OTOY Forums

smicha wrote:Tutor,

Thank you. My two daughters, my wife and me are very good. I talled them that there is a passionate guy, Tutor, full of positive energy on the forum. They were amazed as I am.


The vertical setup with extra fans at the bottom does great job. The temperature scores are so good, especially that you set core clocks so high on air. If you want to keep them clean there are custom demciflex filters.

Is it loud at 3122 RPM?

Smicha,

Thanks for the filter recommendation.

I have eight quad GPU SilverStone Raven 3 systems that I use as render slaves. Typically, I run about 3 systems per room. As set now (the same as was set for the test ran yesterday), that one system running alone is no louder than one of my three 2007 MacPro 2,1s (each with three GPUs) sounds if I'm only GPU rendering on it. With that setup, three systems running together in the same room does increase the noise level. Since I use a very aggressive GPU fan curve on all of my systems, my judgement on what is loud may be a bit tainted. A few years ago I made the decision to run my CPU and GPU fans at 90% as the temps approach about 50 degrees (C) to avoid thermal related failures/throttling since I typically overclock my systems' GPUs and CPUs to get the most performance from them. I sometimes run two render jobs on the same system, rendering with Octane on the GPUs and running the native (or another) renderer simultaneously on the CPU(s). In sum, I'd much rather replace GPU/CPU fans than to replace GPUs/CPUs. Luckily, I haven't had to replace any GPUs or CPUs or their fans for failure. I do, however, shut the room's door when rendering simultaneously on the GPUs and CPU(s).

P.S. Here's my typical fan curve design. I render for warmth in the Winter. Performance pic below is for 4xGTX 780 Ti ACX SC OCs (same as in my last pics) after I smoothed out the curve a little more =>
Directlighting
Pathterm=1
Coherent Ratio=1

~256.08 Ms/s

Here's how 8xGTX 780 TIs ACX SC OCs perform in one of my two Tyan Server systems. Four of the GPUs are ones that I tested most recently. These GPUs, as earlier, are air-cooled. Having 8 GTX 780 TIs in the same system does raise the temperatures. I used a less aggressive fan curve for the GPUs and a more aggressive one for my system. I also significantly reduced the overclock. Tweaking this Tyan system is still a work-in-progress.


Pathtracing
Pathterm=1
Coherent Ratio=1

~191.22 Ms/s


Directlighting
Pathterm=1
Coherent Ratio=1

~421.52 Ms/s

Great setup, still impressive overclock on air. Although temperatures should be measured after 5-10 minutes of rendering.

I am very new to GPU rendering and Octane. My result is

2 x GTX 980
~13.50 Ms/Sec

Is that good enough for non overclocked pair of 980s?

nacho_grande wrote:I am very new to GPU rendering and Octane. My result is

2 x GTX 980
~13.50 Ms/Sec

Is that good enough for non overclocked pair of 980s?

Please post render kernel settings at which you scored like that.

smicha wrote:

nacho_grande wrote:I am very new to GPU rendering and Octane. My result is

2 x GTX 980
~13.50 Ms/Sec

Is that good enough for non overclocked pair of 980s?

Please post render kernel settings at which you scored like that.

I hope the SS below helps.

980 renders close to 780 (not overclocked). So your results are pretty good.
If you can send results on 2.21, not 2.17.

smicha wrote:980 renders close to 780 (not overclocked). So your results are pretty good.
If you can send results on 2.21, not 2.17.

Yeah it would be interesting =)

smicha wrote:Great setup, still impressive overclock on air. Although temperatures should be measured after 5-10 minutes of rendering.

Here are temperatures and results for longer render times - achieved by increasing Max Samples. I lowered the overclock, but the extra work done by maxing samples raised the temperatures.

Temp = low 90s
Pathtracing
Pathterm=1
Coherent Ratio=1

~281.69 Ms/s

Temp = high 80s
Directlighting
Pathterm=1
Coherent Ratio=1

~462.86 Ms/s

Tutor wrote:

smicha wrote:Great setup, still impressive overclock on air. Although temperatures should be measured after 5-10 minutes of rendering.

Here are temperatures and results for longer render times. I lowered the clocks, but the extra work done raised the temperatures.

Temp = low 90s
Pathtracing
Pathterm=1
Coherent Ratio=1

~281.69 Ms/s

Temp = low 80s
Directlighting
Pathterm=1
Coherent Ratio=1

~462.86 Ms/s

...wow...