Hello,
we have wrote an application to render video effects using OpenGL features. This application runs fine but while writing with external buffer method the GTS 250 is able to process more than 390 HD frames for second, both GTX 480 and QuadroFX 5800 are not able to process more than 45-48fps.
can someone tell me what is happening?
the GTX480 is faster or not than a cheaper GTS250?
regards
Vince
That’s not standard terminology. What are you talking about in GL terms?
… the GTS 250 is able to process more than 390 HD frames for second, both GTX 480 and QuadroFX 5800 are not able to process more than 45-48fps.
FPS is bad for comparisons (random link). Converting to something useful like time, we find:
GTS250: 2.6 ms/frame
GTX480: 22.2 ms/frame
So we’re talking 10X slower on the newer board. Sounds like maybe you’re slipping into a software fallback, some driver issue, or something about the GTX480 hardware/drivers/system is just flat slower than the GTS250 system. Do these systems have the same motherboard, CPU, memory, OS, drivers (incl NVidia driver version), and the only thing varying is the GPU? If not, try these tests in the same system swapping only the GPU.
can someone tell me what is happening?
You’re kidding, right? Give us some details to work with here 
Please describe in detail what you’re doing, preferably illustrating with a short compilable test program so folks can help you narrow down the problem.
Hello
Yes we ve compared the performance using the same systems swapping the VGA board only or sometimes changing some drivers too
Our reference systems are based on
Ci7 920 -Asus P6TD Deluxe -6GB RAM (2GBx3Ch)- XP/Win 7 PRO- Nv drivers NV 197 and 257.
Ci7 980x -Supermicro X8SAX -12GB RAM (2GBx6ch)- XP/Win 7 PRO- Nv drivers NV 197 and 257.
HP 8200 - (Dual xeon Nehalem 3,3ghz - 12GB)-Win 7 PRO -Nv drivers NV 197 and 257.
we are using PBO to send/render/retrieve uncompressed video frames to GPU. If we send/retrieve without to render, all is fast as on GTS250 or all other boards like (GT8800, GT9800, GTX260/275/280/285). Using the GTX 480, It seems be like if there is a kind of lock on the rendered buffer that slow-down the exporting of frames.
To export the code relatively to this matter is not immediate but we need for some days to prepare a new tester application.
I’am also experiencing very slow results with the GTX 480 in Motionbuilder while my previous GPU which was a ATI HD4870 was delivering much higher FPS in the viewport and while rendering. Im not sure what the issue is but reakon its related to the new fermi architecture or the gpu’s opengl support with opengl software. I’ll try and post up a benchmark and more details if someone is interested in helping me solve this problem.
This topic was automatically closed 183 days after the last reply. New replies are no longer allowed.