I unistalled the 258.96 driver and used Driver Sweeper to remove all traces of the driver. I then installed 197.45.
I then ran the Benchmark and my Room renders as I did with the 258.96 drivers and have the same results with the 197.45. Adding the second card does gives the same M/s as one card. GPU-Z does show that both cards are almost fully utilized.
xsipower
Two cards the same speed at one !!
Forum rules
For new users: this forum is moderated. Your first post will appear only after it has been reviewed by a moderator, so it will not show up immediately.
This is necessary to avoid this forum being flooded by spam.
For new users: this forum is moderated. Your first post will appear only after it has been reviewed by a moderator, so it will not show up immediately.
This is necessary to avoid this forum being flooded by spam.
Hey,
I'm sorry but i'm out of answers... there's obviously something wrong on the driver/OS/HW level otherwise you would'nt get equal or lower speeds...
We've other customers with dual GTX280 cards with positive results.
We even have a machine at the office that has 2x GTX200 series and it runs at approx 1.8x the speed of one.
Radiance
I'm sorry but i'm out of answers... there's obviously something wrong on the driver/OS/HW level otherwise you would'nt get equal or lower speeds...
We've other customers with dual GTX280 cards with positive results.
We even have a machine at the office that has 2x GTX200 series and it runs at approx 1.8x the speed of one.
Radiance
Win 7 x64 & ubuntu | 2x GTX480 | Quad 2.66GHz | 8GB
Hi Radiance
I told you some time ago (although as I once put a lot of issues, perhaps did not see) that when I tried two cards with the demo on linux, not working properly, but if using the commercial version (although at two architectures different was not much difference, but he had)
There may be some problem with demo versions.
I told you some time ago (although as I once put a lot of issues, perhaps did not see) that when I tried two cards with the demo on linux, not working properly, but if using the commercial version (although at two architectures different was not much difference, but he had)
There may be some problem with demo versions.
LiVux 64bit | Geforce GTX260 Core 216 and GTX470 | i7 860 | 8Gb | Drivers 275.09.07 | Cuda 4.0.17
I really like this render. What I read here on the forum, I see that its a great product. I really want to make sure my system can run it properly before I purchase it.
I am stumped. In a DX10 enviroment such as Futuremark Vantage benchmark my GPU score with no SLI is 10983 but with SLI on I get 17562 which clearly shows that adding the second GPU is doing its stuff. I checked to see if others online with same graphics cards and processors get same results and they do.
The two graphics cards are working as they should at least in DX10.
Thanks for everyone's help.
xsipower
I am stumped. In a DX10 enviroment such as Futuremark Vantage benchmark my GPU score with no SLI is 10983 but with SLI on I get 17562 which clearly shows that adding the second GPU is doing its stuff. I checked to see if others online with same graphics cards and processors get same results and they do.
The two graphics cards are working as they should at least in DX10.
Thanks for everyone's help.
xsipower
I think my issue may be my motherboard chipset. I have been studying the details of PCI-e data bandwidth and my board has PCI-e 1.0 slots which have half the bandwidth of more modern PCI-e 2.0 slots.
I maybe wrong here, but I advise that if you are going to use a powerful GPU (200 series or above) make sure that your PCI-e slots or 2.0 version.
xsipower
I maybe wrong here, but I advise that if you are going to use a powerful GPU (200 series or above) make sure that your PCI-e slots or 2.0 version.
xsipower
Hi,xsipower wrote:I think my issue may be my motherboard chipset. I have been studying the details of PCI-e data bandwidth and my board has PCI-e 1.0 slots which have half the bandwidth of more modern PCI-e 2.0 slots.
I maybe wrong here, but I advise that if you are going to use a powerful GPU (200 series or above) make sure that your PCI-e slots or 2.0 version.
xsipower
That can't be the issue.
Octane does'nt require a lot of bandwidth.
You can run 2 GPUs fine on a PCI express 1.0 8x link.
Once the meshnode is loaded into the GPUs, the only thing that goes over the bus are commands and little updates.
Radiance
Win 7 x64 & ubuntu | 2x GTX480 | Quad 2.66GHz | 8GB
Hello radiance,
Thanks for clearing up my PCI-e bandwidth concern.
As the last attempt to figure out my M/s problems I focused on CPU usage as a possible cause for the bottle neck. I ran windows task manager - (process tab) and monitored octanedemo.exe CPU usage. I have a Quad Core Q6850 3GHz CPU.
