Hi,

Before September CPU usage time was a lot lower than the GPU. For example for a task that required 55000 seconds of GPU time used about 8000 of CPU.
But the GPUGRID team changed something recently because for the last month times are very similar. 57000 of GPU time and 56000 of CPU time. This happens both for normal and long tasks.

I have a free CPU core just for GPUGRID and its always using 99.4% of it.

Is it simply that the CPU will always do as much as it can to accelerate the task and will always be at 100%?
or if I buy a faster GPU it won't process faster because my CPU will slower it down?
Is my CPU already slowing me down?

I have a GTX 275, Q6600@3.2Ghz under Linux 64bits (Fedora 15).

Thanks.

But the GPUGRID team changed something recently because for the last month times are very similar. 57000 of GPU time and 56000 of CPU time. This happens both for normal and long tasks.

This behavior is caused by the SWAN_SYNC environmental variable, probably you've set it, and forget about it.

I have a free CPU core just for GPUGRID and its always using 99.4% of it.

Is it simply that the CPU will always do as much as it can to accelerate the task and will always be at 100%?

Yes, but it's intended for Fermi based GPUs, therefore it won't accelerate a GTX 275 much.

or if I buy a faster GPU it won't process faster because my CPU will slower it down?

A faster GPU will be faster with your existing CPU, but it would be even faster with a newer (Core i3 and up) CPU, because they have integrated memory controller, and integrated PCIe bus (Except the Core i7-9xx series).
I have a Core2 Quad 9650 @ 4GHz (FSB 444MHz), its performance almost reaches the performance of the Core i series.

Is my CPU already slowing me down?

Not significantly.

I have a GTX 275, Q6600@3.2Ghz under Linux 64bits (Fedora 15).

I guess its FSB is 356MHz. The chipset of your mb is relevant, because the NB connects the CPU to the PCIe bus. The P45 and X48 chipset is the best for the Core 2 series.