Originally Posted by areamike
I know this thread is getting old, but what really are the points of benchmarks and overclocking etc. if in the end it's all about getting decent results when playing games?
I have a measly XFX Radeon HD 6870 and BF3 and BFBC2 look and play wonderful. I don't overclock and don't plan to.
Is it all about the "mine is bigger than yours" mentality that it boils down to?
The most useful thing w/benchmarks is providing an environment conducive to 'apples to apples' testing. I.E. it allows for 'controlled' experiments to compare the relative performance of GPU's.
If we see that GPU A scores 30fps in the Heaven bench, and GPU B scores 60fps in the Heaven bench, we can safely assert that GPU B is very likely to provide higher 'maximum playable settings' than GPU A in many actual gaming scenarios, particularly in a game that runs in DX11 and deploys tessellation, since those are kinda the 'focus' of this particular benchmark.
Obviously, no single benchmark (whether 'synthetic' like Heaven, or a benchmark of a particular game) should be relied upon entirely for decision-making, rather, a nice cross section of benchmarks, preferably representing most of the common game engines presently in use (i.e. Source, Unreal, CryEngine, Creation, Frostbite, etc) should be used to assess relative performance. This is important because a particular GPU architecture might perform better than another architecture in a particular engine, but turn out to perform much worse in a different engine.
On a more practical, day-to-day basis, benchmarks are handy for assessing whether a person's system is operating correctly, i.e. trouble-shooting. We often see people on here posting their benchmark results and saying 'does this score look right'? This allows others w/similar gear to provide their results of the same bench, at the same settings, to help the poster determine whether their system is operating correctly ... this is possible because it's a benchmark, which should display consistent performance ... as long as the settings it's run at, and the hardware components it's run on, are kept consistent.
But sometimes, yeah ... we just run benches against each other competitively, solely for the sake of bragging rights/e-peen. There is a special benchmarking forum here on OCN for these types of threads, where people maintain 'leader boards' and what-not.
AFA overclocking goes, to be painfully honest, outside of benchmark competitions, it's a mostly worthless pursuit for a heck of a lot of gfx cards out there. Reason being, there's not a lot of cards that can legitimately overclock ENOUGH that the increase in performance would be noticeable w/the naked eye vs. just running the card at stock clocks.
TBH I cannot recall there being an ATI/AMD card that legitimately had enough overclocking 'headroom' (i.e. the maximum % clock increase, which in turn affects max % increase FPS) to grant the possibility of increasing the 'max playable settings' in a given game to any significant degree.
The best OC'ing card from AMD we've seen lately was probably the 5850, which had enough OC'ing headroom to increase the FPS by a solid 20% via OC'ing ... which is, in my humble esteem, just about the minimum difference in FPS that's actually readily visible to the naked eye. IOW, a maxed out 5850 could very well 'feel' a bit different than a stock 5850, but it's doubtful the difference would be enough to make the game playable at higher maximum settings (such as, to make it so that 4xAA is playable, whereas 2xAA was the most you could do w/a stock card).
However, nVIdia has put out a few cards over the past few years that had HUGE OC'ing headroom available, in particular the GTX460, 465, and 470, the latter two of which were often easily OC'd by 35-40% from their stock clocks, resulting in FPS increases in the range of 30%. And an fps increase of 30% is definitely enough that you'd 'feel it' in gaming scenarios where the FPS is in the range of 30-45 w/a stock card, and could very well be enough difference that it could affect max playable settings, like take you to playing smoothly at 8xAA, instead of only 2xAA with the card at stock.
And if we talk about putting cards onto more exotic cooling solutions like water, there's actually quite a few nVidia cards that can pull out 35-40% increases in FPS, which really gets you into the territory where you're looking at the same perf as cards that cost many hundreds of $ more (when the more expensive card is left stock) ...
For example, our own BallaTheFeared has a couple of 470's on water that will keep up with a pair of stock GTX580's ... considering that 470's go for under $200 these days, and 580's go for around $400 a pop ... even after paying for the water cooling, one can see how a willingness to grab a good OC'ing card, and really crank it up ... can potentially save you a fair chunk of cash vs. the person who isn't willing to OC Edited by brettjv - 12/8/11 at 12:44pm