Gigabyte GTX 670 2GB OC performance

asot

Distinguished
Apr 7, 2012
55
0
18,530
Hi Guys,

Finding out my 550 ti couldn't handle the stress anymore I decided to make an upgrade to Gigabyte's 670 GTX OC 2GB.
http://www.gigabyte.com/products/product-page.aspx?pid=4211#ov

I've read several reviews and they all pointed out it is a very decent GPU that should handle 1080p gaming with the highest details.

I have made some comparison tests with fraps in The Witcher 2, which seemed ok, but I also decided to try the unigine 3.0 test where according to a review my card scored lower than a standard GTX 670 (1500 vs 1800+).

So I'm pretty puzzled now... could you help me out in running some relevant tests ?

I have an AMD Phenom x6 1090T OC@ 3700 , 4 gigs of RAM. From what I've read, these should not bottleneck the GPU.

Also running windows 7, 64 bit.

Appreciate all the help/suggestions I can get!


Thank you!!!
 

asot

Distinguished
Apr 7, 2012
55
0
18,530


I'm running latest Nvidia drivers. Had a clean install last night. Slightly offtopic, do you own an OC version also ? I'm still puzzled with the "boost" on this card. In OC guru it shows GPU clock: base 980/ boost 1058. How do you activate it really ? :S
 
AMD CPUs don't do that well in Witcher 2: http://www.techspot.com/review/405-the-witcher-2-performance/page8.html So that probably explains that pretty well. Just pick some other games to test with.

What Unigen test are you using, Heaven or the new Valley? I thought Heaven was V4 and Valley was V1 . . . but it's been a while since I downloaded them.

If you run the 670 at stock settings rather than the factory overclock does the performance decrease a bit, or actually increase? Just trying to rule out thermal issues, but that's pretty doubtful anyway.
 

asot

Distinguished
Apr 7, 2012
55
0
18,530


Wow, didn't think there was such a CPU impact on TW2. Thanks for the heads up! Still, overall performance there is good, mostly 60 fps, pretty much only drops in cut scenes to minimum 25, but it seems this is caused by cinematic depth of field. Without it it stays at 55+.

I used Heaven DX11 Benchmark 3.0 because this was the one used in that particular review. I haven't made any changes to the graphics settings since I didn't wanna screw up anything while I'm not informed.
I didn't measure the temperature(funny, OC guru doesn't show that) but the fans were used at max 45%.
If I can't find any satisfactory tests to compare my GPU performance I might try to change it to the stock settings as you suggested.



 
Comparing to online reviews will be difficult since drivers will be different, and other hardware will be different. I think I'm running about 1550 in Heaven 4, but I run it on custom settings and in a window generally (and that's just from memory, I only really ran it when I was benchmarking my machine when I first got my 670 a couple months back).

How do you score on Passmarks video card benchmark (videocardbenchmark.net)? Since they aggregate everyone's results it's a better real world comparison to a huge array of other hardware using the same video card rather than a machine designed for video card benchmarking by eliminating other components that could slow things down.

 

asot

Distinguished
Apr 7, 2012
55
0
18,530


Great tip, thanks! Just ran the test. Here are the results:
http://www.passmark.com/baselines/V8/display.php?id=8358403634

It seems average G3D mark for GTX 670 is 5346. I got 5362.

Last 5 Baselines for GeForce GTX 670
Most recent listed first
BL083553 - May 08 2013
5462
BL083519 - May 08 2013
5678
BL083462 - May 07 2013
5818
BL083446 - May 07 2013
5463
BL083372 - May 07 2013
5434

Most are higher than mine so should I assume that they're all OCed ?
 
It could be different OC's, different driver versions, different supporting hardware. If you think about it in percent terms instead, the difference between the lowest and the highest number there is only 472 points, or just under 9%. The stock speed is 915Mhz, so a 9% OC would only take that to 997 . . . under some factory overclock and boost speeds.

Next you could try OCCT, it reads the current GPU speed when running the GPU test and you can see if your OC and Boost is working.

 

asot

Distinguished
Apr 7, 2012
55
0
18,530


Thanks, will do that and let you know!
 

asot

Distinguished
Apr 7, 2012
55
0
18,530
The sad part is that I just installed crysis 1. I maxed it out on 1920x1080 and it seems I can get as low as 17 fps in some areas... wtf really :|

I've even disabled AA, and nothing. Turning down details didn't help much either... I wonder what could be causing this ? I'm fairly sure people have played this with at least 30 fps constant without having a SLI. And a GTX 670 should really be able to run this smoothly...
 

madcap42

Distinguished
Jan 2, 2012
464
0
18,810

sorry been busy i have the boost turned on bios, but that's for my processor,maybe you could try msi-afterburner, i'm pretty sure the graphic card determines whether to use it's overclock or not i could be wrong
 


So you go and pick a notoriously CPU bound older (hint, not thread friendly) game to run . . . Are you just going out of your way to find CPU bound games that run poorly on AMD CPU's? (Of course, what games DON'T run poorly on AMD . . . not many).

It seems to me that any benchmarks that are designed to test the video hardware come in right on average of what you would expect for a 670, but anything that depends on your CPU, and more specifically single thread CPU performance, that your scores are lacking . . . Can you guess what this evidence points to?

Do you happen to have a copy of BF3 that you could test with? Since it's a console port with just some extra flash textures thrown on top then it scales really well with GPU power, but CPU doesn't really matter beyond a certain point. Take a look at the techspot review here:

http://www.techspot.com/review/458-battlefield-3-performance/page7.html (Notice after the Athlon II X2 at the bottom that pretty much every CPU comes in near the same performance).

Yet GPU scaling:

http://www.techspot.com/review/458-battlefield-3-performance/page5.html (You should be about the 580 level or so)

You can read how they tested here: http://www.techspot.com/review/458-battlefield-3-performance/page2.html