Jump to content

Is Nvidia Titan Z a worthwhile card for Houdini?


magneto

Recommended Posts

Probably not.

 

Houdini can use 1 GPU for graphics and 1 GPU for compute, but not both at the same time for the same task (graphics or compute). So there's very little advantage to having the increased bandwidth between the GPUs that you get with a dual-GPU card, especially since that 12GB is two banks of 6GB only accessible by the GPU it's connected to.

 

You'd be better off getting two Titan Blacks (2x $1G) vs one Titan Z ($3G), unless you don't have the space or power connectors for two. You could also shift your budget towards the task that is the more important to you by purchasing two different models - like a GEForce 760 for graphics and a Titan Black for compute, if compute was needing the extra VRAM and/or shader power.

 

To me, the Titan Z seems like a demo unit for Nvidia to show off CUDA. By selling it as a product, they can justify the demo :)

 

AMD is rumored to have a dual-Hawaii card coming out very soon (R9 295?), so perhaps it's just part of the natural competition between the two.

  • Like 1
Link to comment
Share on other sites

Thanks Mark, your advice is solid as always :)

 

What would compute tasks be related to in Houdini? Sims? Would having 2 graphics card really give you a significant boost or is it not very practical for the price?

 

I am not upgrading now but I was looking at some cards like GTX 760 and I thought you could get 2 of these for the price of a single GTX 780 which should be a lot faster even though you have less memory per card than 780?

Link to comment
Share on other sites

I did recently some OpenCL test (smokesolver) on the GTX780 and its very low vram limits its power in OpenCL. That card performs around 18% of usage during OpenCL computation and the speed was comparable with cheaper cards (just a little bit faster). So until there will be 780 with at least 6GB of vram you should go for 760 with 4GB vram (for much better price/power ratio)

  • Like 1
Link to comment
Share on other sites

Thanks pezetko. Did you manually turn on the OpenCL check boxes in those DOP nodes? Because I thought only some parts were using OpenCL and so the performance was still bottlenecked by non-OpenCL parts. I might be wrong though :)

Link to comment
Share on other sites

Thanks pezetko. Did you manually turn on the OpenCL check boxes in those DOP nodes? Because I thought only some parts were using OpenCL and so the performance was still bottlenecked by non-OpenCL parts. I might be wrong though :)

 

 

Not specific to Houdini but the CPU is also critical for GPGPU, it needs to be fast enough to push data to the GPU.

Edited by tar
Link to comment
Share on other sites

Yep vram is more useful. Though I wish more people would use cheap integrated graphics (both Intel and Amd) then wail on support to iron-out all the viewer problems... theory being that if it runs fine on the low-end it'll have no problems with the latest & greatest.

  • Like 1
Link to comment
Share on other sites

  • 1 month later...

Well, at the moment I think it's better to use the CPU openCL (if you have intel). Almost every good simulation use a large amount of RAM so 3/4/6 GB are just useless. ok, you can run a low res version just to see what happens, but most of the time the low res version should be 1/2 or luckily 1/3, otherwise you'll have big changes.

Link to comment
Share on other sites

  • 2 weeks later...

Hey Guys, I thought I would pop in my two cents,

 

Currently I run a AMD R9 290X 4GB and its a super solid GPU with houdini. The drivers are going to be supported in H14 and the OpenCL is top notch! Almost realtime for some things.

 

Thats my recoomendation! 

  • Like 1
Link to comment
Share on other sites

Hey Guys, I thought I would pop in my two cents,

 

Currently I run a AMD R9 290X 4GB and its a super solid GPU with houdini. The drivers are going to be supported in H14 and the OpenCL is top notch! Almost realtime for some things.

 

Thats my recoomendation! 

 

 

Nice! Can you upload a test scene please.  It'd be good to compare against the 7950 OpenCL. Thanks!

Link to comment
Share on other sites

There's an interesting workstation card review of the AMD FirePro W9100 (290X-based) card here:

 

http://www.tomshardware.com/reviews/firepro-w9100-performance,3810.html

 

While it doesn't have any Houdini benchmarks in it, there are several OpenCL and SPECviewperf tests vs. the FirePro W9000 (pro card equiv. of AMD 7970), Quadro K6000 (GEForce 780ti) and Quadro K5000 (GEForce 680).

Link to comment
Share on other sites

That card sounds pretty solid (depsite the super bias in the article) but considering that the consumer level cards will be supported I would wonder when i would need 16GB of VRAM and want to buck over that price! But for sure that card seems super solid!

Link to comment
Share on other sites

  • 9 months later...

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...