|
|
Thread Tools | Search this Thread | Rate Thread | Display Modes |
#61
|
||||
|
||||
May,
spoolsv.exe is the printer spooler process. Googling for it shows that there are some circumstances where spoolsv.exe will consume CPU time again and again; see, for example, the discussion at http://www.annoyances.org/exec/forum/winxp/t1032564815. viewbench does not try to start any print jobs, however, so my assumption initially was that the printer spooler already starts consuming CPU cycles well before you start the benchmark. Then I re-ran some tests on my machine (which runs W2K) - and all of I sudden I could reproduce those hangups. I also observed that spoolsv.exe consumed a few percent of CPU time every now and then, and that the "Printer" applet (Start/Settings/Printers) indicated that the local printer spooler had problems contacting some printers on our local print server machine. So here's my hypothesis: To collect configuration data, viewbench calls Microsoft's msinfo32 utility. This utility scans all kinds of data, probably including status data of the printer spooler. If the printer spooler process hangs for some reason, msinfo32 will also hang - and viewbench will wait forever for the configuration data and never start running its tests. When I found this, I modified some of the parameters which I pass to msinfo32 so that it will only look at certain aspects of the local machine's configuration and avoid polling the printer server status. And guess what, I cannot reproduce the hangup anymore now. I just uploaded the improved viewbench code to ftp://ftp.cocreate.com/sdtestpackage/viewbench. Hope this will also fix the issue found on your test machine. Thanks for sharing your observation! Claus Last edited by clausb; 01-20-2003 at 11:29 AM. |
#62
|
|||
|
|||
I just tried the latest benchmark on the cranky workstation here, and it still hangs on SPOOLSV.exe. The Dell 530, although it has different hardware, is pretty similarly configured, and it doesn't hang.
I'm wondering if this has anything to do with Novell Netware, as the Dell 530 is not running Novell Netware, while the Dell 420 (problem station) is. I'll try to see if there is a way I can temporarily disable the Netware client on this machine to test this out. |
#63
|
|||
|
|||
This is on an older Silicon Graphics workstation we used to run.
Model: SGI-320_ARCx86 CPU: PIII @ 500 MHz RAM: 1 GB Graphics: SGI Cobalt OS: Win NT4 w/ SP5 Version: SD 9.01D |
#64
|
||||
|
||||
Quote:
1280x1024 - 564.824 1024x768 - 526.082 Not dramatic, but faster.
__________________
John Scheffel |
#65
|
||||
|
||||
I have updated the file at http://www.cocreateusers.org/misc/viewbench_results.html
with the lastest results provided by Claus. You may have to refresh your browser window to see the latest version.
__________________
John Scheffel |
#66
|
||||
|
||||
Is CPU Speed a Major Factor?
Looking over the latest results, I was surprised by how slow the Quadro 4 results posted by jmobley are. I would have expected faster performance from these cards. They are quite a bit slower than the Dell laptop with the mobile version of the chip. Looking at the other specs the main difference seems to be the CPU speed. Most of the PCs jmobley posted are 600-900 MHz, but the Dell Laptop is 1.8 GHz.
I didn't expect the CPU to be that much of factor for graphics performance, but maybe it is. Another possibility might be the AGP slot. Most newer PCs and AGP cards are rated 4x, but older PCs were only 1x or 2x. Whatever the reason, the best specs seem to be coming from the PCs with fastest CPUs. Just for grins I did a quick plot of results versus CPU speed (see attachment). Although there are some points way off the line, there does seem to be some correlation. Of course it could just be that the PCs with faster CPUs are newer and in general have better graphics cards, faster bus speeds, faster memory, etc. Another surprise is the Erazor X2 results. These are pretty old cards to be ranking where they are, and were not targetted to the CAD market.
__________________
John Scheffel |
#67
|
||||
|
||||
Is CPU Speed a Major Factor?
Looking over the latest results, I was surprised by how slow the Quadro 4 results posted by jmobley are. I would have expected faster performance from these cards. They are quite a bit slower than the Dell laptop with the mobile version of the chip. Looking at the other specs the main difference seems to be the CPU speed. Most of the PCs jmobley posted are 600-900 MHz, but the Dell Laptop is 1.8 GHz.
I didn't expect the CPU to be that much of factor for graphics performance, but maybe it is. Another possibility might be the AGP slot. Most newer PCs and AGP cards are rated 4x, but older PCs were only 1x or 2x. Whatever the reason, the best specs seem to be coming from the PCs with fastest CPUs. Just for grins I did a quick plot of results versus CPU speed (see attachment). Although there are some points way off the line, there does seem to be some correlation. Of course it could just be that the PCs with faster CPUs are newer and in general have better graphics cards, faster bus speeds, faster memory, etc. Another surprise is the Erazor X2 results. These are pretty old cards to be ranking where they are, and were not targetted to the CAD market.
__________________
John Scheffel |
#68
|
|||
|
|||
Quick question. Is it the same method of running Viewbench between a Win2K workstation and a HP-UX box? I get an error message when I type:
(load "viewbench") in SD 9. It says it was a LISP error and was unable to load the library file. Am I doing something wrong? |
#69
|
||||
|
||||
May,
running viewbench works the same on all platforms. The version of viewbench which I posted to our external web site was compiled using OSDM 2002+, so there is, in fact, a small chance that it won't work properly in older versions. What does the LISP error message say exactly? Is there any additional output in the shell window where you started OSDM? Claus |
#70
|
|||
|
|||
The exact error message says:
LISP error: Loading shared library failed for viewbench. |
#71
|
|||
|
|||
Model: Premio (custom built)
CPU: Intel Pentium III 700 MHz RAM: 512 MB Graphics: ATI FireGL 8700 64MB OS: Microsoft Windows NT Ver 4.0 Build 1381 SP 6 Version: SD 9.01 |
#72
|
|||
|
|||
Model: IBM Intellistation M Pro
CPU: Intel Pentium IV 2.0 GHz RAM: 512 MB Graphics: NVIDIA Quadro4 550 XGL 64MB OS: Microsoft Windows 2000 Professional 5.0 SP 2 Build 2195 Version: SD 9.01 I think this proves that the processor has a lot to do with this test. |
#73
|
||||
|
||||
Quote:
I also find the correlation between viewbench results and CPU speed highly interesting. Food for thought. Claus |
#74
|
|||
|
|||
No problem, Claus. When we finally migrate to OSD 11.X (in a few months, I hope!), we'll have to migrate the Unix boxes to HP-UX 11. We can run the benchmark then.
|
#75
|
|||
|
|||
Here's one more, off a new Dell 650 workstation. It's doing worse than our older 530 stations, so I'm wondering if there is a problem with less-than-optimized drivers or something.
Configuration as follows: Dell 650 P4 2.8 GHz 1 GB RDRAM Wildcat 7110 Windows 2000 w/ SP2 |
Currently Active Users Viewing This Thread: 1 (0 members and 1 guests) | |
|
|