bdeloria
Mechanical
- Sep 3, 2007
- 12
Quick little bit of background information. We're using Unigraphics NX, no TC. We also have a mix of 32 and 64 bit workstations, running NX1 looking to migrate to NX5.
I've tried in the past to try to find the best way to benchmark machines. The only tool that I can seem to come up with or moreso as I make adjustments in the machine settings is the SPEC tests.
They provide results for me but I often question how useful they are. I realize that things such as network performance are also going to be a factor but as I'm only trying to address the issue of general system configuration at this moment this will have to do.
I have in the past simply changed configuration settings and simply loaded up large assemblies over and over again. This however does leave you with the problem of files loading off of the network and too many variables. It's also not suitable to do during business hours as machine loads can change from second to second.
So are these tests worthwhile? Also when does the "multi-threaded" aspect of this come into play? There's multiple options and I'm looking for some feedback on this.
Also one thing I've notice, while running XP-32bit, depending on my memory configuration my score will change.
i.e. I'm running a Dell T7400 dual quad core xeons with a FX4600 card.
I also realize that XP-32 is still only going to be able to recognize >4GB of ram.
Depending on the configuration of the memory the scores change. For example:
Qty (4) 2GB dimms - only addressses 4GB in windows however shows up as quad channel in the bios for memory access
Qty (2) 2GB dimms - 4GB ram, shows up as dual channel in the bios
Qty (4) 1GB dimms - 4GB ram, shows up as quad channel in the bios.
Now when I'm running 4x1GB I get a significant improvement in scores running multithreaded tests. The scores are similar for the 2GB dimms in multithreaded tests. The scores are similar across the board in single threaded tests.
The only thing that I can surmise as a reason for this is that the OS is only accessing 2 of the dimms in the 4x2GB configuration.
Thanks,
Brian
I've tried in the past to try to find the best way to benchmark machines. The only tool that I can seem to come up with or moreso as I make adjustments in the machine settings is the SPEC tests.
They provide results for me but I often question how useful they are. I realize that things such as network performance are also going to be a factor but as I'm only trying to address the issue of general system configuration at this moment this will have to do.
I have in the past simply changed configuration settings and simply loaded up large assemblies over and over again. This however does leave you with the problem of files loading off of the network and too many variables. It's also not suitable to do during business hours as machine loads can change from second to second.
So are these tests worthwhile? Also when does the "multi-threaded" aspect of this come into play? There's multiple options and I'm looking for some feedback on this.
Also one thing I've notice, while running XP-32bit, depending on my memory configuration my score will change.
i.e. I'm running a Dell T7400 dual quad core xeons with a FX4600 card.
I also realize that XP-32 is still only going to be able to recognize >4GB of ram.
Depending on the configuration of the memory the scores change. For example:
Qty (4) 2GB dimms - only addressses 4GB in windows however shows up as quad channel in the bios for memory access
Qty (2) 2GB dimms - 4GB ram, shows up as dual channel in the bios
Qty (4) 1GB dimms - 4GB ram, shows up as quad channel in the bios.
Now when I'm running 4x1GB I get a significant improvement in scores running multithreaded tests. The scores are similar for the 2GB dimms in multithreaded tests. The scores are similar across the board in single threaded tests.
The only thing that I can surmise as a reason for this is that the OS is only accessing 2 of the dimms in the 4x2GB configuration.
Thanks,
Brian
