|
|
On 23.05.10 21:24, Warp wrote:
> Thorsten Froehlich<tho### [at] trfde> wrote:
>> On 23.05.10 20:33, Alain wrote:
>>> As it is now, there is NO preview and the image is NOT saved anywhere.
>
>> That is kind of the point of a processor benchmark and intentional :-)
>
> How much overhead would displaying the image cause in reality? After all,
> it would be nicer to see the image being rendered instead of just some
> percentage...
Well, the overhead would not be much indeed. But that isn't the problem. The
problem is that it cannot be predicted how much overhead it uses, and it
will depend on the GPU (if any). Running this on a system meant as a server,
which is usually fast running headless might add bias of a few percent
consistently even when doing multiple runs...
Or imagine a slow CPU and an X display over an encrypted connection. The
data transfer could draw so much CPU power...
Or a million of other alternative scenarios that could make the benchmarking
go wrong...
In short: No display and no disk output simply eliminates two random
variables from the measurement.
Thorsten
Post a reply to this message
|
|