TR Forums

Chuckula was wondering what the openCL performance would be of an i3-3225 when compared with Semi-Accurates benchmark. I ran those for him (along with another i7 mobile processor and some generations of nvidia cards) and here are the results. It looks like the i3 can hold it's own against the 5800k in most circumstances.

BTW, just click on the results to get the full picture on your screen. Didn't realize it was that wide (was putting it together on a 2560x1440 screen).

Very Interestly! Thanks for running those benchmarks using the latest drivers.

No problem. When I get around to replacing the HTPC motherboard I'll add a GT-520 card to the results. I suspect it will fall somewhere in between the 8800GT and the GTX-275 in performance.

Btw i really wouldn't trust FlopsCL on Intel platforms. On my machine, if I select the Intel CPU from the dropdown list, the OpenCL information shown in the window is for HD4000 and vice-versa. Plus, the HD4000 doesn't even support double in OpenCL.

The HD 4000 does support limited double precision on multiplication and addition. Depending on how FlopsCL determines it's score, it is entirely possible that it is reporting a valid result if division is not used.

Deanjo wrote:

The HD 4000 does support limited double precision on multiplication and addition. Depending on how FlopsCL determines it's score, it is entirely possible that it is reporting a valid result if division is not used.

Well that support is only exposed in DirectCompute and C++ AMP afaik, not OpenCL. From what I see, the FlopsCL result is running on CPUs in this case.

Ya I fired up GPU-z to monitor the load while running it. It looks like you are right, there is zero load on the gpu during that benchmark.