Simple Streams Sample On GTX260 - NO improvement


When running the provided simple streams example on my GTX260 i get almost no improvement when using streams. Can anyone explain why this may be, or are there some settings i need to change to get it to work?

Attached is the console output.

streams do nothing on vista–check the release notes.

Ah… good shout. Cheers

I think I replied too quickly before and might have confused things for some people. So, a clarification:

Streams do two things. First, you get completely asynchronous operation for the GPU. Second, it allows you to overlap one memory transfer at a time from the CPU to the GPU (or vice-versa) with kernel execution.

In Vista, streams still do the former just fine; what they don’t do is overlap memcpy and kernel execution, which is why the stream SDK sample shows no improvement.

Somewhere I heard for GPU programs than can be written in DirectX or Cuda, Streaming was the primary performance advantage that Cuda would have.

So I wonder does this mean the main performance benefit for using Cuda goes out the window on Vista?

Is this something that will be fixed going forward?

Thank you,


Maybe you should check the releasenotes?

Was your reply just a guess that there may be something relevant in the release notes, or are you implying there is a specific point addressing the question but you choose not mention what that point is?

Since an NVidia person commented on the issue it would seem logical that if a fix was imminent hey would have mentioned something.

Well, actually I was writing the post, in the meantime looking for the releasenotes, but as I unfortunately could not find them, sent the post like that. My guess was that the releasenotes would mention it is something that is not working yet (like Tesla e.g.).