So since the announcement of these news cards by both the green and red team I am very interested to see how the VRAY RT rending is compared to the TITAN X and 980ti.
Any updates or info from the devs or people that have access to these benchmarks would be greatly appreciated. I trying to figure about about my next purchase regarding GPUS.
Great! I can hardly wait! Keep us GPU nerds posted.
Also…I have read that AMD and Nvidia are quickly coming out with new Dev tools to unlock more features that the GPU can do. Has there been any recent breakthroughs from the GPU side of things that have seen appealing to adopt for Vray in future updates?
Dev tools - not really, but there seem to be quite a few possibly (very) interesting hardware capabilities, like compute preemption, shared virtual memory, nvlink, native 16 bit support, hbm memory …
We will know if they work once we test them.
I’m genuinely surprised you haven’t gotten access to alteast one GTX 1080 yet being a very popular company and all. Bencharks will start to be announced on May 17th. There is a NDA on it right now so I guess I answered half my question regarding vray benchmarks.
<shrug> We generally have a good relationship with nVidia, and we usually do get new hardware in advance, but we didn’t this time even though we asked several times. So we’ll know more when we actually buy them here.
I finally got my hands on a Titan X the other day (from a Quadro K5200) and lighting cars with Vray Lights/Softbox Maps in RT and being able to look through the lights in the viewport blows VRED and my K5200 out of the water. My boss asked me the other week if it would be better to invest in CPU or GPU’s for Vray power. Originally I thought CPU’s would be better but now I’ve seen what a Titan can do (albeit sometimes navigation is painfully slow - if anyone has any idea how to fix this let me know) I definitely want to explore the prospect of getting two 1080’s - if they are as much improved as the Titan’s (for half the price!) as the benchmarks are showing.
Someone over at Octane forums posted this small benchmark with Arion, seems pretty promising around 25% increase speed, specially since there’s no CUDA optimizations yet, and only recently CUDA 8.0 RC has been out.
Our very preliminary results show about 25% faster renders compared to a GTX 980 and 3% faster to a Titan X on average (some scenes were actually faster on the Titan X though).
Blago still has to run some tests with CUDA 8 to see if there is any difference and we will post the more detailed benchmarks.
Here is the V-Ray RT GPU CUDA test.
As Vlado noted, this is a bit preliminary testing, but since the GTX Pascal GPUs are similar to Maxwell ones, I don’t expect any major changes. In denoising, the 1080 did even better, outperforming the Titan X with a bigger margin. Having monitor attached to the 1080 decreased the performance with up to 10%.
Btw you don’t need special build to use a Pascal GPU.
Every V-Ray 3.30 (or newer) version works well with the 1080.
Currently I have a K5200 quadro but it’s hella slow compared to the Titan X when using RT. Is it possible to use both my K5200 and the 1080 in unison whilst rendering with RT production and RT active shade?
Also, can someone more tech savy explain why the 1080 is faster than the Titan and 980 despite having the least CUDA cores?