System Setup and Notes
I am still using the overclocked Phenom II X6 1100T, and at 3.7 GHz it is still faster than the FX-8150. I will continue using this processor for a while, at least until Vishera-based FX processors hit the streets. A stock clocked X6 1100T is in the same overall performance range as the Intel Core i7 940, but it is pretty seriously outperformed by the i5 2500 (Sandy Bridge) series and beyond. Still, I do think that it is good to see these video cards reviewed on a platform that is simply not covered by other sites. Not everyone has a high end desktop, so it is a valid comparison to view these products on what is most definitely a midrange platform.
I tested the GTX 670 against the MSI R7870 HAWK (overclocked out of the box to 1100 MHz) and the MSI R7950 OC (again overclocked out of the box to 880 MHz). The R7870 retails in the $319 range (though it could receive a price cut here pretty shortly), while the R7950 is not far off at $329. This makes the GTX 670 around $100 more expensive before rebate. The results that you are about to see make this a very interesting comparison.
I tested these boards in a closed case. I am not entirely fond of testing thermals and power in an open test bed, as it is not entirely representative of what the average user will utilize. I use the Enermax Fulmo GT, which is a full size ATX case with an obscene amount of fans. Cooling will not be an issue with this particular product, but it is honestly closer to the realities of most applications.
The full test system configuration is as follows:
- Phenom II X6 1100T @ 3.7 GHz
- Asus SABERTOOTH 990FX Motherboard
- 2 x 4GB GSKill DDR-3 1866 Modules @ 1600 MHz 9.9.9.27 timings
- Samsung 120 GB SSD
- WD Caviar Black 1TB HD
- Lite-On BD Drive
- Corsair AX1200 Power Supply
- Enermax Fulmo GT Case
- Windows 7 Ultimate 64 Bit
I was wondering how much of a
I was wondering how much of a performance gain is seen versus a reference GTX670. Unfortunately, this is not included in the graphs of the game testing. Too bad..
I don’t understand the last part of the article, where you mention that MSI’s default overclock, is actually causing crashes when put into benchmark / torture test. This would mean MSI’s overclock is unstable?
You are looking at most at
You are looking at most at about a 4% increase over a stock clocked GTX 670. The differences are not all that great. You are primarily getting this card for the cooling and unique design/voltage control rather than it being a much, much faster product than the stock GTX 670.
The last part actually reads, "At no time was this card unstable during benchmarking and torture tests…" So basically it was perfectly stable even with the boost speed up to 1170+ right out of the box. The card performed without issue.
Hey Josh, I wanted to say
Hey Josh, I wanted to say thanks for the article. I have been on the fence about upgrading from my old xfx radeon 5870. I had been hunting the various 670/680 cards and the amd 7970 and so forth. After all is said and done I got this card yesterday for $395 with borderlands 2 and mafia 2. I have overclocked it an have the core at 1170 mhz and my boost has been anywhere from 1250-1350mhz. This is of course with a voltage bump and such. Anyway, thank you again for a quality article that has resulted in a massive performance bump for me. Temperatures are still low even under load and even at 50% fan the card is still nice and quiet. 😀
Josh, I was comparing the BF3
Josh, I was comparing the BF3 at 2560×1600 Ultra results in your article to those from the Galaxy GTX 670 GC 4GB article. Your results for the MSI GTX 670 PE card are 48fps min and 63.8fps avg. The results for the Galaxy card at the same settings are 26fps min, 41fps avg and 66fps max.
Can the BF3 results for the MSI and Galaxy cards from the two articles be compared? If so, why pay more for the extra 2GB of memory on Galaxy card? Any chance the results for the MSI card are avg and max?