Do we need new PC's?

[MikeBoswell]

We are considering a dell 360 , 3 ghz machine w/ 1 gb ram and nvidia quadro fx 500 grapics, 10k scsi drives.

Currently we have HP p-class 1 gh w/ 1gb of ram w/ hp fx5 graphics, 10k scsi drives.

They still work well, but on occation when you want to move a complex face or some simiular heavy request, they are slow to respond.

I was wondering if anyone out there has gone thru a simiular upgrade and was it the bang you expected for the buck?

Thanks,

[John Scheffel]

You might want to take a look a the Viewbench results table which I just updated.

http://www.cocreateusers.org/misc/viewbench_results.html

Nobody has posted results for the quadro fx 500 yet, but based on the current table I would expect the PC you describe would get a sub-200 score. That would make it about 2-3 times faster than your current hardware. As noted in the table this is only graphics speed (redraws and dynamic viewing), but modeling operations and drawing updates should be much faster as well since they mostly depend on CPU and RAM.

You might want to consider going to 2 GB of RAM if you are working on large models. If you hear a lot of disk activity when you work on your current models using 1 GB, then you are likely swapping out to virtual memory and need more RAM.

If you do get this hardware, please run Viewbench and post the results. We don't have any data on that CPU speed range and graphics card yet. You can find instructions in this thread.

http://www.cocreateusers.org/forum/showthread.php?s=&threadid=173

[clausb]

With today's very fast graphics chipsets, the rule of thumb is that the performance bottleneck these days usually is NOT the graphics card, except for very few special cases:

Support of occlusion culling; this is particularly important if you're using the Econofast mode in Annotation.
Model clipping with capping planes enabled; a good implementation of this mode in the graphics driver can speed up these operations very considerably
Working with many graphics windows at the same time.


The fx500 should be going fine in all these respects, so I'd rather save some money for the graphics card and invest it in memory or other system components.

Claus

[MikeBoswell]

John, Claus;

Thanks for your replies.

A situation w/ a user the other day prompted my orig post. He selected ~15 slots on a flat sheet. The sheet had some repeating patterns and many more slots. He wanted to do a move face on the selected slots. It took ~42 seconds. His PC is the one I am looking to replace.

We did the same test on another workstation. dell 360, 2.7 w/1gb ram. fx500 card, scsi. The dell finished in ~20 seconds.

I dont feel the hp fx-5 card (vs the quadro fx500) was the bottlle neck in this case, either. I was giving the processor all of the credit.

So, I to restate my query, What factors would you base the need to replace a workstation?

Thanks,

[clausb]

Simplistic answer: You buy a new system if the costs for buying it and setting it up are less than what you gain my making an engineer faster in daily work.

As you have found, many modeling operations in OSDM scale quite nicely with CPU speed. This is something we have been observing for several years now. In many cases, we've even seen linear scaling.

One way to look at the decision problem is to count and measure the length of modeling operations on a typical day. For instance, let us assume that the engineer spends roughly 2 hours on raw computational (modeling) tasks on a given system. From your measurements, you can assume that the new system will speed those modeling operations up by at least a factor of two, so you gain more than one hour of engineering time per day.
Plus, of course, your engineers will then drink less coffee and save additional time on those trips to the coffee machine :-)

If your engineers are in China, saving some of their time is financially less attractive, of course, but with the typical running rates of a US or European engineer, you might be able to amortize the investment within a few months.

Claus