GTX580 vs GTX670 for good min FPS

[xchaotic]

As many of you probably now by now, Max or avg FPS is not as good an indicator of playability and smoothness as min FPS.
I currently have GTX580. It is a power hog, for that reason alone, I want to upgrade to GTX670.
I currently game at 1920x1200 so it seems both cards will do fine, but I also plan to upgrade my screen to 2560x1440 before XMas.
At such resolution I expect both card to struggle at least a little with max or high settings in some games (which is fine by me - my GTX580 feels underutilised atm).
Now which of those two cards will offer a smoother gameplay, with less dips below 30fps in most games, Skyrim, Borderlands 2, BF3, next CoD etc?

 

[brandon888]

get 670 ... or even 680 if you have money 670 is better bang for buck for sure then 680 ^^ but on 680 you will get better minimum fps ....

 

[Jaydip]

Actually I would wait,670 might not be able to maintain a constant 30fps in games like Skyrim, Borderlands 2, BF3, next CoD etc @ 2560x1440 if you want to have all the eye candy

 

[felang]

Actually I would wait,670 might not be able to maintain a constant 30fps in games like Skyrim, Borderlands 2, BF3, next CoD etc @ 2560x1440 if you want to have all the eye candy

I´d bet real money that a 670 will run the next CoD at a minimum of 100 fps avg @ 1440p.

 

[Jaydip]

Yup you are spot on that one

 

[Durvelle27]

HD 7970 GHz in my opinion

 

[cmdrdredd]

Actually I would wait,670 might not be able to maintain a constant 30fps in games like Skyrim, Borderlands 2, BF3, next CoD etc @ 2560x1440 if you want to have all the eye candy

It does, I have tried it. If you mod skyrim with enb you kill the framerate anyway even on sli. Running fxaa injector with 4k textures is fine tho and I prefer that to enb mods.

 

[Jaydip]

It does, I have tried it. If you mod skyrim with enb you kill the framerate anyway even on sli. Running fxaa injector with 4k textures is fine tho and I prefer that to enb mods.

Thats great :thumbsup: the choice becomes much simpler then

 

[xchaotic]

Well, I plled the trigger on the GTX670 - doesn't make sense to spend more on something hypothetical...

 

[Jaydip]

great choice congrats

 

[djnsmith7]

Pretty sure you'll enjoy the card.

 

[rgallant]

another nv Koolaid drinker should have waited for the real upgrade of the 580
-like to see a 772 mhz 670 [locked no auto boost] oh sorry forgot you have no control over clock's to compare , funny that .vs a 580 772 mhz see if it is $400.00 faster . [I think = ish] balls out both + 20% 670 so @ 30 fps for the 580 , a 670 will do 36 fps and we know that will give 100% game play over the 580.
I did not vote . as the same crew will fight over a $20.00 difference in a card price as to which is the best card regard less of which one is better , but at other times tell op to spend $400.00 on a side grade. IMO>>

 

[TakeNoPrisoners]

The GTX 670 isn't really an upgrade from a GTX 580. No single card besides the GTX 690 would really be an upgrade.

 

[boxleitnerb]

min fps is no good indicator either. No single value ever is.

I would like to see something like FEAR introduced:

25% fps above 60
25% fps between 30 and 60
50% fps below 30

You can make a neat traffic light diagram out of it. Green, yellow, red. The more green, the better.

 

[toyota]

The GTX 670 isn't really an upgrade from a GTX 580. No single card besides the GTX 690 would really be an upgrade.

well in most games it will only be about a 25-30% increase but it will be better than that in BF3 though and its huge upgrade for Clear Sky.

I dont think the gtx670 will make a big enough impact on minimum frame rates in most cases though.

 

[toyota]

min fps is no good indicator either. No single value ever is.

I would like to see something like FEAR introduced:

25% fps above 60
25% fps between 30 and 60
50% fps below 30

You can make a neat traffic light diagram out of it. Green, yellow, red. The more green, the better.

well he is clearly meaning a consistently higher minimum framerate.

 

[moonbogg]

I thought benchmarks show a 670 being almost as fast as a 590. 670 is a good deal better, just maybe not a night and day difference. Maybe when Nvidia releases a real card you will see the difference you are looking for.

 

[imaheadcase]

Just get a 2nd 580 for SLI. problem solved. You won't get much more performance going from 580 to 670/80 ath that rez.

 

[boxleitnerb]

well he is clearly meaning a consistently higher minimum framerate.

"Consistently" across multiple settings and games doesn't change the fact that it is still only one value. Imagine playing for an hour and recording the framerate every second. What is the minimum fps? The lowest value during that hour. Minimum fps will only tell you about gameplay experience for a miniscule part of that hour, maybe a minute at most, so its usefulness is quite limited.

 

[toyota]

"Consistently" across multiple settings and games doesn't change the fact that it is still only one value. Imagine playing for an hour and recording the framerate every second. What is the minimum fps? The lowest value during that hour. Minimum fps will only tell you about gameplay experience for a miniscule part of that hour, maybe a minute at most, so its usefulness is quite limited.

if a game is consistently above 50 or 60 fps then that is much better than seeing lots of framerates in the 70s and 80s with tons of minimum drops into 40s and 30s. sometimes there may be a huge area where one card will struggle yet another one wont. I am not talking about one or two dips here so again comparing minimum framerates is very useful.

if you still can not understand that then I will give an example. in Metro 2033 there might be some settings where I potentially average 85 fps yet the game is NOT playable with vsync on and even a litle choppy with it off because there are just way too many spots that dip into the 40s and even 30s. you cant tell that by just looking at the average so you need a graph to see how much the minimum framerates were actually dropping.

 

[moonbogg]

Minimum FPS is sometimes due to CPU also. Just depends.

 

[Arkadrel]

Which card is better for 2560x1440?

The real answear is the 7970 1ghz edition though (since that wasnt a option voted 670)
3GB of vram can come in handy when you play at those resolutions, and want to use high levels of AA:

 

[boxleitnerb]

if a game is consistently above 50 or 60 fps then that is much better than seeing lots of framerates in the 70s and 80s with tons of minimum drops into 40s and 30s. sometimes there may be a huge area where one card will struggle yet another one wont. I am not talking about one or two dips here so again comparing minimum framerates is very useful.

if you still can not understand that then I will give an example. in Metro 2033 there might be some settings where I potentially average 85 fps yet the game is NOT playable with vsync on and even a litle choppy with it off because there are just way too many spots that dip into the 40s and even 30s. you cant tell that by just looking at the average so you need a graph to see how much the minimum framerates were actually dropping.

You just proved my point, thank you. You're talking about a graph - that is something else entirely because it conveys much much more information, namely how many dips there are, how strong they are and how long.

With 3 values (min/avg/high) you don't know all that. You just know, that for 1 second only the fps dropped to the minimum value. What else happened in the benchmark beyond that is pure speculation. That's why a minimum fps value is useless and not helpful - a graph on the other hand very much is.

 

[toyota]

well a graph is what I was referring to the whole time. there is no other way to show consistent framerates.

 

[boxleitnerb]

Fair enough, that was not clear. It is just irritating when min values alone are presented as the holy grail, because they aren't.