[funxion]

Hey guys,

So right now I'm sitting with a GTX 560 Ti Classified 448-Core that I've had for about a year and a few months. I used to run it at stock settings (797/1594/1950 @ 1.013V), but I decided to overclock it about 3-4 months ago. After toying around with overclocks, I found that 880/1760/2150 @ 1.025V was stable and beneficial enough to keep.

Obviously when you overclock and concurrently increase voltage, temps will be higher. My temps are at 34-35 idle and anywhere from 45-55 in-game (CoD4, Black Ops II, PlanetSide 2, BF3).

My buddy, who has a GTX 580 Twin Frozr, says he idles at 50.

Are these temps okay to play at? Am I not pushing my card to its max? I also have my fan curve set to this.

I'd love to get some input, so let me know what you all think.

 

[MrTOOSHORT]

Voltage, overclock and temperature are all very good. Keep load temps below 80'C is the set rule for gpus.

 

[funxion]

Quote:

Originally Posted by MrTOOSHORT

Voltage, overclock and temperature are all very good. Keep load temps below 80'C is the set rule for gpus.

Would going up to 900/1800/2200 be better? Just trying to get the most out of this card without breaking it.

 

[MrTOOSHORT]

Quote:

Originally Posted by funxion

Would going up to 900/1800/2200 be better? Just trying to get the most out of this card without breaking it.

If the temps are good, 1.1v is no problem for that beefy card. What can 1.1v get you? That's where I'd leave the overclock at.

 

[funxion]

Quote:

Originally Posted by MrTOOSHORT

If the temps are good, 1.1v is no problem for that beefy card. What can 1.1v get you? That's where I'd leave the overclock at.

Are you sure that 1.1V is safe? I'm only at 1.025V, haha!

 

[juano]

I will second that 1.1v is reasonably safe voltage to run your card at, and honestly I only say "reasonably safe" instead of "perfectly safe" because anything can happen and there is always some risk to overvolting.

 

[MrTOOSHORT]

1.1v is safe for that card, doesn't mean you need to go that high. Seems like you have a great overclocker on your hands. Maybe just go 1.05v instead so the card is still quiet and cool. I run 1.113v on my Asus DCii 570 for 900Mhz core to give you an example.

 

[funxion]

Quote:

Originally Posted by juano

I will second that 1.1v is reasonably safe voltage to run your card at, and honestly I only say "reasonably safe" instead of "perfectly safe" because anything can happen and there is always some risk to overvolting.

Thank you for chiming in! It's good to get a second opinion, not that MrTOOSHORT's was good to begin with, but you know what I mean. I agree.
Quote:

Originally Posted by MrTOOSHORT

1.1v is safe for that card, doesn't mean you need to go that high. Seems like you have a great overclocker on your hands. Maybe just go 1.05v instead so the card is still quiet and cool. I run 1.113v on my Asus DCii 570 for 900Mhz core to give you an example.

I will start moving up slowly, but my fan already gets pretty loud with my current fan curve setup. What do you think is a good max idle temp to hit? It's also probably good to mention that I'm on stock cooling (although EVGA's stock cooling on my model is actually very good).

 

[MrTOOSHORT]

Don't worry about idle temps too much, load is where you look. Like I said, under 80'C load and you'll be set.

 

[juano]

Idle temp is almost meaningless, try to stay away from the 80s on your load temp and that will be great. FYI your chip is specified up to 97C by Nvidia, I personally want to stay 10C from the Nvidia limit at all times no matter what, and then I like to have another 5C for comfort and wiggle room if the case gets dusty or the house gets hot etc. So if I was personally OCing your card I would stop raising voltage when I hit 80-82C load temps and then see what the max core clock that voltage will get you.

 

[funxion]

Quote:

Originally Posted by MrTOOSHORT

Don't worry about idle temps too much, load is where you look. Like I said, under 80'C load and you'll be set.

Quote:

Originally Posted by juano

Idle temp is almost meaningless, try to stay away from the 80s on your load temp and that will be great. FYI your card is specified up to 97C by Nvidia, I personally want to stay 10C from the Nvidia limit at all times no matter what, and then I like to have another 5C for comfort and wiggle room if the case gets dusty or the house gets hot etc. So if I was personally OCing your card I would stop raising voltage and call it a day when I hit 80-82C.

Alrighty gentlemen, I'll report back with where I end up!

 

[funxion]

Quote:

Originally Posted by juano

Idle temp is almost meaningless, try to stay away from the 80s on your load temp and that will be great. FYI your chip is specified up to 97C by Nvidia, I personally want to stay 10C from the Nvidia limit at all times no matter what, and then I like to have another 5C for comfort and wiggle room if the case gets dusty or the house gets hot etc. So if I was personally OCing your card I would stop raising voltage when I hit 80-82C load temps and then see what the max core clock that voltage will get you.

