Originally Posted by Skrillex
And cut fps in half. It would prove innacurate outcomes.
People always ask will x cpu bottleneck y gpu.
I do not have a plethora of CPU's laying around, I'm just providing what benchmarks I can then giving an overall opinion of the CPU. People can leverage my results agaisnt review sites and then draw that think as the winner.
Coming from an i7 system I can honestly the say the performance seams to have taken a dip at the same clocks as the Phenom II aswell.
Minimum FPS on the i7 920 @ 4Ghz was much better but on the Phenom I definitley see dips and places where the CPU is holding the GPU back.
But that is my opinion take it with a pinch of salt, don't take it as gospel nor write it off as rubbish.
Using Fraps to log your FPS doesn't cut framerates more than maybe 1%. It's using Fraps to record gameplay that wipes out your FPS.
And you're right: "People always ask will x cpu bottleneck y gpu."
And it's an annoying question, every time they do it
Because it's impossible to answer w/o knowing what test (i.e. what bench/game), and what settings you are running that test at.
Heck, probably nobody would say my i7 at 4.2 'would bottleneck' my 470's in SLI ... but it happens all the time, actually. In fact just last night I determined that my (very noticeable) FPS drops down into the 40's in the Witcher 2 (totally maxed out at 1920x1200, minus Ubersampling) ... are in fact being caused by CPU bottlenecking.
W/regards to the present thread, my point is that just running a bunch of benchmarks at one CPU clock and saying 'here's my results, now go look at review sites and figure out if the CPU is a bottleneck' ... is not useful. No offense, but it's just not. Esp. when the review sites would likely run different settings, or may not use the same benchmark, etc.
Not to mention, their 'goal' is not to tell you 'what FPS you'll get at home' w/a given card. Their goal is to compare cards against each other in an identical testbed. As such, one should really never look at reviews and think 'this is what I *should get* on this test'. Those guys do all kinds of random stuff that they don't tell you about. As long as they keep the testbed the same for each card, their results are 'valid', even if they are not 'accurate' vs. what you'll see at home.
If you want to provide a really useful study, you should run the same test twice, once w/the CPU at stock, and once w/it OC'd, then do a breakdown of the performance improvements that came w/the OC Edited by brettjv - 6/10/11 at 6:31pm