Has anyone done, or can anyone point me to any information comparing GTX gpu’s and the Quadro K series cards with Blender? I’m thinking of investing in a high end design and/or rendering machine. I think, from the blurb I can find I’d be looking at the K5000 series.
I understand the upper end Quadro cards will handle the design environment well, but will they also render cycles quickly and efficiently?
I have to work out how heavily I need to re-mortgage before I decide if I dedicate one machine to design with a high end card and another for background render.
The Nvidia help line people were very helpful, they say comparing render times between Geforce and Quadro cards is like comparing Disk drives with Operating systems. I’m not quite sure how they come to this!
I guess if I had the money and the motherboard real estate I would use a K series for running the monitors for design, and have the 3 GTX780’s sat there doing nothing until rendering!
A quick thought on SLI. I know it would make no difference at all to rendering, but is the Blender Application written at all with SLI in mind? Would SLI improve the design environment?
Any ideas on this would be very much appreciated. I’m really struggling with what I would call pretty low poly scenes. Yes I’ve been through the wiki on speeding it up, and I have all the subs turned right down. etc. etc…
Hi! I was loocking for an older post of mine and saw your thread.
I have a GTX 750ti and recently I bought a quadro k620 (600 but the newer model)
I can confirm that the quadro runs noticeably quiter and is more power efficient in terms of componentry than the regular geforce.
the k620 has a max power draw of 45 W so it doesn’t require any extra cables while the 750ti needs one 6pin extra.
the geforce is almost twice as fast in CUDA compute with Cycles due to the number of cuda cores and the frequency.
I didn’t buy the k620 for Blender alone nor did I do it for CUDA compute( almost useless but still faster than my i73820 stock)
Quadro is primarily for OpenGL display tasks. Blender doesn’t get any special treatment like 3ds max does with quadro in terms of OpenGL optimizations but I can definately say the quadro is more responsive and fluent in Blender than a regular geforce.
If you strictly need Single FP32 precision with CUDA compute in raytracing then you’re best bet are one or more high end GAMING geforce coupled with one disply only card (keep this one out of the compute area)
You also need a little FP64 then GTX TITAN BLACK or better TESLA for compute only
My only gripe with GPU rendering is that I cannot set how many CUDA cores are left free for a dispplay card like the quadro.
That way if I had lets say one quadro k620 with 384 cuda cores + one gtx 980 for compute only I could activate the full 2048 cores on the GTX and have only half of the quadro helping out thus keeping a stutter free display.
This becomes apparent when you have a high end quadro model such as the k4200 and gtx 980 both with 4gb Vram and you wouldn’t want to have the cores on the quadro idling but helping out.
thanks. so win7 recognise 750 without issues then? i read somewhere that there has to be a dummy connection to the gpu in order for os to recognise it… but it seems it what rubbish afterall
I’ve recently had a chance to test an AMD firepro w2100 card and I can say that the k620 is both faster and much quiter (you can barely hear it in idle mode) while the AMD is much noiser even when idle.
In load it sounds like a hairdryer and it has some problems with 10 mil + triangles not showing up in the viewport at all. Be glad you’ve got a nVidia card. It’s also much more energy efficient in idle than AMD.