This is my card.

Oh, and how high should memory clock be going?

 

[juano]

Quote:

Originally Posted by funxion

This is my card.

Oh, and how high should shader clock be going?

Approximately twice as high as your core clock.

LOL I was just pulling your leg, it's exactly double the core clock, ZOMG I'm so funny!

No but seriously on the Fermi architecture the shader clock is always locked to two times the core clock.

 

[funxion]

Quote:

Originally Posted by juano

Approximately twice as high as your core clock.

LOL I was just pulling your leg, it's exactly double the core clock, ZOMG I'm so funny!

No but seriously on the Fermi architecture the shader clock is always locked to two times the core clock.

Oh man...I'm an idiot. I meant memory. This is why I shouldn't be doing this when I first wake up lol.

 

[juano]

I don't bother OCing memory on Fermi because of how little difference it makes to performance as well as how difficult it is to actually ascertain stability. The VRAM on both of our cards has error correcting code (ECC) which means that it typically will not shows instability errors in the expected manner of artifacts which are easy to spot, but will instead fix them itself but at the cost of performance. So to test stability of the ECC VRAM you need to get multiple baselines of benchmarks at stock memory speed then repeat each of those benchmarks every time you increase the speed and you need to see a improvement in the scores, when you start to see scores falling instead of rising that will be the only indication you've reached instability of the VRAM in that specific benchmark. Basically it's a whole lot of work where you' might not even be 100% positive it's actually stable for not much gain, and I've never felt it's worth the trouble.

 

[funxion]

Results
797/1594/1950 @ 1.013V | Idle: 32°C | Load: 62°C | Stable
880/1760/2150 @ 1.025V | Idle: 34°C | Load: 68°C | Stable
905/1810/2150 @ 1.025V | Idle: 35°C | Load: 71°C | Stable
910/1820/2150 @ 1.025V | Idle: 35°C | Load: 71°C | Stable
920/1840/2150 @ 1.025V | Idle: 35°C | Load: 72°C | Stable
925/1850/2150 @ 1.025V | Idle: 35°C | Load: 72°C | Unstable (1 artifact)
930/1860/2150 @ 1.025V | Idle: 35°C | Load: 72°C | Unstable (8 artifacts)
930/1860/2150 @ 1.038V | Idle: 33°C | Load: 70°C | Stable
935/1870/2150 @ 1.038V | Idle: 33°C | Load: 70°C | Unstable (1 artifact)
935/1870/2150 @ 1.050V | Idle: 34°C | Load: 70°C | Stable
940/1880/2150 @ 1.050V | Idle: 35°C | Load: 72°C | Stable
950/1900/2150 @ 1.050V | Idle: 35°C | Load: 71°C | Unstable (1 artifact)

The bolded OC is what I'll be using as my 24/7 OC. I don't really do competitive overclocking, so that's good enough for me. A 50MHz increase on my core clock with the same or lower idle and load temps. This card seems to like 1.038V. Anything higher is just pushing it.

 

[juano]

What have you been using to test stability? If it's something like OCCT, kombuster, or Furmark those are poor tests for the 500 series because of the fact that the clocks will throttle down to 405MHz intermittently in those tests despite what MSI AB or GPU-z may be telling you. Source 1, source 2.

I'd recommend multiple tests to determine stability, for example Unigine Heaven, 3d Mark, and multiple games.

 

[funxion]

Quote:

Originally Posted by juano

What have you been using to test stability? If it's something like OCCT, kombuster, or Furmark those are poor tests for the 500 series because of the fact that the clocks will throttle down to 405MHz intermittently in those tests despite what MSI AB or GPU-z may be telling you. Source 1, source 2.

I'd recommend multiple tests to determine stability, for example Unigine Heaven, 3d Mark, and multiple games.

I use EVGA OC Scanner X. It runs on full load. Obviously it won't detect all artifacts, so I just play GPU intensive games to determine how stable the OC truly is. But if it's getting artifacts in OC Scanner X, I rule it out immediately.

 

[dbthump]

I have the same card, MSI version, modded with the Antec 620. I've been running 950/2100 for a couple weeks now at +150mV. I pretty much only play BF3 at 1080, mostly high settings, capped at 80fps, and my temps are never above 47*. I ran Heaven at maxed out for an hour without any issues.

My card has a nice heat spreader covering the ram and the vram, so of course I left that on. I made a bracket, like dwood's, that holds a 90mm fan and dumps right onto the vrams. In a week I'll be adding 14mm cooper heat sinks onto the heat speader right over the vrams.