my EVGA GTX670 FTW vs Galaxy GTX570 results

[toyota]

I just got finished doing an out of the box comparison of my EVGA GTX FTW model vs my Galaxy GTX570. no the EVGA card is oced a bit from the factory and the Galaxy GTX570 is stock. now the EVGA card is oced a bit from the factory and the Galaxy GTX570 is stock. I ran the GTX570 at stock speeds anyway for 99% of the time I owned it so I want to see how the EVGA GTX670 directly compared. all of the following benchmarks were done IN GAME and represent real GAMEPLAY. all tests were done at 1920x1080.

its about a 43% improvement overall if we leave out the one anomaly at the end.


Alien vs Predator with all max DX11 settings with 4x AA

GTX570
Min, Max, Avg
41, 81, 57.123

GTX670
Min, Max, Avg
63, 124, 84.040


GTA 4 every setting maxed

GTX570
Min, Max, Avg
27, 81, 45.830

GTX670
Min, Max, Avg
35, 114, 59.254


Crysis DX10 with all very high settings and 4x AA

GTX570
Min, Max, Avg
31, 55, 40.734

GTX670
Min, Max, Avg
45, 72, 60.045


Metro 2033 with DX11 very high with only tessellation

GTX570
Min, Max, Avg
26, 80, 49.345

GTX670
Min, Max, Avg
35, 120, 74.189


Metro 2033 with DX11 very high with tessellation and 4x MSAA

GTX570
Min, Max, Avg
21, 83, 43.866

GTX670
Min, Max, Avg
27, 115, 56.409


Metro 2033 with DX11 very high with tessellation, 4x MSAA and Physx

GTX570
Min, Max, Avg
18, 65, 36.291

GTX670
Min, Max, Avg
17, 97, 51.896


Metro 2033 with DX11 very high with tessellation, 4x MSAA, Physx and DOF

GTX570
Min, Max, Avg
16, 46, 29.342

GTX670
Min, Max, Avg
20, 56, 35.259


Alan Wake fully maxed and 8x AA

GTX570
Min, Max, Avg
34, 67, 51.676

GTX670
Min, Max, Avg
50, 104, 77.903


Batman Arkham City with DX11 extreme settings and 2x MSAA with high Physx

GTX570
Min, Max, Avg
29, 90, 50.552

GTX670
Min, Max, Avg
25, 114, 72.613


Batman Arkham City with DX11 extreme with 8x MSAA and high Physx

GTX570
Min, Max, Avg
23, 63, 33.628

GTX670
Min, Max, Avg
29, 89, 51.217


Crysis 2 DX11 ultra with high res textures

GTX570
Min, Max, Avg
37, 54, 45.716

GTX670
Min, Max, Avg
61, 83, 68.671


Just Cause 2 all settings on maximum with 8x AA

GTX570
Min, Max, Avg
36, 75, 49.211

GTX670
Min, Max, Avg
41, 120, 75.566


Red Faction Armageddon on maximum settings with 8x AA

GTX570
Min, Max, Avg
38, 60, 50.744

GTX670
Min, Max, Avg
49, 80, 67.065


Clear Sky with all max settings including 4x AA

GTX570
Min, Max, Avg
10, 27, 17.151

GTX670
Min, Max, Avg
16, 39, 24.868


ok I saved the one anomaly for last. I have no idea what is going on here in Clear Sky as my gpu was ramping up fine so it must be a driver issue. it seems odd that on full max settings it was faster than GTX570 but on reduced settings it was slower than the GTX570.

Clear Sky with max settings except sun quality, sun rays and SSAO lowered to medium but still 4x AA

GTX570
Min, Max, Avg
30, 60, 43.405

GTX670
Min, Max, Avg
22, 46, 31.576 (WTF?)

EDIT:

after a forced reboot from a blue screen during testing of Clear Sky the game now runs like it should so go figure.

GTX670
Min, Max, Avg
61, 117, 78.764

 

[d3fu5i0n]

Thanks for the benchmarks.
The results for 'Clear Sky' @ "with max expect sun quality, sun rays and SSAO lowered to medium but still 4x AA" are very strange though. Hmmh...

 

[TakeNoPrisoners]

The Clear Sky results don't mean much. Since the game is old and not really that popular the driver team most likely does not care about it.

A better game to use would be Stalker SOC, that one has DX 11 and is a much better game than Clear Sky. It is more popular and recent.

 

[toyota]

The Clear Sky results don't mean much. Since the game is old and not really that popular the driver team most likely does not care about it.

A better game to use would be Stalker SOC, that one has DX 11 and is a much better game than Clear Sky. It is more popular and recent.

um it does mean something if you want to play it and it is a still fairly modern game. no game should go down in performance when you "upgrade" especially if it becomes unplayable compared to the previous card. and you must mean COP not SOC because SOC was the very first STAKLER game.

 

[Leyawiin]

Thanks Toyota. I can see that having a solid CPU and decent monitor size makes a huge difference when you buy a current gen GPU. I just got an HD 7850, OC'd it to 1100 Mhz and it shows almost no improvement over my GTX 460 1GB. Problem is I have an X4 980 BE and a 1680x1050. Finally experiencing a bottleneck situation. Your results show me I need to start saving up for a new system.

