GALAXY GeForce GTX 670 GC Video Card Review @ [H]

GALAXY GeForce GTX 670 GC Video Card Review - GALAXY's new GeForce GTX 670 GC features a dual fan setup on a custom built PCB. GALAXY also tuned 670 GC with a strong 1006MHz overclock. We see how this compares to other video cards in its price range including a Radeon HD 7950 and GeForce GTX 580. Then we put the GALAXY GeForce GTX 670 head to head with a GeForce GTX 680.

Damn impressive. Temps are amazing, even with 75% fan speed. Not that it's an issue since it's only as loud as the rest of the system is, anyway.

Great job, Galaxy.

So basically gtx 680 performance for 60 bucks less AND its available. Awesome card and awesome review. It amazes me how lackluster the gtx 580 is looking now though.

I agree direfox. I jumped the gun and picked up a gtx 580 matrix three months before this gen of cards released and I'm kicking myself now. At least its not struggling at 1080p for the games i'm currently playing. This galaxy card does look nice though. Just wish the pcb wasn't blue.

gee - I am almost having buyers remorse after getting a Galaxy 680 stock card. Not too much though.

Even though I am not overclocking yet, I am currently enjoying Witcher 2, MW3, Batman Arkham City, and a few other games on my 30 inch 2560x1600 monitor with all the eye candy turned on and am getting a great gaming experience.

Good review! Great card but am I the only one one that does not see much point to this card? As is, I find the 670 to closet to the price of a 680. This card just forces that point. So an OC 670 can pretty much match a 680. If I can afford 400USD (440USD in this case), then I can afford 500USD and get a 680 with even more performance head room from OCing.

19% improvement in Witcher 2 from a 9% oc on boost and 8% oc on memory? now that is some good scaling.

I really like this card but it was 20 bucks more than the EVGA FTW model plus no free shipping and I already had trouble justifying the EVGA card.

I'm confused by the way to card boosts. It has a base clock, a boost clock, and an in-game clock? Is this only for the Galaxy model or for all 670's? It's interesting to see a base clock of 1111, boost clock 1190 and a Top boost clock of 1280. So does that mean you have achieved an overclock of 1280 or does that number just fluctuate? Sorry for the silly question, I am just a little confused on how to boost clocks work.

Maverick1337 said:

I'm confused by the way to card boosts. It has a base clock, a boost clock, and an in-game clock? Is this only for the Galaxy model or for all 670's? It's interesting to see a base clock of 1111, boost clock 1190 and a Top boost clock of 1280. So does that mean you have achieved an overclock of 1280 or does that number just fluctuate? Sorry for the silly question, I am just a little confused on how to boost clocks work.

Click to expand...

it seems odd and confusing at first but its actually pretty simple. the advertised boost clocks are just a conservative number that the card will always boost too unless their is an issue. the real boost clocks will be quite a bit higher as long as temps and tdp are not exceeded. all you have to do is run a benchmark to see what your actual boost clocks are. lets say the max boost you see is 1100 then you increase the offset to 50. that will mean you will now see 1150 max boost. of course you need to raise the the power target first just to make sure that the extra boost will not exceed the card's tdp. now when you go past 70C the card will knock off 13mhz and it will knock off another 13mhz at 80C too.

Nice review and looks like a great card.

Why do these always come along AFTER I buy mine.

I have two MSI GTX 670s and have been very impressed with the performance.
They go toe to toe with my son's CF 7970s in every game @ 5760 x 1200.

This Galaxy looks sharp and the cooler appears to work well to boot. cudos.

was debating on getting the 670 a little while back, then i realized there is no point since i already have my 570 overclocked to above satisfactory speeds, and the only thing that's going to come out that's going to be demanding anytime soon is Metro:LL , and by then, the GK110 cards will already be out by then.

Re-sale value for the 570's will still be decent 6-8 months down the road IMO

Quick question if you guys don't mind...

I recently picked up Batman Arkham City on STEAM. The game runs smooth but when I enter doors it lags a lot and takes a bit to "catch up". This normal or are there any tweaks available?

System Specs: See sig (CPU @ 4.2GHz)

cannondale06 said:

it seems odd and confusing at first but its actually pretty simple. the advertised boost clocks are just a conservative number that the card will always boost too unless their is an issue. the real boost clocks will be quite a bit higher as long as temps and tdp are not exceeded. all you have to do is run a benchmark to see what your actual boost clocks are. lets say the max boost you see is 1100 then you increase the offset to 50. that will mean you will now see 1150 max boost. of course you need to raise the the power target first just to make sure that the extra boost will not exceed the card's tdp. now when you go past 70C the card will knock off 13mhz and it will knock off another 13mhz at 80C too.

Click to expand...

Meh, I much prefer my 7970 just running at max overclock all the time - and not deciding by itself when and where it wants to run at max frequency like the 600 series seems to do.

I had the same problem. I had to use FXAA and turn down settings to high from very high.

