SPECapc SW 2007
in [SolidWorks]
|
From: Anna Wood on 13 May 2007 03:03 Here are my scores for my other home machine, SW2007 SP2.1, EVGA e- GeForce 7100 GS video card. The original Nvidia FX550 died and I replaced it with a low end gamer card. FWIW, Anna Wood SolidWorks 2007 Workstation Benchmark User Name : ------------ Computer Name: ------------ Manufacturer : Dell Inc. Model : Precision WorkStation 390 OS : Microsoft Windows XP Professional OS SP : Service Pack 2 CPU : Intel(R) Core(TM)2 CPU 6600 @ 2.40GHz # of CPU : 2 Memory : 2045 **** Overall Test Results **** Note: All results are in seconds. Lower scores are better. Test Number 1 Test Total = 227.5 Graphics = 104.45 CPU = 53.58 I/O = 69.47 Test Number 2 Test Total = 226.46 Graphics = 104.41 CPU = 52.83 I/O = 69.22 Test Number 3 Test Total = 226.11 Graphics = 103.96 CPU = 52.97 I/O = 69.18 Test Number 4 Test Total = 225.53 Graphics = 103.76 CPU = 52.72 I/O = 69.05 Test Number 5 Test Total = 226.05 Graphics = 103.9 CPU = 52.93 I/O = 69.22 Test Averages for 5 tests(s). Test Total = 226.34 Graphics = 104.1 CPU = 53.01 I/O = 69.23
From: TOP on 13 May 2007 07:14 Here is the cream of the crop on the 2005 Benchmark. Apparently CPUs and IO have gotten a whole lot faster and graphics cards have gotten a whole lot slower. SPECapcSM SolidWorks2005 Benchmark Result Dell Precision Mobile Workstation M90 2.33GHz NVIDIA Quadro FX 2500M Submitted by: Dell, Inc. Test Number 1 Test Total = 447 Graphics = 88.75 CPU = 183.03 I/O = 175.22 Test Averages for 1 tests(s). Test Total = 447 Graphics = 88.75 CPU = 183.03 I/O = 175.22 Dell Precision Workstation 690 3.00 GHz nVidia Quadro FX 3500 Submitted by: Dell, Inc. Test Number 1 Test Total = 325.68 Graphics = 64.96 CPU = 132.23 I/O = 128.49 Test Averages for 1 tests(s). Test Total = 325.68 Graphics = 64.96 CPU = 132.23 I/O = 128.49
From: Art Woodbury on 13 May 2007 07:49 In article <1179021594.768590.27830(a)k79g2000hse.googlegroups.com>, kellnerp(a)cbd.net says... > Compared with the 2005 benchmark your processor scores are awesome > unless you somehow got graphics and CPU mixed. > > TOP > > My report was a cut-and-paste of the text file created by the test. I don't have any previous results for comparison, mainly because I never thought the SPEC tests related directly to real CAD usage. Art P.S. The graphics and CPU fans spooled up higher than I've ever heard them go. It didn't sound like a Dell product -- more like Pratt & Whitney.....
From: TOP on 13 May 2007 09:07 I am trying to make a couple points about SPECapc for SolidWorks. 1. SPECapc is really not what you would call a benchmark in the strictest sense of the word. Maybe a relative benchmark is as far as I would go. You can't compare results from one year's benchmark to another. It doesn't run reliably across different releases of SW. I can't run the SPECapc2007version on SW2003. And if I run SW2007 on SPECapc 2005 there will be the overhead of file conversion. 2. SPECapc does not reflect real world. If you look at the CPU scores between 2005 and 2007 you will see a tremendous difference. Likewise the graphics scores. If you take a 2005 assembly, one that you struggled with in 2005, convert it to 2007, do you see the same level of improvement that SPECapc shows on the new hardware you are testing? 3. SPECapc does not reflect the real world in that it is heavily biased to graphics card performance. Just visit the results page on their site or look at the source code. Years ago when SPEC was a hot item on the NG I did a lot of testing with it. I found that a good graphics card could make a mediocre CPU look stellar and a bad graphics card would make a good CPU look like a dog. In the real world you can work around a slow graphics card by setting SW graphics to their lowest settings. Sure circles will look like hexagons, but work can get done. You can't work around a poor performing CPU and that is where a lot of time is spent waiting on large assemblies and complex parts. 4. SPEC does not benchmark a significant subset of SW functionality. And it doesn't go out of it's way to test the functionality in SW that causes bottlenecks.You won't see a lot of complex large assembly drawings or assemblies with lot's of mega-multi configuration parts in them. A lot of it is concerned with graphics eye candy. 5. Given that SPECapc 2007 is just out and that 2008 is now in beta it has a limited lifetime so that in six months it will be relegated to just another curiousity. Here is a list of uses for SPECapc: Home hardware evaluation Business purchasing evaluation Internal hardware development Internal ISV development User system evaluation/optimization Research study Vendor competitive analysis Magazine or online publication In other words you can't use it to evaluate one version of SW against another. I should mention that you hit on one of the strengths that SPEC has always had, the ability to really stress a system. Any system that won't run through all five iterations of SPEC probably has problems. I have used SPEC in the past to evaluate registry tweaks. The newer versions aren't as handy because they take so long to run. It is the first thing I would run if hardware problems are suspected. Why SW raises the CPU temperature so much still is a mystery to me. But it does and more so that other types of software like FEA. TOP
From: Dale Dunn on 13 May 2007 12:40
> Why SW raises the CPU temperature so much still is a mystery to me. > But it does and more so that other types of software like FEA. This one I think I know. FEA is almost entirely floating-point math and block memory transfers, which only exercises those areas of the chip. SW will exercise the integer and logic portions of the chip as well and probably work the caches harder, thereby causing another portion of the CPU die to generate heat. This will be more pronounced in recent chip designs that can lower the clock speed or even switch off unused parts of the chip. Currently, Intel's Core processors are the most agressive about this. It's a significant portion of their power saving (in addition to the 65 nm process, etc.). |