Why does my reference GTX780 run a lot cooler than Tom's says it runs? Prioritize Temps or Power?

ericjohn004

Honorable
Oct 26, 2012
651
0
11,010
I just got my new GTX 780. It's an EVGA GTX780 SuperClocked, with the awesome aluminum reference design cooler. Stock, this thing runs at 941Mhz-993Mhz Boost, with 1502Mhz Memory. I have it at 1041Mhz-1093Mhz Boost Clock, with 1752Mhz Memory in EVGA Precision X. Which is a fairly good overclock for the GTX 780 without changing the voltage at all. It's at +100Mhz on the core and +500Mhz on the memory. I can go as high as +150 on the core and +650 on the memory without changing the voltage at all. If I go higher than that I'll have artifacts. I'm also running it on a 1920x1080p 120Hz BenQ display and I have a 3570k@4.7Ghz with an H100i cooler and a Gigabyte UD5H Motherboard with an 850w PSU with an 8 pin and 6 pin connector on the GTX780. You can read all of my specs if you hover over my picture.

Now, In the Tom's Hardware review they point out how with the reference design cooler, at stock clocks, their GTX780 get's up to the 83c temp target and starts throttling down a bit. That NEVER happens with my card. With the overclock I have, my card never even get's above 69c during an EVGA OC Scanner "Furry Donut" stress test benchmark at 1920x1080p with 8x MSAA. That's with running it for like 2 hours. It's the most stressful benchmark I have. And during gaming it'll never go above 67c, and that's while playing Crysis 3 at the highest settings possible. Even during Heaven 4.0 it never goes above 67c.

So I'm wondering why in the Tom's Hardware review of the GTX 780 they said it got up to 83c and throttled down? And that's just with stock settings. I'm overclocked and getting 14c lower temps than they are, and they are throttling while I'm not. That doesn't seem possible. I'm also getting GTX780 performance, so my card isn't messed up. My temps were taken with my house at around 74c. While I think they had their room at 78c.

I also have a couple more questions.......

Should I prioritize Temp. or Power, in EVGA Precision X?

And should I link them together?

I've tried all the different ways while I was an hour into a GPU burn, and nothing made any difference at all. The only thing that makes a difference is when I set the power from 100% to 106%. That makes a difference. But if I change the temp from 75c to 96c, I see no difference. If I link or de-link the temp and power, I see no difference. That's probably cause my card is so cool though. I've read somewhere that you should set it to prioritize Temps over Power to get more performance, so that's what I do, although it makes no difference.

Also, my GPU usually always runs at a clock that's higher than my overclock. I see it running at 1163Mhz a lot. And I only have it at 1093Mhz Boost clock. I'm guessing that's cause I'm at such a low temp and have plenty room for GPU Boost 2.0 to do it's thing.

Boy, I'm glad I don't have a 290 or 290x, or heat would actually be a problem. This card is so awesome, I highly recommend a GTX770, 780, or 780Ti over the R9 290 and 290x. Those reference Nvidia coolers make a world of difference, and they look a lot better too.

So let me know what you think, thanks for reading all of this. I try to be as specific as I can so you won't have to ask any questions.
 
Solution
normal temps for evga 780 sc... i have few friends with them

all play on 1440p, so they are heavily overclocked, rare to see them hit 70
Furmark doesn't work as a stress test because of driver/bios restrictions. There may have been a BIOS revision to the reference 780 to make the fan profile more aggressive, hence your lower temperatures, that or you live in Antarctica and have an excellent case.

Anyway to overclock you want to get rid of all the stupid restrictions. That means maxing out the temperature target and prioritizing it (easier to hit a power limit wall than it is a temperature one), max out the voltage and power target, do not link temp and power limit. Then just see how far you can get. Typical 780 overclock is +500 on the memory and 1100Mhz core, if you are already at 1093Mhz with no voltage changes you probably have a pretty decent overclocker.
 

ericjohn004

Honorable
Oct 26, 2012
651
0
11,010
Ok cool. I didn't use Furmark by the way cooky. I used EVGA OC Scanner. And I also get the same temps with MSI Graphics test, and Furmark. All using the same "Donut" Furry test. They all go no higher than 69c.

And currently I have the temp/power unlinked, and I prioritize temp. But what I'm saying is that, if I change it to power, or even if I go from 96c down to 70c with my temp target, it makes NO difference whatsoever. I'm guessing it makes no difference because my card already runs so cool that allowing it extra "room" isn't even needed.

And yeah, I think I do have a good overclocker. I get up to 1163Mhz most of the time while playing games. It only becomes unstable when I go past +150Mhz, and my Memory gets unstable going past +600Mhz. In GPU-Z it says my ASIC quality is almost 80%, it's like 79.3%. Whereas my older 660Ti which couldn't overclock at all was at 62%. So the ASIC quality might be some indication of overclocking prowess, MAYBE.

Thanks for the answers guys, any idea why I see no difference between prioritizing power over temp or temp over power?