Originally posted by: Matthias99
Originally posted by: DaveSimmons
Originally posted by: CaiNaM
does doom3 tell us how HL2 or STALKER will play? NO. will HL2 tell us how Doom3 will play? NO.
sorry, but as the old adage goes, "if it's good for the goose....", so applying the same logic, doom3, etc. also "sucks" for benchmarks. his selective logic just doesn't cut it.
while i'd certainly agree that an actual game holds a bit more relevance, even within a "real" game, performance will vary depending on which part of the game you choose to benchmark. just look at far cry - depending on the area chosen, and advantage can be shown for ati or nvidia...
A 3DMark score gives no idea about performance in any real game. It's not the same logic, because a Doom3 benchmark gives some idea about performance in not just D3, but also in D3 engine games and for other game engines that are optimized along similar lines. Same with HL2. If you see a review with D3, UT2K4, Far Cry, and (someday) HL2 you have a very good idea of how that video card will work for games now and for the next couple of years. What does adding a 3DMark score tell you?
It gives you (theoretically) a score that is specifically NOT tied to a particular game engine, but indicative of a video card's performance wrt a specific DirectX/shader level. If you want to measure overall raw graphics performance in a stable, repeatable manner, this is not a bad thing.
If you want to ballpark how well a card might run a future game engine utilizing a lot of SM2.0b/SM3.0 shaders, 3DMark05 would give you an estimate.
And as for your example -- take your scores in D3, UT2K4, Far Cry, and HL2/Source. What does that tell you about, specifically, STALKER (which is probably not any more demanding than HL2, but could be very different than all those games)? Not much -- at least not a whole lot more than 3DMark03 would.
It's NOT a replacement for application-level benchmarking, but it is a useful measurement precisely because it ISN'T tied to a particular game or engine, and it's more sensitive to small performance changes than a lot of game engines are. It also has features (like specific frame capture, and batch run modes) that are useful for doing IQ comparisons and the like (well, at least the Pro version does), which a lot of game engines just don't supply. The folks trying to look cool by spouting "3DMock is teh suXX0r!!!" for the millionth time are barking up the wrong tree; synthetic benchmarking programs are just tools, and they have their uses.
And I fully agree that Kyle from |H| can say whatever the hell he wants about FM and 3DMark, and any legal action that FM takes will likely be thrown out. But if I had spent a whole lot of time and money developing a new benchmarking program, and then a major hardware site said on their front page (with no further explanation!) that it "sucks as a benchmark", then I'd be pretty mad, too.
They are a bunch of scumbags and I intend to influence any who care to listen to boycott their products. You do what you want, but don't try running that condencending nonsense about what purpose they serve when we know damned well already, and just don't give a shat 
Facebook
Twitter