OCTANE BENCHMARK: Submit performance scores for your rig!

[PolderAnimation]

Here is our Octane-PC: a

intel i7 4770
5 x 780 TI @1160 mhz
16 GB ram
Asus Maximus VI extreme
And 2 meters of aluminium foil .

We searched for a nice solution for 5 videocards but couldn't find something that we wanted. So we decided to build our own case (photo attached). This case is made out of 2 old computer cases. We used risers to get the cards out of the case. When we where done building we tried out our new system,..... only problem was, the computer didn't work.

After a lot of testing we found out that the risers caused the problem. The risers we not good enough shielded.
The solution was do put aluminium foil around the risers. At the moment it seems that they are properly working.
So if you are going to use risers, buy good shielded ones =D.
Also we found out that max tdp doesn't mean max watt. So something this system peaks a 2200 watts of a very short time. This only happens during shading and test rendering. It doens't peak this high when rendering an animation (dunno why).

The cards that we are using are the overclocked 780Ti windforce (the cards run on 1160 mhz when rendering).
But even with this overclock the temps are between 40-55C (full load) depending on the card (idle 19-20C).

Version 1.2:
pathtracing 38.67 Ms/sec

[acc24ex]

whoa .. cool setup, good info on the shielding of risers - I'm using them with gtx 580s, my mainboard is the Asus P6T7 WS with 7 PCIe slots - and I had issues when connecting the 590 gtx card via riser (the system did not recognize)
- also - did you use those risers with additional power (or is it for more power hungry cards)

.. finally I just wanted to check which card is the fastest yet, is it the titan or 780ti? - and how does that still compare to 590gtx (I can still manage with 1.5 gb), the pricing on the new cards are ridiculous as always

.. really hating the fact nvidia keeps on naming products just to confuse us, titan, and 780ti (implying there is some "titaning" going on)
- also reading about the number of CUDA cores is now total bullshite - since 590gtx is almost the same speed as titan which has twice the number of cores .. real bullshit, now I just wish we had another option because nvidia suck a bit with their "how can we false advertise this" policy
- I mean CUDA cores are basically for GPGPU apps .. and they don't mean anything now to the people buying them for that, it's a niche, which they are trying to confuse in the same manner as they do with gamers..

[kavorka]

Hey guys, I am looking to upgrade my work computers.
Currently we have a couple 580's and 690's in our comps.

On 1.52, PT, we get:

GTX 580: 3.97
GTX 690: 7.15


I am trying to figure out the speed of of the 780 and 780ti.
SO far, i see one person who is getting the 780ti at 8.4 and another at 7.7. I'm guessing they are different cards or different versions of Octane.
I'm wondering if anyone has the EVGA SC 780ti and can test it with 1.52?
I found them on Amazon for $709. If I can get it more towards the 8.4, might pick up a couple, even sell our 690s and 580s since we will get more VRAM.

Even then, speed upgrade isnt even that much over the 690. I'm wondering if we should just wait.

[voon]

The 600 and 700 Series is comparable with regards to speed per core at a given frequency. They're all Kerpler cores. Card makers however provide various degrees of overclocking. That can be the reason you see speed differences, given the same driver and octane version etc. of course, if you start varriing everything, you create a mess and can't deduct anything

The earlier 500 series cards are Fermi cores. They use far less cores, but more powerful ones, hence a 780 isn't that different from a 580. Sadly, Nvidia didn't really advance speeds there^and we're not much farther than we were with 580s.

In gaming of course we talk about a huge difference on the GTX cards ... but not in CUDA calculations.

[acc24ex]

- so far, the speed increase from 590gtx to 780ti or titan is like a few percent
- if you can live with less memory I really do not see the point of upgrading now.. more powerful maxwells are probably going to be released near the end of this year, and octane team is yet to discover if it's the real deal or not

- when you think about it, 580 and 590 gtx brought the revolution in GPGPU, for the past few years there was almost no significant push in technology, according to Moore's law, every 2 years the technology doubles - this has obviously changed - since gaming performance was only pushed - but still gamers should be aware that with brigade they can receive far superior graphics - and for developers it should be as easy as using octane.. so what is really upgraded for gamers? I guess trying to hype and push the almost same technology around for 2-3 years as new is great for managers, not that much for users.. who are really actually monopolised right now, with basically having only cuda..

- how can you explain a CUDA core having more "power" for 580 than for 780? I am just assuming its false advertising.. god damn it nvidia!

[haknslash]

.. finally I just wanted to check which card is the fastest yet, is it the titan or 780ti? - and how does that still compare to 590gtx (I can still manage with 1.5 gb), the pricing on the new cards are ridiculous as always

.. really hating the fact nvidia keeps on naming products just to confuse us, titan, and 780ti (implying there is some "titaning" going on)
- also reading about the number of CUDA cores is now total bullshite - since 590gtx is almost the same speed as titan which has twice the number of cores .. real bullshit, now I just wish we had another option because nvidia suck a bit with their "how can we false advertise this" policy
- I mean CUDA cores are basically for GPGPU apps .. and they don't mean anything now to the people buying them for that, it's a niche, which they are trying to confuse in the same manner as they do with gamers..

A 780Ti is going to be faster than a Titan. A quick useful website for comparing performance of cards against another is hwcompare. There is "Titaning" lol if you think about it because the 780, 780Ti, Titan and Titan Black are all variants of one another where you trade off performance for VRAM and visa-versa, ultimately with the Titan Black being the most combined of them all having the perfromance of the 780Ti with the VRAM of the Titan. Hope that clears it up for you.

There is nothing "BS" about the CUDA cores between a Fermi card and a Kepler card. They are two different architectures so 1 Fermi core represent roughly 2.5/3 Kepler's cores. That's why the number throws people off as they think "X" core = "X" core when that's not really the case. That's like thinking all tires (Fermi, Kepler, Maxwell) are the same just because they are made of rubber (CUDA) material . Think of Kepler as being the more efficient of the two when it comes to heat and power, so you have to trade off speed. Can't have your cake and ice cream too

[ASyme1]

Hi guys-
I just re-tested my "uber rig" I have here at work. This workstation has seven Titans using a Netstor breakout chassis. I've upgraded four of those Titans to Black Editions. Great cards if you can find them. Super hard to find now but I'm sure that'll change. This rig is doing a 61 in Pathtracing and a 45 in PMC. DL is over 200 and I can't get a screen grab because it's over before I can catch it! Funny. This machine can RENDER.

Alec

[BorisGoreta]

Great results ! Could you please measure how much faster is single TITAN black compared to TITAN ?

[ASyme1]

Great results ! Could you please measure how much faster is single TITAN black compared to TITAN ?

Sure! Good idea. Just tested. With one normal Titan in Pathtracing I'm getting 7.8 or so. With the BLACK Titan I'm getting 9.5 in Pathtracing. That's more than I thought It would do. Isn't that like 15-20% faster? I'm impressed. I wonder if that's the same score as a 780ti? Anyone want to run a single 780ti to find out? Curious.

Alec

[BorisGoreta]

I'm getting 6.4 on a single TITAN in Octane 1.35. How come ? Are your TITANs overclocked ?

Actually it drops to 5.5 after it warms up.