- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello,
I have an Ivy Bridge desktop running a clean Windows 7, which I use to test the HD4000 with OpenCL. My application is a software h264 encoder multi thread. For now it only uses the CPU.
The thing is the test application has a big variation in performance, for different tests with the same configuration. Whereas it is pretty much stable when I run exactly the same thing on my Sandy Bridge desktop.
I tried to disable turbo mode and power management to make the frequency stable but it changes nothing. I have no other processes running in parallel (no more than on the sandy bridge).
I look at the I/O while the app is running and there is nothing different from the sandy bridge. (The source file to encode is stored onto the RAM memory).
That's why I am wondering which parameter could affect the performance that much, and how to solve this problem. The variation is about 10% in encoding time per frame, or even more.
Thank you,
Chris
Link Copied
1 Reply
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
It is hard to guess what is causing this variance. I suggest that you use a profiler like Intel Amplifier XE 2011or a similar tool to verify if the same code paths are executed, same number of threads is used,andif the same amount of time is spent in the same functions. This will give you a hint what might cause the variance.
Reply
Topic Options
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page