 on domain   2:    4.08322 elapsed seconds.
Timing for main: time 2009-08-30_18:58:53 on domain   3:    0.78804 elapsed seconds.
Timing for main: time 2009-08-30_18:59:26 on domain   3:    0.78720 elapsed seconds.
Timing for main: time 2009-08-30_19:00:00 on domain   3:    0.78676 elapsed seconds.
Timing for main: time 2009-08-30_19:00:00 on domain   2:    4.09103 elapsed seconds.
Timing for main: time 2009-08-30_19:00:00 on domain   1:   13.33776 elapsed seconds.
Result? The AMD is about 25% faster while modeling. I was very surprised - especially at the sheer size of the gap. I tested a couple of other runs just to verify. Had the same results.

The AMD is slower at the very beginning when initializing the nests, but after it launched into the actual modeling, there's no mistake it was the much faster processor for doing this stuff. I know the intel would come out ahead in some gaming benchmarks, etc.. but for the purpose of modeling, at least according to this test, the AMD definitely seems to be the way to go (and it's a little cheaper to boot!).

So there you go.. your mileage may vary, but that's my experience for whatever it's worth...

-Bill

(edit: squash some typos)

wettersat
Posts: 32
Joined: Fri Aug 21, 2009 9:30 pm

Re: results -> AMD Phenom 2 945 -vs- Intel Core2 Quad 9400

Post by wettersat » Mon Aug 31, 2009 6:37 am

Bill, try to run original benchmark of WRF system. Then it is easier to compare the results with other systems. Benchmark output is a file with all important data (cpu, calculation time, .....)

Gerald

cubemonkey
Posts: 13
Joined: Wed Aug 26, 2009 1:48 am

Re: results -> AMD Phenom 2 945 -vs- Intel Core2 Quad 9400

Post by cubemonkey » Mon Aug 31, 2009 12:25 pm

That's a good idea.. I had forgotten about that benchmark. When I get some time, I'll do that and post the results.

-Bill

meteoadriatic
Posts: 1603
Joined: Wed Aug 19, 2009 10:05 am

Re: results -> AMD Phenom 2 945 -vs- Intel Core2 Quad 9400

Post by meteoadriatic » Mon Aug 31, 2009 3:14 pm

We could make one thread specifically to this benchmarks here on forum and try to do as many testing of different machines as we can. It would be nice hardware buying guide for modelling purporse! What do you think?

cubemonkey
Posts: 13
Joined: Wed Aug 26, 2009 1:48 am

Re: results -> AMD Phenom 2 945 -vs- Intel Core2 Quad 9400

Post by cubemonkey » Wed Sep 02, 2009 1:35 am

Sounds good to me :) I haven't gotten around to doing the official benchmark yet, but I'll post it there when I do..

-Bill

pattim
Posts: 199
Joined: Sun Jun 24, 2012 8:42 pm
Location: Los Angeles, CA, USA

Re: results -> AMD Phenom 2 945 -vs- Intel Core2 Quad 9400

Post by pattim » Fri Jun 29, 2012 6:54 pm

cubemonkey wrote:Sounds good to me :) I haven't gotten around to doing the official benchmark yet, but I'll post it there when I do..

-Bill
I just tried "ganged" vs "unganged" memory controllers on my 1090t and it made a ~50% increase in speed to ungang the memory controllers. (Increasing the NorthBridge clock didn't help when the controllers were ganged.)

Post Reply