Unfortunately like all other custom GeForce GTX 600 series video cards, the GALAXY GeForce GTX 670 GC sells for $439.99 at Newegg.

Click to expand...

The Gigabyte 670 uses custom cooler + 680 pcb and is only $400...

Darksword said:

Meh, I much prefer my 7970 just running at max overclock all the time - and not deciding by itself when and where it wants to run at max frequency like the 600 series seems to do.

Click to expand...

you don't understand how it works then.out of the box my gpu will go to 1189 boost and stay there unless my temp goes over 70 C then it will drop to 1176. you can simply raise the fan curve if you have a card that gets that hot and you want it to sat at full boost. the only other way it wont stay at full boost is if I have a framerate limiter or vsync on. and that makes perfect sense because the gpu will only work as hard as it needs to for it to stay at whatever limit I set.

cesium666 said:

The Gigabyte 670 uses custom cooler + 680 pcb and is only $400...

Click to expand...

Yep... the one sitting in my sig (Gigabyte GTX 670 OC w/ custom windforce cooler) at 1354mhz core/+250 mem for 24/7 use, cost $399.99 MSRP from amazon as well.

I can attest to these cards being damn good. I used the 4gb 680OC version of this same cooler, and its definitely a beast. It overclocks to 1300 extremely easy and is rock solid.

The cooler is very good - not as quiet as an accelero or twin frozr 3 but not bad at all considering the size of the card (small!). Definitely a thumbs up from me.

Damn good review, I'm gonna buy a 670 on Thursday and this review just cause me a huge issue. I was having trouble to justify 400$, now 440$? damn it! lol.

I still don't understand [H] reviews sometimes. Your tests are good but they have one major flaw. When you compare Overclocking capability of the card against others, you don't overclock the other cards. Why do you do that? If you're going to overclock one shouldn't you overclock the ones it's being pitted against? The 670 is battling against the 7950 because they are almost the same price and it would be very beneficial if everything was in one review instead of switching back and forth between reviews that have each card overclocked.

I tried to tell a person these 670s were better than 2 295s in SLI but I forgot you can't put brains in a billycart - thnx for the headsup [H], this will go nice in my new 'old' build RE3, 950, Raven rv01 - need a dedicated car-sim machine!!! Want to be a racer!!!!

Maverick1337 said:

I still don't understand [H] reviews sometimes. Your tests are good but they have one major flaw. When you compare Overclocking capability of the card against others, you don't overclock the other cards. Why do you do that? If you're going to overclock one shouldn't you overclock the ones it's being pitted against? The 670 is battling against the 7950 because they are almost the same price and it would be very beneficial if everything was in one review instead of switching back and forth between reviews that have each card overclocked.

Click to expand...

You can't do that because there are lots of custom HD 7950 cards. Which one do you pick ? Unless you have a face-off kind of article which picks the best custom HD 7950 cards (1 or 2 cards) and best custom GTX 670s (1 or 2 cards) and does an overclocking performance article. anyway there are reviews from hardocp like XFX HD 7950 Black edition which give you an idea of HD 7950 OC performance should you wish to compare

Nice card. I really want to see 670s with custom coolers using the normal PCB though or a custom one that is the same size.

Darksword said:

Meh, I much prefer my 7970 just running at max overclock all the time - and not deciding by itself when and where it wants to run at max frequency like the 600 series seems to do.

Click to expand...

Can you provide some sort of evidence that a fixed overclock performs better than adjusting clock based upon GPU loading? Or is this more based upon "this concept is foreign to me and I do not understand it..therefore I must disagree with it".

Maverick1337 said:

I still don't understand [H] reviews sometimes. Your tests are good but they have one major flaw. When you compare Overclocking capability of the card against others, you don't overclock the other cards. Why do you do that? If you're going to overclock one shouldn't you overclock the ones it's being pitted against? The 670 is battling against the 7950 because they are almost the same price and it would be very beneficial if everything was in one review instead of switching back and forth between reviews that have each card overclocked.

Click to expand...

When the [H] reviews 7xxx cards...they do the same thing. When reading the review you must take it in context (e.g. you have to stop looking for the answer you want to see or the argument you want to force). The point of the OC is to compare it to the factory defaults it comes for THAT card. I know you might like to see it in a seperate graph (in this case factory vs OC)..but that is just an extra graph that provides no value other than arguing symantics.

They aren't directly pitting the 670 against a 7950. That is a construct you seem to have forced in your mind that some how it is a always a "fight/duel" and you need to defend your favorite. The 7950 is a reference point. They are reviewing the card in the title and not a pure 6xx vs a pure 7xxx! You really need to read more than one review article to really understand how the card techonology and what specific vendors are providing you..

Video cards are so much more than just FPS. It is a larger construct of value proposition based upon your desired level of gaming.