I loaded the benchmark scene and immediately the CPU usage went to 25% (full usage of a single core) solid never going above 25% which indicates that octanedemo.exe is a single threaded application. Total CPU usage on my system during the renders was ~27-29% indicating that 2-4% was used by other resources on my system.
I then made different complexity obj scenes and imported them into Octane. What I found was that once the scene (with pathtracing) was simple enough for the CPU usage to no longer peg at 25% the fps and M/s began to increase otherwise with it pegged it remained fixed at ~1.43 M/s no mater how many GPUs were being used or what scene was used.
From my tests I would deduce that Octane is sensitive to the raw single core CPU power of the processor. Adding GPUs to a system that has already saturated on core of the CPU will not speed up the render.
My question is, will Octane become multithreaded so that it takes advantage of multi core processors? Otherwise I would think that it would be neccessary to make sure the CPU one uses scales properly with the GPUs one uses.
Thanks,
xsipower
Thanks for clearing up my PCI-e bandwidth concern.
As the last attempt to figure out my M/s problems I focused on CPU usage as a possible cause for the bottle neck. I ran windows task manager - (process tab) and monitored octanedemo.exe CPU usage. I have a Quad Core Q6850 3GHz CPU.
I loaded the benchmark scene and immediately the CPU usage went to 25% (full usage of a single core) solid never going above 25% which indicates that octanedemo.exe is a single threaded application. Total CPU usage on my system during the renders was ~27-29% indicating that 2-4% was used by other resources on my system.
I then made different complexity obj scenes and imported them into Octane. What I found was that once the scene (with pathtracing) was simple enough for the CPU usage to no longer peg at 25% the fps and M/s began to increase otherwise with it pegged it remained fixed at ~1.43 M/s no mater how many GPUs were being used or what scene was used.
From my tests I would deduce that Octane is sensitive to the raw single core CPU power of the processor. Adding GPUs to a system that has already saturated on core of the CPU will not speed up the render.
My question is, will Octane become multithreaded so that it takes advantage of multi core processors? Otherwise I would think that it would be neccessary to make sure the CPU one uses scales properly with the GPUs one uses.
Thanks,
xsipower
Hi,
Octane already runs one thread per GPU.
I already said this is a driver/OS/HW issue, as octane really does'nt need much CPU at all. there can't be any CPU bottleneck with your CPU, even with 8 GPUs.
All octane does is load the scene into your GPU, and then sends simple commands and very small updates.
There is no requirement for PCI-E bandwidth or CPUs other than a bare minimum of a dual core CPU of 2GHz or so, and you can easily use octane on PCI-E 1.0 4x slots with multiple GPUs on such a system.
Radiance
Octane already runs one thread per GPU.
I already said this is a driver/OS/HW issue, as octane really does'nt need much CPU at all. there can't be any CPU bottleneck with your CPU, even with 8 GPUs.
All octane does is load the scene into your GPU, and then sends simple commands and very small updates.
There is no requirement for PCI-E bandwidth or CPUs other than a bare minimum of a dual core CPU of 2GHz or so, and you can easily use octane on PCI-E 1.0 4x slots with multiple GPUs on such a system.
Radiance
Win 7 x64 & ubuntu | 2x GTX480 | Quad 2.66GHz | 8GB
This will be my last post with my present hardware using Octane.
I formated my HD and installed Windows XP 64 Pro SP2. Installed 197.45 nvidia driver. Downloaded Octane demo and scene files. Ran benchmark.ocs as I did with Windows 7 64 Ultimate and guess what? I got the same M/s as with Win7. My CPU usage went from 25% in Win7 to 45% now in WinXP64.
What frame rate and M/s should I expect with the benchmark.ocs with demo resolution with pathfinding turn on? I am using two 280 GTX and I get 0.70M/s with Quad Core Extreme Q6850 @ 3GHZ. Is this way slow? By how much?
Before I run out and purchase new hardware I need to know how bad my system is?
Thanks,
xsipower
I formated my HD and installed Windows XP 64 Pro SP2. Installed 197.45 nvidia driver. Downloaded Octane demo and scene files. Ran benchmark.ocs as I did with Windows 7 64 Ultimate and guess what? I got the same M/s as with Win7. My CPU usage went from 25% in Win7 to 45% now in WinXP64.
What frame rate and M/s should I expect with the benchmark.ocs with demo resolution with pathfinding turn on? I am using two 280 GTX and I get 0.70M/s with Quad Core Extreme Q6850 @ 3GHZ. Is this way slow? By how much?
Before I run out and purchase new hardware I need to know how bad my system is?
Thanks,
xsipower