GTX470 Performance Test Part 2: Windows 7
2xAA
Here the GTX470 sees a performance drop on Win7 compared to XP in Crysis Warhead and in Jericho.
We do however see a slight performance gain in Hell’s Highway with Win7 compared to XP. The latter was already faster than the GTX285, which means Win7 extends the GTX470’s lead further.
But the clear standout here is Far Cry 2, where even the DX9 path sees a massive 48.79% performance gain on Win7, making it far faster than either the GTX470 or GTX285 XP scores.
In fact, when I set Far Cry 2 to DX10 at 2560×1600 with 2xAA and maxed out its DX10 settings, it still ran faster than DX9 did at 1920×1200 with 2xAA, scoring about 48 FPS. Far Cry 2 is one of the rare games that actually runs faster under DX10 and looks better at the same time.
The rest of the games showed no real performance change by moving to Win7.
Load up Windows 3.1 next!
Seriously, legacy software on modern hardware? If you’re trying to point out driver issues, fine. But there are less ridiculous methods to do so other than ancient games on an obsolete operating system with seemingly random resolution/settings.
Many thanks for this BFG, supremely interesting results there! I need to go off and digest them.
Initial impressions are that the drivers are erm….. slightly dodgy to say the least
XP an obselete OS? Have a look at the Steam Hardware Survey for April 2010. XP 32 bit is the most popular OS with 38% of people using it. Windows 7 is on 25%.
Really Marauder, you should check out your figures before spouting rubbish.
Marauder, are you saying Windows 7 (64 bit) is old? That’s the OS that this article tested, and it was released in 2009.
As for XP, the GTX285 was the fastest card overall, and it was running on XP. So that makes it the reference score to which the GTX470 is compared to.
I don’t know about you, but I play more than just the six cherry picked games that are constantly benchmarked by just about every website out there. I actually play these games regularly and the settings used aren’t random. In most cases they’re the settings I use when I play them.
Call me crazy, but I actually expect a GTX470 to be faster than a GXT285. That’s why I purchased one.
Wow, a breathtaking article! I still play the original Unreal Tournament using 16xS AA, forced by nhancer on WinXP. Both of my rigs have two hard drives each. Two are WinXP-32 and one is Vista x64 SP2, and the other Win7 x64. Some games just play much better on WinXP today, still.
It’s really nice having both, but it means more maintenance (driver updating, etc..). If I wanna play DX10 games, I simply boot to Win7 and Vista (especially if playing with my friends over LAN). If it’s DX9, I’d still prefer WinXP for most games because of greater flexibility in forcing TR-SSAA, triple buffering, refresh rates, etc.. not to mention better performance in some games.
Thank you, BFG10k.. when I own a Fermi, I’ll be sure to play Doom 3 on Win7 instead of XP!
It just seems that GTX 470 is hampered by the reduced # of TMU’s (56 compared to 80 on the GTX 285) along with lower clock (600MHz vs 648MHz). Its memory bandwidth is also considerably lower than GTX 285’s–despite a large increase in shader units, I suspect the bandwidth is the reason why the 470 performs so much worse at 2560×1600 than at 1920×1200 when compared to GTX 285 (see xbitlab’s article).
Keep it up and you’re making Alienbabeltech the most comprehensive and resourceful enthusiast site for GPU owners!
Glad to see you around, Bo_Fox. Your positive comments are always nice to read.
Regarding the performance issues, I don’t think they’re from architectural limitations, but mainly from drivers.
In particular, the 16xS OpenGL games use little to no shading and are all about basic texturing and memory bandwidth, yet the GTX470 flies through the results. Also Serious Sam 2 is a texture-heavy game, but the GTX470 flies there too. There was also a lot of movement in some titles by simply using a different OS.
Then we have really shader-heavy games like Stalker Clear Sky and Crysis, which basically showed no performance gain.
This to me looks like driver issues, especially with regards to TrAA and non-standard AA modes.
I admire the effort given to test in a unique way. Sure, it’s not what everyone else is doing, but that’s why it has information that the others don’t, too.
Posted at Benchmark Reviews: http://benchmarkreviews.com/index.php?option=com_content&task=view&id=10576
Olin, thanks for dropping by, and for the link @ Benchmark Reviews.