Jump to content

finalRender Distributed R. Farm...Athlon MP or Xeon?


Recommended Posts

I already have 4 Dual MP2800+ machines in my office, networked for finalRender's Distributed Rendering. It works like a charm.

 

I'm thinking of adding 2 more render boxes.

 

I got the quote for Dual Xeon 3.06 with 3GB RAM, Supermicro motherboard X5DAL G, 40 GB drive, CD ROM for $ 1,855. I didn't ask for case and power supply. Good deal?

 

Another quote was for a Athlon MP2800+, 3GB RAM, 40GB HD, CDROM for $ 1,945

 

Which one should I go for?

 

95% of our renders are single, hi rez images, mainly interiors with very large polygon databases, Global Illumination with tons of textures. We always exceed the 2GB limit per render.

I'm thinking....if I go with Xeon, I'll be faster than with Athlon MP...no doubt about it.

But will finalRender D.R. like having Athlon and Intel on the same image render? Some users have experienced trouble with differenct CPU cores. You'd get some buckets in the image being brighter or darker than the other.

With Athlon MP I can be safe...on the downside....it's waaay slower than Xeon and like Greg said many times, it's phased out and dead.

 

Athlon or Xeon, that is a question.

 

 

any suggestions?

 

 

Beno

Link to comment
Share on other sites

I have tried mixing amd's and intels myself in DR with finalrender and have gotten exactly what you are talking about. Some buckets are brighter/darker than others. I never got an answer from cebas as to how to fix that but I didn't dig too deep since I was only testing DR. My recommendation would be to stay safe and go with the amd's.

Link to comment
Share on other sites

I have the same opinion...maybe AMD Opteron or Athlon-64

 

I have seen your work in fr-forum ...it's very impresive!!!

good luck!!!

 

Thanks Fernando

 

...Opteron or ATHLON-64 is definitely a thought but...wouldn't that make a different CPU core from Athlon MP? I'm not sure if fR's DR feature would like that either...can someone confirm or deny this?

 

cheers

 

Beno

Link to comment
Share on other sites

I raised this issue before. I, personally, have not experienced it. But technically it is possible, and as you seem to show, practically it is possible too. AMD and Intel chips treat floating point calculations differently. Basically if you do a float calc on an AMD and one on a Intel, there is a chance that you will get a different answer. Raytracing (including GI) is heavy on the float calculations. Therefore, depending on your rendering engine, you could get a different image on an AMD machine that you would one a Intel machine. I think some lightwave people have experienced this. I raised the issue with the Chaosgroup guys and they tested Vray under many conditions and they claimed that they matched pixel for pixel. I generally trust them.

 

I would do a couple of tests. Borrow someones Intel machine. Set up a simple scene with some GI. It doesn't have to be set to the maximum accuracy. Just make sure that GI is a large part of your illumination. Render it twice. Once on the AMD, one on the Intel. Put the two images together in photoshop layers and check to see if there are differences. If there are, you have an issue.

 

Solutions? Use only one time of proc on your farm... AMD or Intel. I am not sure if the float form AMD and AMD64 are the same. If you do go with AMD64... that will be the future, but you may have to wait 2 years for the future to take full effect. The other solution is to test other rendering engines to see if they have the same issues.

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
 Share

×
×
  • Create New...