 

[exar333]

Thanks Toyota. I can see that having a solid CPU and decent monitor size makes a huge difference when you buy a current gen GPU. I just got an HD 7850, OC'd it to 1100 Mhz and it shows almost no improvement over my GTX 460 1GB. Problem is I have an X4 980 BE and a 1680x1050. Finally experiencing a bottleneck situation. Your results show me I need to start saving up for a new system.

A new monitor would be a great way to start out. 1080P/1200P would still work great on your current rig.

 

[Termie]

Great set of benchmarks. I figured you'd get a 50% improvement, and you got almost exactly that in some games. In others it wasn't as big an improvement. Still, that's a significant enough upgrade that you'll feel the difference. Now, hopefully you can sell your GTX570 to make up for some of the sting of this generation's price/performance issues.

By the way, how does the FTW cooler sound at idle and load? Are you able to keep it under 70C without serious noise? That's the threshold for keeping max boost.

 

[Zanovar]

Cheers for the benchmarks,that card a decent overclocker toyota?

 

[toyota]

remember too that this card is basically even with a gtx680 so reference clocked gtx670 would probably be about 7-10% slower.

out of the box, the card hits 1189 for boost but I have not tried to oc it yet. I will be doing another little review later to see how those games scale with just adjusting power target, offset and memory speeds.

 

[Don Karnage]

Max overclock vs max overclock?

 

[toyota]

Max overclock vs max overclock?

I did not want to fool with that because my gtx570 was not a great overclocker unless I upped the voltage which I dont really like to do.

 

[Termie]

I did not want to fool with that because my gtx570 was not a great overclocker unless I upped the voltage which I dont really like to do.

I think the GTX670 would overclock at least as well as the 570, which across the board was not a great overclocker. Now, you're pre-overclocked by about 100MHz here, so at most I'd say you have another 80-90MHz to go. Still, that's going to be a bigger overclock than you would have had on the 570, at least without voltage.

You got about a 40% improvement, stock-to-stock or OC-to-OC, by my estimation.

 

[SirPauly]

Thanks!

 

[toyota]

I think the GTX670 would overclock at least as well as the 570, which across the board was not a great overclocker. Now, you're pre-overclocked by about 100MHz here, so at most I'd say you have another 80-90MHz to go. Still, that's going to be a bigger overclock than you would have had on the 570, at least without voltage.

You got about a 40% improvement, stock-to-stock or OC-to-OC, by my estimation.

yes its 43% if we leave off the one anomaly at the end.

 

[Termie]

yes its 43% if we leave off the one anomaly at the end.

Right, but your 670 has a 10% overclock, so stock-to-stock, it would be more like 40%.

 

[guskline]

toyota, first thank you for the time it took to run all those benchmarks. Clearly the GTX670 is a faster card and anyone that doubts that need only look at all of your benchmarks. With all of your specs, I thought I would see a ssd! The Samsung is a great HD but I have gone to a hybrid setup of ssd for at least OS and some of my most used programs on the ssd with the mechanical HD for data and less used programs.

 

[toyota]

Right, but your 670 has a 10% overclock, so stock-to-stock, it would be more like 40%.

yeah it would be even less that 40% if this was reference clocks for gtx670.

 

[Rvenger]

FTW 670 pulls very close to a Reference 680. I owned a 680 and now a 670 and I cannot see a difference in performance at all. Even the benches are extremely close.

 

[JohnnyChuttz]

Awesome, just makes me want a 670 that much more.

 

[Crap Daddy]

Thanks. Good to know what to expect for those of us who have GTX570. Still not worth an upgrade for me, I'm on 22" 16:10 and the difference there should be more like 35% stock vs. stock for roughly 80$ more that was paid for my card a year ago.

 

[thestrangebrew1]

Awesome info. I was curious about how my 570 would stack up against the FTW card and I now I at least have an idea. I'm still going to hold off on a new card right now because my 570 is doing fine for me, but this is great info nonetheless.

 

[Gordon Freemen]

Thanks Toyota. I can see that having a solid CPU and decent monitor size makes a huge difference when you buy a current gen GPU. I just got an HD 7850, OC'd it to 1100 Mhz and it shows almost no improvement over my GTX 460 1GB. Problem is I have an X4 980 BE and a 1680x1050. Finally experiencing a bottleneck situation. Your results show me I need to start saving up for a new system.

Why get a new rig if you are getting acceptable performance already perhaps logic failed I presume ?

 

[Gordon Freemen]

A new monitor would be a great way to start out. 1080P/1200P would still work great on your current rig.

Some people prefer performance over resolution as in ultra fast response time and color quality and contrast I will stick with my stellar Samsung Syncmaster 226BW 1680x1050 reso for some time.

 

[blckgrffn]

Why get a new rig if you are getting acceptable performance already perhaps logic failed I presume ?

Wanting to have his GPU stretch it legs? *shrug*

@ Toyota - nice set of benches!

Time consuming but worth it.

The 670 is very capable card, clearly!

 

[toyota]

the gtx670 would have been a killer card for 250 bucks. :biggrin:

I am still miffed about the Clear Sky results. I submitted a ticket with Nvidia and reported it on their forums but I doubt they will really care. there might even be some other games that are worse as that's the only ones I tested. and the reason I did test those were because those are about the only demanding games I have.

EDIT: BFG10K had a 64% improvement going from a gtx580 to a gtx680 so something is really odd there.