NVIDIA GTX EVGA 780 ti SC vs EVGA 980 SC tested

Generic forum to discuss Octane Render, post ideas and suggest improvements.
Forum rules
Please add your OS and Hardware Configuration in your signature, it makes it easier for us to help you analyze problems. Example: Win 7 64 | Geforce GTX680 | i7 3770 | 16GB
User avatar
ralf_breninek
Licensed Customer
Posts: 88
Joined: Fri Mar 22, 2013 8:33 pm
Location: Germany Cologne / Köln
Contact:

Hello everybody,

First off I am a graphics designer not a hardware specialist, therefore my tests are not professional.

But sofar the 980 seems to be 15-25% slower which is a big bummer :( This means one minute longer computing per frame on my test scene which is a huge dissappointment. Although several people have pointed out that this might be a drivers problem either on NVIDIA´s, EVGA´s or Octane´s side. I suppose the next weeks / months will show the true capabilites of the 980 GTX.

Setup:
I ran the test on my 780ti superclocked first, then deinstalled all the drivers and the card itself, replaced it with the 980 GTX superclocked and installed the drivers again. So I ran every card single, that way it had the full x16 PCI express bandwith. Both cards ran on stock settings.
780ti 1006 MHZ with boost up to 1162MHZ
980 1239 MHZ with boost up to 1352
NVIDIA driver 344.11 (released on 19.09.2014)

I rendered a 5 frame sequence to avoid VRAM reloading issues.

FRAME__TIME 780ti__TIME 980
325_____4:06_____05:31
326_____3:45_____05:06
327_____3:45_____05:07
328_____3:42_____05:05
329_____3:44_____05:05


Conclusion:
One minute per frame faster rendering on my "old" 780ti is a critical number. Also because I did not run into VRAM issues yet and I plan to watercool it anyways I am going to stay with my 780ti.

The lower power consumption is not that big of an issue, because a 100 watt difference would only mean a 100€ higher power bill per year on 10 hours a day for 356 days which is unlikely to happen so its more like 50-100€ difference. (My power company takes 26,72 per kWh)

Hopes:
I hope NVIDIA will make GPU computing more feasible again for freelancers like me. I am looking into a tripple / quad EVGA 780ti SC setup again now.
Last edited by ralf_breninek on Mon Sep 22, 2014 11:19 am, edited 2 times in total.
C4D R21.207 / Octane for C4D 2020.2.4 / Watercooled 4 x EVGA Geforce RTX 2080 ti
http://www.pixelasm.com
Interface and motion graphics design for interactive interfaces and animations on TV, web, mobile and automotive from Köln / Cologne Germany
User avatar
bepeg4d
Octane Guru
Posts: 10345
Joined: Wed Jun 02, 2010 6:02 am
Location: Italy
Contact:

hi,
thanks for sharing your tests so far :)
which driver version have you used for the test?
the blender cycle test posted some days ago was promising, i strongly hope that you can get faster results with newer drivers, finger crossed ;)
User avatar
ralf_breninek
Licensed Customer
Posts: 88
Joined: Fri Mar 22, 2013 8:33 pm
Location: Germany Cologne / Köln
Contact:

I updated my initial post because driver Version is a good point. Thanks bepeg4d :)

I downloaded a fresh version of the NVIDIA "suite" and thought it would contain the newest version already. Just checked on that and realized it does not :( So I updated them to 344.11 (released on 19.09.2014)

Unfortunately the default boost clock was decreased by 12MHZ to 1354MHZ. Because I wanted everything to stay in an "out of the box" condition I left it that way. I guess that explains the 1-2 seconds longer rendertimes compared to my first post.

Why would a company not update their software with the newest drivers? The way it is right now users have to download about an extra 100mb. They could shave off Giga or even Terrabytes of their traffic just by updating their software bundles as soon as new drivers are released. Although I got to admit that the updating process is a lot easier compared to several years ago when you had to look up the newest driver version, download and install it yourself...
C4D R21.207 / Octane for C4D 2020.2.4 / Watercooled 4 x EVGA Geforce RTX 2080 ti
http://www.pixelasm.com
Interface and motion graphics design for interactive interfaces and animations on TV, web, mobile and automotive from Köln / Cologne Germany
User avatar
splittmatt
Licensed Customer
Posts: 42
Joined: Tue Dec 10, 2013 4:51 pm
Location: Gold Coast, QLD Australia.

Thanks for the data.

