Originally Posted by Cyro999
If one person has a 760 and one person has two 780ti's, they will see different results on ultra. At times, the 760 will be holding back FPS, making any effects that RAM and CPU have, invisible, and diluting the averages and maximums.
If you put both to 720p for example, GPU no longer has any effect - you see the maximum and average FPS that the system is, at maximum, able to produce, >and may not be able to reach without faster RAM or CPU<<br />
Testing at ultra is good, but to be clear, you're testing performance that's limited in at least two ways, and getting results that are not the clearest performance-wise because of that. It's my firm belief that you should benchmark in three ways:
First, 720p low settings, for CPU/Memory stress and to show the performance achievable at max;
Secondly, 1440p everything max pushing AA etc, for GPU stress and to show how GPU's do relative to eachother;
Thirdly; 1080p max or near max settings to show what many people will experience.
^All three of those results are very helpful. If you only take the third one however, you'll get a very narrow picture of performance - same with never looking at frametime numbers. I think it's important to glance at them a few times and see if any trends pop up, or if not, have an idea of how the game will feel at a given average FPS because of how the chart looks.
I agree, but also disagree with the above.
1. 720p test won't push any limits on a capable system. Unless you were doing 120FPS+ tests would it, to me, make sense
2. 1440p this is the other extreme
3. I think that's the sweetspot as MOST players game on this.
Of course you can do all three, to see the variances, but where do you draw the line and where do you understand the differences.
You'll have three different results and they might not all correlate (maybe because your data is invalid OR because the tests just don't test the same thing).
A CPU/RAM bench will yield different results than a GPU bench.
If that's the ideal for you and the optical, I guess that's great for you - but also if people want to do proper tests, they're going to have to spend a lot of time benching and most people don't have the time or patience -> and I'm not speaking about an 18second bench on an empty server
At least, on my end, I know of no person that's done as many benches as I did and isn't a professional reviewer and/or paid for their work (be it through advertising banners on their site or being literally paid to conduct the review by an employer: ie. Corsair's benches)
In fact,t hat raises the question: You ask quite a lot about benches and make some valid points about them - but why haven't you done any benches?
Say you contribute to the win7 vs win8 thread I created?
Maybe even this RAM thread with ram benches at different times?
It strikes me quite a bit that the people that often request data are the ones that love reading the info, but can never be asked to collect the data themselves to share with the public.
That's not a dig to you, but a general point and observation I've made through the years I've been reviewing.
I'm sure you'll agree with me - the more the merrier.
More people with information, the better. It doesn't give 1 persons perspective from 1 PC, but several, which is exactly my point with the RAM benches: I want to see more benches done. If I could have done them myself, I would have.Edited by Totally Dubbed - 11/17/13 at 11:57am