Different results on both MSi Gtx 680 Twin Frozr

[litwicki22]

Hi all. I just bought 2 x Msi Gtx 680 Twin Frozr, and i tested single card on 3dmark 11.
On first card i have more different variations fps between tests. And on second card i have smaller variations fps between tests.
I run few times 3dmark 11 on both cards. It is normal or not? Why on first card i have more different variations on fps between the same GT tests?


1 card:

And now second card ( look closely at variations in fps between the same GT tests, they are just smaller ):

RMA that first card? Or what now?


Why i on first card i have more fps variations between the same GT on next loops? For example: 9458 to 9399 and 9438 to 9396 scores.

I mean card is working properly?

 

[notty22]

The scores are within a few percent of each other. That is absolutely normal.

 

[boxleitnerb]

Normal turbo variation. It's normal.

 

[litwicki22]

Yes but why on second card i have a smaller variations between GT fps and score? On second card scores are just closest

While on first card i just have more differents in score and fps, like that: 9458 to 9399 and 9438 to 9396 scores.

 

[notty22]

Even without turbo, I rarely got the exact same score, almost never. Same with cpu benchmarks. You can run them consecutively , large or small. There is often a slight variation.

edit : I cringe that you would consider RMA'ing for this reason.

 

[litwicki22]

OK but i am asking about differences on scores in first card between loops. While on first card i just have more differents in score and fps, like that: 9458 to 9399 and 9438 to 9396 scores.


While on second card are :9417 ->9427 --> 9414 --> 9438 -->9428

 

[Puppies04]

Dude it is really nothing to worry about. Firstly unless you run several dozen tests you have no way to know if the tests would eventually end up with very similar results on both cards. Also, as already stated the difference between your lowest and highest is just over 0.5% which in the world of benchmarks is about 8 times smaller than a "normal" margin of error.

 

[lavaheadache]

Jesus some people just worry way too much. Rma because you have less than a .5% variation in 3dmark? Seriously?

 

[The_Golden_Man]

litwicki22

LoL! Just relax! There is nothing wrong with your cards dude! I have two ASUS GTX 670 Direct CU II. My upper card run a higher turbo VS my lower card. Also my upper card overclocks like a dream, while my lower one overclocks like shit.

But hey, even my lowest quality card still has a much higher turbo clock then what is advertised. So cannot complain.

Also, 0.5% is far lower than the margin of error in benchmarks.

 

[fastamdman]

As everyone else stated it's not the card. Hell at 0.5% it could have been a program loading in the background. Whether it was a temperture monitoring program taking up to much power, a virus protection loading, who knows. If you are 100% sure everything was off in the background, it's still a reasonable margin of error. It's normal for cards to have good runs and bad runs because you are being judged on soooo many things.

Try running several other tests on both the cards, check there temps, check the gpu usage etc and make sure they are both within reason. Currently, you have nothing to stress. Go do some gaming and make sure they are both stable without driver crashes

Then onto overclocking!

 

[RussianSensation]

1-2% variation in benchmarks is generally within the expected standard deviation/margin of error.

 

[guskline]

There's a great song titled " Don't worry, Be Happy!" That surely applies here. I'd be overjoyed to have your "slower" card whichever one it might be!()

 

[cmdrdredd]

OK but i am asking about differences on scores in first card between loops. While on first card i just have more differents in score and fps, like that: 9458 to 9399 and 9438 to 9396 scores.


While on second card are :9417 ->9427 --> 9414 --> 9438 -->9428

I have two identical cards. One will boost higher by default and not throttle as quickly by default (even before overclocking). It's how these GPUs work. Don't waste time RMAing the card, it's 100% fine. I can loop heaven all day and one card will throttle at 70c and boost lower while the other card never touches 70c and boosts higher. You'd think the higher boost clock would cause that card to get hotter and throttle, not always so.

 

[aaksheytalwar]

It is due to Kepler boost.

 

[chimaxi83]

Yep, thanks to "GPU Boost", no two cards will be alike.

 

[litwicki22]

So if first card have lower boost and throttling faster with scores between "9458 to 9399 and 9438 to 9396 scores" is not bad and dont RMA?

While second card scoring are: "9417 ->9427 --> 9414 --> 9438 -->9428" ( smaller differences )

 

[DigitalWolf]

So if first card have lower boost and throttling faster with scores between "9458 to 9399 and 9438 to 9396 scores" is not bad and dont RMA?

While second card scoring are: "9417 ->9427 --> 9414 --> 9438 -->9428" ( smaller differences )

Well think of it this way. What exactly is going to be your reason for the RMA? Because if they test the card its working fine... Telling them its slightly slower on some benchmark is not a valid reason to RMA a card.

 

[litwicki22]

Yeah but if i got that more variant scores between loops maybe card is overheating?

 

[cmdrdredd]

So if first card have lower boost and throttling faster with scores between "9458 to 9399 and 9438 to 9396 scores" is not bad and dont RMA?

While second card scoring are: "9417 ->9427 --> 9414 --> 9438 -->9428" ( smaller differences )

It's like taking two identical CPUs. One CPU runs at 70c max and the other will go to 74c max. You cannot predict that and both are within design specifications.

No two GPUs are ever identical in every way.

The card is not overheating. If it were you'd crash, get artifacts, or it would give driver errors, even slow down to 2D clocks. It isn't doing that so the card is fine.

 

[Wall Street]

Sorry but you need to do the math yourself and understand how little difference there is. I think that you are just looking at the absolute number difference and not the percentage. The fastest result (9458) is 0.66% faster than slowest (9396).

For an example of how close this is, divide each score by 200 and assume that was a framerate. If one benchmark run got 47.3 FPS and another got 47.0 FPS, would you complain about this? They both round to 47 FPS, so you likely wouldn't care. Yet this is the same difference between these high and low test runs.

In anything, 3DMark shouldn't give four-digit and five-digit scores because they tests really don't measure to that accurate a level.

 

[litwicki22]

I mean scores on my first card,like : 9458 to 9399 then 9438 to 9396 , then 9436 to 9391 and 9391 to 9442 are OK or not good ? I mean that variant score points differences.


While on second card i have smaller differences ,see :


9417 ->9427 --> 9414 --> 9438 -->9428


What do you think about first card?

 

[razorhawx]

hey guys i got a asus m5a99x evo can i run a gtx 680 msi twin frozr on it.....i am actually planning to buy it this week any answers is highly appreciated...
its got 2 2.0 x 16 pcie slots(x16)
and 1 2.0 x 16(x 4 ) pcie slot
i currently own a 6870...since 680 requires gen 3 i was having doubts