1. What is the best playable experience I can get with current games with THIS card?
2. What is the power draw for THIS card?
3. How is the driver support for THIS card?
4. How loud is THIS card?
5. How much does THIS card cost?
6. How is the vendor support/warranty for THIS card?

The [H] reviews seem to cover all these items. Thus when I do decide to buy a card i'm not dealing with a "well, I looked at canned benchmark" as my soul source of information. Sorry..while Anand (and others) do a lot of nice detail work...getting down to vendor specifics is not in their mindset. Their review methdology just does not have the ability to show why Vendor X is better than Vendor Y for a card with any degree of credibility/confidence.

Why are temps lower when overclocked? was it just because the fan was manually set at 75%?

I mean there's a HUGE difference with temperatures under load.

Maverick1337 said:

I still don't understand [H] reviews sometimes. Your tests are good but they have one major flaw. When you compare Overclocking capability of the card against others, you don't overclock the other cards. Why do you do that? If you're going to overclock one shouldn't you overclock the ones it's being pitted against? The 670 is battling against the 7950 because they are almost the same price and it would be very beneficial if everything was in one review instead of switching back and forth between reviews that have each card overclocked.

Click to expand...

It's been done
http://www.hardocp.com/article/2012/05/14/geforce_680_670_vs_radeon_7970_7950_gaming_perf/

great review!!! so good in fact it drove the price of the card up $70 on amazon!! wtf, i was going to buy this card today too.

Link goes to 504$ 670 GC card with no free shipping. Wow, I guess that didn't last...

Rahh said:

Link goes to 504$ 670 GC card with no free shipping. Wow, I guess that didn't last...

Click to expand...

Nah it's a private seller not Amazon. Private sellers gouge the fuck out of GPU prices on Amazon.

Derangel said:

Nah it's a private seller not Amazon. Private sellers gouge the fuck out of GPU prices on Amazon.

Click to expand...

Well, in their defense, Amazon does charge a pretty hefty commission for electronics sales (or at least they did, I haven't sold anything there recently).

Maverick1337 said:

I still don't understand [H] reviews sometimes. Your tests are good but they have one major flaw. When you compare Overclocking capability of the card against others, you don't overclock the other cards. Why do you do that? If you're going to overclock one shouldn't you overclock the ones it's being pitted against? The 670 is battling against the 7950 because they are almost the same price and it would be very beneficial if everything was in one review instead of switching back and forth between reviews that have each card overclocked.

Click to expand...

You should visit the GPU section of [H]ard|OCP more. You would have caught this review right away!


[ninja edit]

Stoly already covered this one.

Stoly said:

It's been done
http://www.hardocp.com/article/2012/05/14/geforce_680_670_vs_radeon_7970_7950_gaming_perf/

Click to expand...

I have seen all the reviews for all the cards and I know how thorough they are. I know this card is factory overclocked and they are going against reference models that are at stock speeds, which is completely understandable. It's when they manually overclock it and then compare it to the cards that they left at stock speeds. In the conclusion they talk about its great speeds and how it performs better than x,y,z well you damn right it did you overclocked it more. I don't think it's fair to draw conclusions based on manually overclocking one and leaving ones at stock speed like they say here: "The manually overclocked GALAXY GeForce GTX 670 GC was the fastest video card in this bunch, providing the best performance in every game. We think it is fair to say that GALAXY stepped up its game with the GALAXY GeForce GTX 670 GC, allowing it to match the performance of the GeForce GTX 680." Although it's true, yes it was faster and matched it but there should be a disclaimer/asterisk noting to people they pushed the 670 to achieve those speeds. It was not stock out of the box like that. And if they're going to push a 670 to those speeds, push the other cards to their fullest potential to do a correct apples to apples comparison.

Let's be honest, 90% of the time you'll be able to achieve factory overclocked settings on any given card. For people that are shopping for cards it makes things a lot easier. I don't have to get a card review from one thread, pit it against another thread with all at the settings of another review that had a different setup from both the first hard reviews. In my mind it gets confusing. I mean I take the time to look at every review but I am just wondering why all cards aren't overclocked if they're going to be overclocking one.

It just makes it easier to stroll into a Galaxy 670 GC thread and go, "oh they have a comparison at stock speeds with other cards, that's great. Oh goody an overlocked comparison! Now I can see if it beats an overclocked 7950! Oh...wait, I have to go to another thread to find that...Oh wait, the 7950 was overclocked on a different test setup. Oh wait, they were tested at 1600, not x1200 like this review. Hmmm, so is the 670 better or not?" Makes it more difficult to compare cards that way. Let's be honest, when we all see reviews we want to see how the perform against other cards so we know if we should purchase that specific card or another.

TL;DR: Manually overclocking one card and comparing it to stock-clocked cards (in the conclusion section mind you) is confusing for potential buyers. If you want to show a true comparison, put the data from a past [H] review that was done to make things simpler so users don't have to go searching through different threads to get the data they were expecting.

....and it was deactivated on newegg, and amazon no longer carries it.....