OTOY Forums

Have an ASUS Z9PE-D8 WS and previously had 4 x 780 Ti GPUs on it, just upgraded to 3 x 1080 SC GPUs and before swapping them out, I made a test scene for myself... on the 780 Ti's, it rendered in 10:00 min... with the three 1080 SC's, it rendered in 10:04. Even tried keeping one of the 780 Ti's in there and the results with three 1080 SC's and one 780 Ti were exactly the same as just three 1080 SC's.

Needless to say, I'm not impressed with the results. Some machine specifics:

[*] ASUS X9PE-D8 WS motherboard
[*] All PCI-E slots set to 3.0 in the BIOS
[*] NVIDIA drivers up to date (did a clean install after installing the new GPUs)
[*] C4D R18 up to date
[*] Running the latest 3.x.x C4D Octane plugin
[*] Running the latest 3.x.x Octane standalone
[*] Windows 10 Pro up to date

If there are any specs I left out that would be helpful to anyone, please let me know. Are the 1080's just not optimized yet for Octane? Trying to figure out why my render speeds have not improved.

Thanks

hi, Edensoto,

Here is a thing, & I'll try to explain what happened =) the reason OctaneBench has different scenes.. is because You can not judge cards performance based on one specific kernel or one specific scene. As far as I understand cards from different architectures behave differently..on some scenes the difference is bigger on some smaller or even non existent.

now consider this, that on v2.17 based OctaneBench Your 780Ti would be a bit faster , compared to v3 OctaneBench, so the question what benchmark You've used to test both configurations (aparently Pascal based GPUs do not run on v2.17). What I meant to say is that if You tested 780Tis on v2 while 1080s on v3 & got the same speed, that's not to say these results are the same =)

in order to have full picture, You need to test Your PC with different scenes & only then You can compare total performance. If You were upgrading from 3GB card to 8GB cards You gained a lot more vram, & a bit of efficiency. Last thing to do would be to install MSI Afterburner, tweak fan curves & make offset on clock speeds with 1080s You can get extra performance for free.

Another thing, depending on kernel You use, You can tweak few settings there to get additional boost in rendering speed. So there are things to be done if You want to make Your setup render faster. If You want some additional help, please run OctaneBench, open four instances of GPUZ with sensor tam opened & different GPU selected in the bottom. Make print screen under full lead & we'll see whether Your computer utilize GPUs to the maximum.

glimpse wrote:hi, Edensoto,

Here is a thing, & I'll try to explain what happened =) the reason OctaneBench has different scenes.. is because You can not judge cards performance based on one specific kernel or one specific scene. As far as I understand cards from different architectures behave differently..on some scenes the difference is bigger on some smaller or even non existent.

now consider this, that on v2.17 based OctaneBench Your 780Ti would be a bit faster , compared to v3 OctaneBench, so the question what benchmark You've used to test both configurations (aparently Pascal based GPUs do not run on v2.17). What I meant to say is that if You tested 780Tis on v2 while 1080s on v3 & got the same speed, that's not to say these results are the same =)

in order to have full picture, You need to test Your PC with different scenes & only then You can compare total performance. If You were upgrading from 3GB card to 8GB cards You gained a lot more vram, & a bit of efficiency. Last thing to do would be to install MSI Afterburner, tweak fan curves & make offset on clock speeds with 1080s You can get extra performance for free.

Another thing, depending on kernel You use, You can tweak few settings there to get additional boost in rendering speed. So there are things to be done if You want to make Your setup render faster. If You want some additional help, please run OctaneBench, open four instances of GPUZ with sensor tam opened & different GPU selected in the bottom. Make print screen under full lead & we'll see whether Your computer utilize GPUs to the maximum.

Thanks, glimpse, for that info. I guess where I'm at is... all the technical stuff aside, I just assumed 3 x 1080 Superclock's would be faster than 4 x 780 Ti's... I'm not really concerned with benchmarks and all that as I am in "will these cards render my project that I'm working on faster?" and the answer so far has been: No.

Okay, this is strange, as well... all three GPUs are maxing out at only 80% usage... I have them all set to use the max 120% power target. How do I increase that to get them all to 100% usage?

to give You propper reply. based n numbers & speeds..

gk110 based cards like 780Tis or Titan Blacks could score around ~110 in OctaneBench
1080s for now score ~140 & in some cases 160.. so 3x 140+ is around the same as 4x 110..

Consider the fact You're comparing High end card from one generation to mid end from other generation..
Again the same mistake as Guys made comparing 580s (hi end) to 680 (mid end)

High end from Pascal line is TitanX & probably upcoming 1080Ti. TitanX (Pascal scores over 160 out of the box for now).
So the results are as expected, but You expectations were a bit too high based on misinformation
(maybe You looked at gaming performance comparisons or so..)

if You're running Path Tracing kernel, bump PArallel samples & MAx tile samples to maximum. GPU usage will not change too much but on some scenes You can cut render time by 10-15%

glimpse wrote:to give You propper reply. based n numbers & speeds..

gk110 based cards like 780Tis or Titan Blacks could score around ~110 in OctaneBench
1080s for now score ~140 & in some cases 160.. so 3x 140+ is around the same as 4x 110..

Consider the fact You're comparing High end card from one generation to mid end from other generation..
Again the same mistake as Guys made comparing 580s (hi end) to 680 (mid end)

High end from Pascal line is TitanX & probably upcoming 1080Ti. TitanX (Pascal scores over 160 out of the box for now).
So the results are as expected, but You expectations were a bit too high based on misinformation
(maybe You looked at gaming performance comparisons or so..)

No... no other expectations other than "newer + faster = better"... I just relate it to the rest of the computing world components. Still can't figure out why the cards are topping out at 80% usage, though.

edensoto wrote:... Still can't figure out why the cards are topping out at 80% usage, though.

try unchecking "low priority" in your devices settings in c4d.

mesut wrote:

edensoto wrote:... Still can't figure out why the cards are topping out at 80% usage, though.

try unchecking "low priority" in your devices settings in c4d.

None of them have priorities set (low or high)... none of them are checked

edensoto wrote:

mesut wrote:

edensoto wrote:... Still can't figure out why the cards are topping out at 80% usage, though.

try unchecking "low priority" in your devices settings in c4d.

None of them have priorities set (low or high)... none of them are checked

It's normal behavior on v3 builds.. - even Your older 780Ti would not run as high as they were running on v2.