Question
Leaderboard question
11 years 1 week ago #503
by Michal 'bonzaj' Staniszewski
Hi,
Yes that one is difficult.
The best way to compare GPUs is by setting them to the same frequency and testing them on the same PC components. And this is what I did with single GTX and R7970. they are present in the leaderboards with the name "catzilla reference". But of course we won't be able to do it since, we would need to test every card by ourselves. So the current proposal for algorithm is:
1. if we have at least 5 GPUs of the same type in the leaderboards
2. add their results and divide by 5
3. use this mixed result for comparison
Of course it may be influenced by users, so if one GPU will have 5 users that overclocked it highly - then it means that you are able to clock it to such degree.
The next thing is that we would need to grab results from at least "tiger" board for GPUs and from KItty for CPUs. The crew will be back online tomorrow so we will be able to move forward for the final release
.
Yes that one is difficult.
The best way to compare GPUs is by setting them to the same frequency and testing them on the same PC components. And this is what I did with single GTX and R7970. they are present in the leaderboards with the name "catzilla reference". But of course we won't be able to do it since, we would need to test every card by ourselves. So the current proposal for algorithm is:
1. if we have at least 5 GPUs of the same type in the leaderboards
2. add their results and divide by 5
3. use this mixed result for comparison
Of course it may be influenced by users, so if one GPU will have 5 users that overclocked it highly - then it means that you are able to clock it to such degree.
The next thing is that we would need to grab results from at least "tiger" board for GPUs and from KItty for CPUs. The crew will be back online tomorrow so we will be able to move forward for the final release
. Please Log in or Create an account to join the conversation.