If you get a chance maybe you could read this post, bench test and fill in the spreadsheet? It only takes a minute but it's good to have all the comparisons in one place.

http://render.otoy.com/forum/viewtopic.php?f=9&t=7144
Win 11. Intel I-9 14900K. 64Gb. RTX 4080 Super.
User avatar
glimpse
Licensed Customer
Posts: 3740
Joined: Wed Jan 26, 2011 2:17 pm
Contact:

thanks for info! =) hope something could be done with optimisation on a code side =)

pluss really curious how these cards would run a little over-clocked =)..but we'll see =)
User avatar
smicha
Licensed Customer
Posts: 3151
Joined: Wed Sep 21, 2011 4:13 pm
Location: Warsaw, Poland

Guys,

I just got 3x780 6GB asus strix and I will put them on water together with my titan on xxl ekwb waterblock.

We've tested 980 on overclock.net and it is 3.5-4ms/sec vs 6.5-7.8 for titan or 780. I am so happy I got last 3x 780 6gb in Poland. These are so hard to find today.


PS. 340eur net/780 6GB :)
3090, Titan, Quadro, Xeon Scalable Supermicro, 768GB RAM; Sketchup Pro, Classical Architecture.
Custom alloy powder coated laser cut cases, Autodesk metal-sheet 3D modelling.
build-log http://render.otoy.com/forum/viewtopic.php?f=9&t=42540
User avatar
glimpse
Licensed Customer
Posts: 3740
Joined: Wed Jan 26, 2011 2:17 pm
Contact:

smicha wrote:Guys,

I just got 3x780 6GB asus strix and I will put them on water together with my titan on xxl ekwb waterblock.

We've tested 980 on overclock.net and it is 3.5-4ms/sec vs 6.5-7.8 for titan or 780. I am so happy I got last 3x 780 6gb in Poland. These are so hard to find today.


PS. 340eur net/780 6GB :)
Nice! =) I was looking to grab STRIX 780 6Gb, but at the last minute seen lightly used TitanBlack (still with warranty) for the same price & with reference cooler! =) picked that beast =) will be way easier to fit water block (as some users on this forum already noted that there's no wb for strix, EK confirmed =)
User avatar
smicha
Licensed Customer
Posts: 3151
Joined: Wed Sep 21, 2011 4:13 pm
Location: Warsaw, Poland

Glimpse,
there is one, actually three different ones :) Just ordered 3x :)

http://www.coolingconfigurator.com/step ... _gpus=1452


I'll post some photos from the build log soon.

PS. Great you got the black!
3090, Titan, Quadro, Xeon Scalable Supermicro, 768GB RAM; Sketchup Pro, Classical Architecture.
Custom alloy powder coated laser cut cases, Autodesk metal-sheet 3D modelling.
build-log http://render.otoy.com/forum/viewtopic.php?f=9&t=42540
User avatar
glimpse
Licensed Customer
Posts: 3740
Joined: Wed Jan 26, 2011 2:17 pm
Contact:

It's just a visual match - that doesn't keen it will fit 100%

quite a while ago I've seen a video from extreme rigs where they tested different blocks for GPUs (titans/780). The surprising thing is that GPU chip itself isn't always the most heated part under load - that comes as a warning, that good block (with perfect fit) is simply a must if You going to OC, especially with flashed bioses etc..- but I guess You know all that already =)

jop & thinking about 280 on that Black (as I have small neat box - fits perfectly under the front grill =)..
User avatar
smicha
Licensed Customer
Posts: 3151
Joined: Wed Sep 21, 2011 4:13 pm
Location: Warsaw, Poland

Glimpse,

Two things:

1. Thanks for some feedback on waterblocks for 780. I am looking for some more precise info about problems users encountered with strix and watercooling - if you find some useful links please post them.

This is the best info I found so far.
http://www.overclock.net/t/1502942/toas ... 80-oc-6gd5

2. If you can, go with 280 (2x140 I assume) or even more. My 2x560 are outside the case and until I get some money for case-labs STH10 I don't care how these look:) Measure the max distance you can mount the radiator and get yourself the thickest one in push pull with low rpm fans - be-quiet are just fantastic. Will you have room for pump and reservoir? Don't you want to cool your cpu at once?
3090, Titan, Quadro, Xeon Scalable Supermicro, 768GB RAM; Sketchup Pro, Classical Architecture.
Custom alloy powder coated laser cut cases, Autodesk metal-sheet 3D modelling.
build-log http://render.otoy.com/forum/viewtopic.php?f=9&t=42540
Post Reply

Return to “General Discussion”