4
$\begingroup$

I recently added a new card to my system. I am running a Titan X and 770. When testing my render speed, I found that, for the same scene and same samples, the 770 took 1:53 to render it, and the Titan took 1:20. However, with both working together, it took 1:16...only a 4 second improvement over the Titan alone.

I know that you don't get the full power of the second card, but this seems a little extreme. In Blender's options where I select the compute device, it lists 770+Titan, with the 770 first (The titan is in the primary motherboard slot and runs 1 of the 3 displays.) Does this mean it's primarying the 770 and then not making full use of the Titan? How do I fix this?

EDIT: When watching my GPU monitor, I can see the 770 being used, but the Titan rarely shows activity above a couple %. It seems like the 770 is getting all the tiles or something. What's with that?

EDIT 2: When testing viewport rendering, things actually slow down when using both cards vs using just the Titan.

$\endgroup$
  • $\begingroup$ how big are the tiles you are rendering? Are the monitor(s) hooked up to any of the cards? How much ram do the cards have? $\endgroup$ – cegaton Aug 16 '15 at 1:02
  • $\begingroup$ @cegaton I've tried a variety of tile size, from x128 to x1024 using Autotilesize addon. While they change render speed slightly, it doesn't make a huge difference. x512 seems to be optimal for both cards from testing them on their own. My primary screen is on the titan, both secondary screens on the 770. The 770 has 4gigs VRAM, the titan has 12. $\endgroup$ – Drudge Aug 16 '15 at 1:17
  • $\begingroup$ When using different GPUs memory use during render is limited to the smallest of them, In this case you can only use 4GB, so a lot of the Titan's RAM is not going to be used no matter what. If on top of that the 770 is driving two monitors you have even less resources available for rendering. Do a test using only one monitor hooked up to the Titan and see if that makes a difference. $\endgroup$ – cegaton Aug 16 '15 at 1:33
  • $\begingroup$ It seems to me that the most efficient way to use your system is to have the lesser one of the cards deal with the displays only and use the most powerful one for rendering... $\endgroup$ – cegaton Aug 16 '15 at 1:36
  • 1
    $\begingroup$ It's hard to say what is wrong with your setup. Just the fact Titan takes 1:20 and much cheaper 770 takes 1:53 seems wrong (also are we talking about bmw benchmark scene?). On developer.blender.org there were couple problems and bugs with Titan drivers before, you might want to check that if that's resolved. If your cards are not in SLI and one of them is below 99% while rendering your best bet is to file a report and contact devs. $\endgroup$ – Jaroslav Jerryno Novotny Aug 18 '15 at 5:48
6
+50
$\begingroup$

So I just tested this:

No, card order does not make a difference. Doesn't really matter which one is set as system (if the system is doing nothing) or to which the monitors are plugged. The render times were off by fractions of a second (BMW benchmark scene with 2 cars).

Both my cards are in PCI Express 3.0 at x16 slots.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.