Octo CORE Disappointment

[Ramon Aranguren]

Hello folks, I have a PC that I built myself with a core duo and 4 and 3 Gb of RAM, I've been very happy with it for more than a year now. Yesterday I received a computer sent from my company with Dual XEONS E5472 @ 3Ghz, 16 Gb or RAM, Nvidia quadro 4500.

 

Well the thing is that I did a quick test last night of an image that takes around 57 seconds on my old computer, and surprisingly it lasted 48 seconds on the new one, I thought that it was going to be like 30 seconds tops, so my question here are:

 

Am I especting too much from the machine?

Is there something else I don't know, like some settings on vray or max?

Is the new PC wrongly constructed?

 

Both computers are running on windows vista 64, 3ds max 2008 and vray 1.5.17.

 

Thanks for any advice and/or suggestions

[Phabz]

Maybe you should run a test with a scene that on your original machine takes 30-45min... I am sure you will see some improvements.

Also, I think I saw either here or in the vray forums a benchmark scene that you can download.

If you are not that satisfied with the new Xeons, you can ship it to me?!

Let us know the result of the tests

[AJLynn]

There's a Vray 1.5.17? Is that a beta?

 

Anyway. What John said - part of the time is taken up by translating the scene, etc. - processes that don't multithread well or at all. The rendering itself is where you save time, and this will be more noticeable the longer the render takes. If you really got the 8-core to make 57 second renders take longer, what you'll need to do is message John to get his address and ship the 8-core directly to him. The etiquette is that he should reimburse you for the shipping costs.

[Fran]

I would make sure the BIOS is up to date. I've heard of an issue with the 54XX's on an ASUS board where Cinebench was much slower than expected. Updating the BIOS solved the issue.

[SandmanNinja]

Oh Fran!

 

My heart skipped a beat when I read your post.

Too bad Virginia is so far away...

 

[litleboy]

does vray/max use all the cores when preparing the scene to render? If not that could be the problem.

[Ramon Aranguren]

Thanks for the answers... well I wouldn't really care to give it away since I didn't buy it, it's not even really mine. And about the scene the preparing issue sounds logical, I'll try to look for that benchmark scene and will post the results here if anyone's interested.

[Hazdaz]

So is this a case of diminishing return, or simply a system that wasn't set up properly?

[Crazy Homeless Guy]

If the scene was already rendering at 57 seconds on your old machine, I am going to guess that both, the old machine was a decent 32 bit, and that the scene is simple. You are going to see the biggest benefit of 64 bit when you start rendering heavy scenes with lots of geometry.

 

I am guessing your old machine was some what decent since you were running Vista on it.

[BOXXLABS]

Hello all -

 

This is my first post here on CGarchitect but I've been lurking for a long while and have been active on some other CG-related communities for years...

 

Regarding 8-core performance: depending on the scene construction, there will be some number of seconds (or even minutes) that are spend tessellating and pre-processing the scene's lights, shadows, etc. Many of these processes are single-threaded.

 

So if it's a relatively small scene, you might see very little in the way of overall render time performance improvements over your older machine.

 

Something you might try if you are using the VRAY "light cache", try nailing the "number of passes" setting to "8". This will help a little.

 

Also, try to minimize the use of large displacement maps and shadow maps. It may be faster to actually model and render the actual geometry/shadows using the raytracer...

 

Also, if your scenes use a lot of very large textures, you might want to make sure you have a fast disc subsystem or else rendering will stall until all the maps have been loaded, translated, etc...

 

In any case, as was stated before, you should certainly see much better perofrmance on 8 cores when you go to render larger or more detailed scenes and use the appropriate settings.

 

Cheers,

 

Adam

BOXXlabs

Edited July 7, 2008 by BOXXLABS

[Ky Lane]

I went from a core2duo to a dual quad xeon system (mac pro) and the difference is immeasurable. Scenes that would take 2-3hrs at high res/settings now took 30-45 mins. Ill never look back.

[Tommy L]

I went from a core2duo laptop to 32 core DR. I'll definitely never look back...