Which GTX cards to use for IRAY rendering workstation? 780ti or 980

T

ty1er

Limp Gawd
Joined
Mar 13, 2007
Messages
129
I'm in the process of building a new workstation for mainly running 3ds max and creating images with its IRAY render. Since IRAY can be processed on both the CPU and benefits greatly on a fast CUDA equipped GPU. This leads to my final puzzle piece of my PC build....the graphics card(s).

Which GPU will have the best performance when rendering with IRAY?

I'm either going to go with a three or four way setup using the 780ti or the newer Maxwell based GTX 980 cards. I'm having a tough time deciding between the two. The benefit of the newer 980 GPUs is that they are more efficient and have a 1gb more memory then the older 780ti's. But if the older 780ti are a large percentage quicker, I'll go that route.

Over at Maxforums forum there's a IRAY benchmarking thread but hasn't been updated with a lot of recent hardware.

http://www.maxforums.org/threads/iray_gpu_cpu_comparison_test/0001.aspx
http://dl.dropbox.com/u/23316655/iray_test_just_hit_render.zip

If anyone has a GTX 980/970 and runs 3ds max 2015 (with the latest update that supports Maxwell gpus). It will be interesting to see yours results.

http://www.nvidia-arc.com/products/mentalray/maxwell-gpu-support.html


Please post separate results of CPU, GPU(s), and combined CPU & GPU(s) times.


danke
-ty1er/
 
Last edited:
im for sure no expert on the subject...but there's no way i would invest in the older generation of multiple cards when the newer ones are pretty much the same price or am i wrong? hard to imagine an older generation doing better than a newer gen of cards but im sure more people will chime in.:)
 
In some cases, non-game related rendering applications (like IRAY, I would expect) need more processing power from the parts of a rendering pipeline that are less used in games. While the 980 is faster than a 780ti in pure gaming applications, in some non-gaming rendering and FP performance aspects it does not have that same level of advantage over the 980.
I did find this:
http://www.migenius.com/products/nvidia-iray/iray-benchmarks-2014-5

It shows a 780ti superclock being faster than a 980, but the advantage is not huge in many benchmarks, but in the Global Illumination benchmark - wow, that's a big difference. Tough call. I would look into used 780tis, maybe, if they are price competitive.
 
sethk said:
In some cases, non-game related rendering applications (like IRAY, I would expect) need more processing power from the parts of a rendering pipeline that are less used in games. While the 980 is faster than a 780ti in pure gaming applications, in some non-gaming rendering and FP performance aspects it does not have that same level of advantage over the 980.
I did find this:
http://www.migenius.com/products/nvidia-iray/iray-benchmarks-2014-5

It shows a 780ti superclock being faster than a 980, but the advantage is not huge in many benchmarks, but in the Global Illumination benchmark - wow, that's a big difference. Tough call. I would look into used 780tis, maybe, if they are price competitive.
Click to expand...

Morning sethk,

I feel the same way after looking at the benchmarks over at Migenius. Generally the 780ti SC is faster according to those benchmarks. In there tests they are listing a non SC "overclocked" 980 card(?) I'm wondering how well a gtx980 SC fairs against the 780ti.

The only other information I have found between the two series cards in regards to gpu rendering was these two threads below. But they are comparing other GPU rendering engines besides Iray.
This one is a discussion about the Octane Render
http://www.overclock.net/t/1514149/octane-render-benchmark-for-970-980-please-help

This one is for VRAY-RT. Comparing a Titan Black against the 980/970 GTX. From what I gather, the Titan Black is very similar to the 780ti, except it sports 6GB instead of only 3gb. O' and costs more then twice as much!.
http://blog.boxxtech.com/2014/11/17/geforce-gtx-rendering-benchmarks-and-comparisons/

Both favor the older cards.

What to do.... what to do....
 
Last edited:
ty1er said:
Morning sethk,

I feel the same way after looking at the benchmarks over at Migenius. Generally the 780ti SC is faster according to those benchmarks. In there tests they are listing a non SC "overclocked" 980 card(?) I'm wondering how well a gtx980 SC fairs against the 780ti.

The only other information I have found between the two series cards in regards to gpu rendering was these two threads below. But they are comparing other GPU rendering engines besides Iray.
This one is a discussion about the Octane Render
http://www.overclock.net/t/1514149/octane-render-benchmark-for-970-980-please-help

This one is for VRAY-RT. Comparing a Titan Black against the 980/970 GTX. From what I gather, the Titan Black is very similar to the 780ti, except it sports 6GB instead of only 3gb. O' and costs more then twice as much!.
http://blog.boxxtech.com/2014/11/17/geforce-gtx-rendering-benchmarks-and-comparisons/

Both favor the older cards.

What to do.... what to do....
Click to expand...
The TITAN also has full double precision for floating point numbers, which could be worth the extra money in those rendering programs. If you're going to go 3- or 4-way SLI with GeForce for a rendering workstation, it might be worth looking into a Quadro K5200 for that much money.
 
Also woth noting, if running in 3 or 4 way sli, you're going to have thermal concerns to deal with, especially with the 700 and titan series cards. I've built a few 3 way sli systems with 780's/780ti's/titans/and titan blacks, and all of them needed custom fan profiles to prevent the cards from throttling (they would often have to run the fan at 100% at stock speeds).

Given they 980's lower tdp, power and heat should be easier to manage, allowing a full speeds for longer times (and also less noise).

However, if you're planning on water cooling or using a cryptocoin mining open frame chasis, this may not be an issue.
 
I ended up opting for the 780ti SC cards. Might be a little while before I get the system up an running. When I do i'll post my results.
 
Test file
http://dl.dropbox.com/u/23316655/iray_test_just_hit_render.zip
Just load file and hit render.

3dsMax 2015, and Nvidia driver 347.09

All stock clocks

CPU Only 5960X: 45 secs

1x GTX 780ti SC KP (no cpu): 20 secs
2x GTX 780ti SC KP (no cpu): 11 secs (82% faster then a single card)
3x GTX 780ti SC KP (no cpu): 07 secs (186% faster then a single card - 57% faster then dual cards
4x GTX 780ti SC KP (no cpu): 06 secs (233% faster then a single card - 17% faster then three cards)

CPU & ALL GPUs: 5 secs (800% faster then just the CPU alone)
_______________________________________
 
Last edited:
I cant help but wonder what your test result would have been with cpu and only 1 or 2 card's....I mean looking at your own data from a cost point of view the extra cards don't do a whole lot
 
primetime said:
I cant help but wonder what your test result would have been with cpu and only 1 or 2 card's....I mean looking at your own data from a cost point of view the extra cards don't do a whole lot
Click to expand...

This was more of a test to demonstrate the performance/diminish and return of multiple GPUs.

I agree, the cost of the fourth card outweighs the performance. But in most situations I would be only using three cards max. The fourth would be dedicated for the OS. This way, you are still be able to use the machine while rendering.
 
Last edited:
You must log in or register to reply here.
Back
Top