Nvidia Titan

[vlado]

Vlado, should we create a new benchmark scene to have a new ground base for next setups/releases?

It will be a good benchmark for single GPUs for quite some time However it might be a good idea to post a separate benchmark for multiple GPUs. Or maybe post new scenes completely, just for a change

Best regards,
Vlado

[jpmartin]

I have compared an EVGA GTX 680 4gb Classified, overclocked with a slightly overclocked EVGA GTX Titan Superclocked. I used the test file referred to in this thread and ran the latest RT. Here are the results, at 1920x1152 sized image:
680 OpenCL = 1m 48.3s
680 CUDA = 0m 44.1s

Titan OpenCL = 1m 42.7s
Titan CUDA = 0m 29.9s

I am not an expert VRay user, by any means. I have appreciated all the help I have received in this forum, so I wanted to share my interesting results!
Also - the 680 GTX was tweaked for max performance, the Titan is not.

Thanks,
Brian

The results of this benchmark test are for a scene rendered at 1920x1152. I believe the actual benchmark scene is 800x480. Changing the output size of the render would make a huge difference in render times. Any chance you could run the Titan benchmark at the original 800x480 size so we can have a level comparison?

Many Thanks!
-Jesse

[stilgarna]

The results of this benchmark test are for a scene rendered at 1920x1152. I believe the actual benchmark scene is 800x480. Changing the output size of the render would make a huge difference in render times. Any chance you could run the Titan benchmark at the original 800x480 size so we can have a more level comparison?

Many Thanks!
-Jesse

+1

Could be a lot better...

[Bpositive]

No new tests!?

[tammo]

I didn't get the card,
It happened exactly as i feared.
The online shop has send me an email that something went wrong with their software,
and they don't have any cards, I have to wait until they get some, which might happen in a few weeks.

I wonder how many fantasy cards they sold,
and most will have paid them in advance with non-fantasy money.

I wrote them back and told them that we get a new 3d guy in two days.

Realizing that i now had a problem, they immediately took advantage.
One sales guy told me he just found a card that he could ship tomorrow,

But because of some extra work they had to do to get it,
I would have to pay another 12% of the cards value in advance...

[eyepiz]

Wow these Titans are so darn hard to find at retail price!...I somehow managed to get one ordered from EVGA that showed up in stock for like 1 min. I should have it tomorrow.

I'll post test on the bench scene once I get it up and running...I'm also super curios about viewport performance. The 680's classifieds do just slightly better than the Quadro 5000.

-Eric

[tammo]

I just found a Benchmark: GPU-Computing - CS Raytracing

It's on the lower third of the page:
http://www.computerbase.de/artikel/g...-gtx-titan/12/

Score
____________

Gtx Titan : 851
Gtx 680 : 495
Gtx 580 : 270
____________

[BKE]

No new tests!?

Hello all! I apologize for the delay in responding. I had to get a few projects out the door with deadlines. I see what I did wrong last time. I did not leave the scene on ActiveShade. I changed it to Render. I was consistent, though, in the settings for the 680 GTX and the Titan, so the comparison between the cards was accurate.

I EDITED THIS POST. See posting #70 below for actual render time.

[eyepiz]

Looks like your missing textures?

Hello all! I apologize for the delay in responding. I had to get a few projects out the door with deadlines. I see what I did wrong last time. I did not leave the scene on ActiveShade. I changed it to Render. I was consistent, though, in the settings for the 680 GTX and the Titan, so the comparison between the cards was accurate.
The CUDA time is 57.0 seconds. I am attaching the rendering with my settings for your reference. [ATTACH=CONFIG]13301[/ATTACH][ATTACH=CONFIG]13302[/ATTACH][ATTACH=CONFIG]13303[/ATTACH]

[BKE]

Looks like your missing textures?

Thank you. I found the link that includes the textures. I am re-posting the results: 1m 48.5s (Also 1m48.1s - I ran it twice.)

[vlado]

Thank you. I found the link that includes the textures. I am re-posting the results: 1m 48.5s (Also 1m48.1s - I ran it twice.)

That's very nice Only slightly slower than a 690, and that one's a dual card...

Best regards,
Vlado

[Bpositive]

I just tested my gtx 580 in cuda with the latest drivers and vray 2.4 so its directly comparable.

1 came in at 2m 41.2s and 2x at 1m 21.6s.


So thats 108 seconds versus 161 = 49% faster

Thats not so bad. Though one could have hoped for better, it can justify the investment.

[Alan Iglesias]

...So thats 108 seconds versus 161 = 49% faster

Thats not so bad. Though one could have hoped for better, it can justify the investment.

Thanks and very interesting...that puts it right about where we predicted speed-wise.

Yes, not a bad investment at all when you consider the 6GB VRAM. I think many will be surprised at the size of a file you can get in a seemingly small amount of GPU RAM compared to what Max reports physical RAM used in a rendering. Remember that only the scene rendering information is being sent the the GPU, and you are not holding the OS, Max file, etc. in RAM as you do in a normal render.

Two of these (roughly equalling 3 580s speed-wise) would make a very compelling real-time GPU rendering solution for a considerable amount of folks out there for just a couple of thousand dollars.

Anyone out there want to buy a couple of used 580s? HA!

We are gettting there...

-Alan

[Bpositive]

Anyone out there want to buy a couple of used 580s? HA!

Haha, you can have mine for a very good price too!

[Jiri.Matys]

So it was 1m 48s with GTX Titan? In a scene with textures? BKE?