GTX 680 GPU Boost levels - what's yours?

[Kaotika]

So Nvidia stated, categorically, the average (and guaranteed) increase you can expect is 1058Mhz, GPU Boost is also limited to 1110Mhz on every card anyway, so no review cards went over that speed. Also the performance difference between the guaranteed Boost clocks (1058) and the maximum (1110) is less than 2%. Negligible.

Actually the difference should be more than 2%, as the theoretical difference is 4.9% and even at 1084 (2.4% theoretical) the average performance difference is already 1.5% according to hardware.fr's benches.

I use Afterburner, latest edition that has full support for the 680. Apparently, when the cards are under heavy load, they don't fluctuate the clocks at all, it finds a happy medium and sticks to it. Anands review gives a very good rundown of what's going on, pretty much matches up what's going on with mine, as far as i can tell anyway. As i've said before, heat plays a crucial role, if the GPU is over 70c it simply will not hit the maximum Boost of 1110, it scales the clocks according to load, heat and power consumption, although never dropping below 1006Mhz.

So people should use Afterburner to find where their card sits at then I suppose (oh, and it can drop under 1006MHz, too, for example at TPU had their press sample to drop under)

[QuadDamage]

At 220+ GPU boost I get 1344Mhz. Do the math.

[kitch9]

Nvidia was very transparent with the information regarding GPU Boost, from the reviewers guide:



So Nvidia stated, categorically, the average (and guaranteed) increase you can expect is 1058Mhz, GPU Boost is also limited to 1110Mhz on every card anyway, so no review cards went over that speed. Also the performance difference between the guaranteed Boost clocks (1058) and the maximum (1110) is less than 2%. Negligible.



I use Afterburner, latest edition that has full support for the 680. Apparently, when the cards are under heavy load, they don't fluctuate the clocks at all, it finds a happy medium and sticks to it. Anands review gives a very good rundown of what's going on, pretty much matches up what's going on with mine, as far as i can tell anyway. As i've said before, heat plays a crucial role, if the GPU is over 70c it simply will not hit the maximum Boost of 1110, it scales the clocks according to load, heat and power consumption, although never dropping below 1006Mhz.

This is the thing I'm sure I read reviews that stated their cards went higher than those you describe.... Mine goes to 80c before the fan starts to really ramp and will boost to 1250mhz quite happily at 80c, with no sign of throttling. (Overclocked settings.)

[icemanchilled]

Actually the difference should be more than 2%, as the theoretical difference is 4.9% and even at 1084 (2.4% theoretical) the average performance difference is already 1.5% according to hardware.fr's benches.

Fair enough, but that means it's still only between 2 and 3% at the most. Regardless, the difference isn't large enough to matter, would it matter if the 7970 was 2-3% faster than the 680 or vice versa? No, it's close enough to simply not matter.

So people should use Afterburner to find where their card sits at then I suppose (oh, and it can drop under 1006MHz, too, for example at TPU had their press sample to drop under)

I should have clarified, it won't drop below 1006 when gaming, it may do so in stress tests to keep your card from frying, which is a good thing of course.

This is the thing I'm sure I read reviews that stated their cards went higher than those you describe.... Mine goes to 80c before the fan starts to really ramp and will boost to 1250mhz quite happily at 80c, with no sign of throttling. (Overclocked settings.)

That's because it's overclocked, then Boost simply adds it's increase to whatever you added. Left stock, with no manual overclock, Boost will max out @1110Mhz.

[QUAKERVAL]

That's because it's overclocked, then Boost simply adds it's increase to whatever you added. Left stock, with no manual overclock, Boost will max out @1110Mhz.

I run 2 different benches(UNIGINE HEAVEN and FLUIDMARK) and both showed boost max out @1124(using PRECISION X)..

[QuadDamage]

I run 2 different benches(UNIGINE HEAVEN and FLUIDMARK) and both showed boost max out @1124(using PRECISION X)..

My boost is also higher than 1110.

edit : Actually it's 1124 just like yours, Val.

[Kaotika]

Both of yours are at stock clocks & settings (no extra power, oc's etc)?

[Gordo]

From watching the behaviour of mine I can see how it would be very easy of Nvidia to tweak the performance of binned cards to send to reviewers whilst the rest of us mere mortals get what we are given and it would be very difficult to quantify any difference without knowing how the reviewers cards boosted and exactly when.

This "feature" has murkied the waters somewhat, you can pay as much as the next man for a product and actually end up with one the performs substantially worse out of the box and there's no real way of knowing if that is the case.

I'm not a massive fan of it.

And that's different from the 'normal' manual overclocking of video cards that has been done in the past in exactly what way?

[Kaotika]

And that's different from the 'normal' manual overclocking of video cards that has been done in the past in exactly what way?

The difference is that the boost clock is in use in all the reviews, but you can't know you'll get as high boost clock when you buy one from a shop. Manual OC is and has always been a draw of luck, in reviews too. But this doesn't really belong to this thread, which purpose was just to collect data from 680 owners

[QUAKERVAL]

Both of yours are at stock clocks & settings (no extra power, oc's etc)?

yes everything @stock(100% power and 0 on GPU and MEM.) is there more then 1 revision of this card? if yes mine is A2.