Thea Render at BaseCamp 2012
-
About GPU renderer tecnical questions, please wait for our official announcement. We can discuss those in detail at the time, but I can tell one thing at the moment, a CUDA card is needed. I'd go with GTX 5xx or GTX 6xx, but some older cards are fine (but not first generations).
-
Notareal,
How do you tell whether or not your card is a first generation product?
-
@hieru said:
Notareal,
How do you tell whether or not your card is a first generation product?
As some technical specs might change it's hard to say definitive limit, but at the moment the compute capability version 1.2 is required (so 1.0 and 1.1 cads are out). You can find a compute capability versions list at http://en.wikipedia.org/wiki/CUDA or http://developer.nvidia.com/cuda/cuda-gpus
http://www.techpowerup.com/gpuz/ can be used to verify compute capability version. -
Thanks for the useful info Notareal.
Looks like I should be OK as far as trying out the GPU engine is concerned. Now I just need to figure out the best upgrade for the future.
-
GTX 680 with 4 GB ram would be ideal.
-
Just watched 'em - unreal Unreal!
Really cool stuff, can't wait for it to come out....I'm looking at a GPU upgrade in a bit, that's for sure! -
@hieru said:
Notareal,
How do you tell whether or not your card is a first generation product?
It will say it on the card itself with the serial number. Not sure but I think you can also get the specs from system preferences and compare online.
I used Octane Render for Sketchup when it came out with a GTX5 and that was two years ago so most newer cards should be fine (I think).
Cant wait for this its been a long time coming Thea but you have pulled it off in the end!!
-
@notareal said:
GTX 680 with 4 GB ram would be ideal.
I figured that I'd be looking at a 580 or 680, but I'm sure I remeber someone saying that the 680 was very poor at GPGPU as it's more of a gamer card?
-
@hieru said:
@notareal said:
GTX 680 with 4 GB ram would be ideal.
I figured that I'd be looking at a 580 or 680, but I'm sure I remeber someone saying that the 680 was very poor at GPGPU as it's more of a gamer card?
In certain GPGPU operations 680 is indeed "crippled" and inferior to 580, but fortunately this has no affect on Thea GPU engine. The preview video was done using GTX 580. After the release some speed comparisons can be done.
-
Nice! I believe iRay is the only GPU raytracer to require workstation-type GPUs. The rest work best with gaming-type ones.
-
Thanks for the clarification Notareal!
-
Excuse my ignorance, but I couldn't really find any info on this:
Is this the new version of the Thea SketchUp plugin? Or is it a new independent product, which will be charged for? IOW, if I buy Thea + SU plugin now, will I have to pay again to get this new tool?
BTW, a pity that there is no Thea for LightWave.. -
It's a new version of the current plugin and (AFAIK but there is notareal to correct me) these features above (both plugin and GPU rendering) will bee included in a free upgrade for existing license holders.
-
@notareal said:
In certain GPGPU operations 680 is indeed "crippled" and inferior to 580, but fortunately this has no affect on Thea GPU engine.
Could you elaborate on that?
I just bought a 3gb GTX580 for its "said" greater GPGPU performance.
Was that a bad move...?
-
Thanks Csaba
Guess it's time to delve into the mysterious realm of graphics cards again, to pick the right one
Is this site http://www.videocardbenchmark.net/high_end_gpus.html worth looking at?
Looks like GTX670 would be a good choice? -
how can i check if my motherboard would support one of these high end GTX cards??? You see.. i am noob.
-
GTX580 is not a bad at all. But as used architecture between kepler and fermi is rather different, cuda core to core comparison is difficult. There been talk that GTX6xxx series is inferior to GTX5xxx series, but this seems to be true only in some spesific instruction that Thea does not use. When some new CUDA features are used, GTX6xxx has some benefits. Still Thea tries to be usable for rather wide varity of CUDA cards, so some of these features might not be used at the moment.
As card vs card comparisons are hard to made in this developement phase (propably missing some optimizations and so), I'd rather not say if there any meaningfull difference between xtg580 and xtg680 - as xtg580 works fine with Thea.
Anyhow this does just proof that GPGPUs are so fast developing area that software developement has some challenges and hardware spects do change so fast. I believe this was the reason for originally planned Thea OpenCL renderer to CUDA render move. Naturally OpenCL is not in my undertanding excluded from future use, but it just, unfortunatelly, looks not that mature to be used at the moment. -
-
@jenujacob said:
how can i check if my motherboard would support one of these high end GTX cards??? You see.. i am noob.
Any modern motherboard core 2, i5, i7 should be fine or att least there should be a PCI Express bus (v.2.0). Proper PSU is more important, often computers do come with rather low end power suply.
Usually these high end GPU cards do require around 550W and greater PSU... it's best to look on rated specifications at http://www.geforce.com/hardware/desktop-gpus -
@notareal said:
GTX580 is not a bad at all. But as used architecture between kepler and fermi is rather different, cuda core to core comparison is difficult. There been talk that GTX6xxx series is inferior to GTX5xxx series, but this seems to be true only in some spesific instruction that Thea does not use. When some new CUDA features are used, GTX6xxx has some benefits.
It would be very interesting to see a speed comparison between a 580 and a 680...
Advertisement