• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

GTX580 Overclocking Help

InvisiblePrawn

InvisiblePrawn

InvisiblePrawn

Associate
Joined
15 Jun 2012
Posts
200
Hi,

I bought my EVGA GTX580 SC in march along with the rest of my rig (from OCUK of course!) and have been using it at stock clocks since - and a very fine job it's been doing.

Recently, I read a couple of guides and began to overclock it using EVGA Precision X.

I started increasing memory clock in 50MHz increments, testing using EVGA OC Scanner as well as Furmark. I was doing 5 minutes of the Furry test. 5 minutes of the Tessy (Tesselation?) test in OC Scanner followed by 5 minutes of Furmark at full screen with 4x MSAA

I then started increasing core clock in 25MHz increments, testing similarly, but using no AA on Furmark (following the reccomendations of a guide).

I reached a maximum of 900MHz core and 2350MHz memory after the tests crashed the display driver and I reduced the core clock. Thinking all was well, I tested using Furmark for 15 minutes to make sure the clocks were stable.

Throughout testing, temperatures were around 70°C, maximum of 75°C during the Tessy test. I did not change the voltages or any other setting s on the card, either.

I then proceeded to play a game (Skyrim if it matters). Played fine for around 5 minutes before the screen glitched out and the display cut off. When it returned, I noticed the clocks had returned to stock. Similar things happened in other games (Crysis, Borderlands) but only when using the upped clocks. Even reducing the clocks to just above stock to around 825/2100 still caused the problem.

This obviously means that the clocks I set aren't stable. If not, how were they able to pass all the tests without fault? I was lead to belive that Furmark was just about the most taxing thing you could do to a card.

So, I would really appreciate any advice on my problem. Is my testing method flawed? Should I be testing just using games? Also, what kind of clocks have other people achieved on similar cards? Mine seemed about right from what I read on various forums - but if they aren't stable, what's the use!

Wow, I've been rambling. Sorry for the wall of text.
Any help would be great,
Thanks.
 
Hi mate, I don't know the answer to this and i will love to know the answer too as my oc 570 passes all stress tests but in bf3 it shuts on me, resulting in me having to lower the overclock a bit!
 
I started increasing memory clock in 50MHz increments, testing using EVGA OC Scanner as well as Furmark. I was doing 5 minutes of the Furry test. 5 minutes of the Tessy (Tesselation?) test in OC Scanner followed by 5 minutes of Furmark at full screen with 4x MSAA
Click to expand...

I think I remember reading in some reviews the GTX5xx series are designed to throttle back when running programs like Furmark. Try running something like unigine heaven 3 for a couple of laps to do a quick test of an oc.
 
Thanks for the replies guys.

The throttling behaviour would explain why the card only seems to crash in games. I will do some tests using Unigine Heaven to se if I can reach my real stable OC.

To anyone fammiliar with the program, what are the ideal settings for stability testing? I would imagine full screen with maximum settings would be the most stressful but I've been wrong before!

Also, how long of a test would be sufficient to test stability? The card would need to be stable for hours to be useable in games but would a 15-30 minute test be enough for quick stability testing?

Thanks again.
 
I've had several overclocks be benchmark stable (e.g. heaven or 3dmark or OCCT) and then crash in BF3... seems BF3 is a harsher task master than any of the benchmark / stress test and you have to notch down slightly to get back to fully stable
 
andybird123 said:
I've had several overclocks be benchmark stable (e.g. heaven or 3dmark or OCCT) and then crash in BF3... seems BF3 is a harsher task master than any of the benchmark / stress test and you have to notch down slightly to get back to fully stable
Click to expand...

Indeed... games are your ultimate test.

There's a nice Metro 2033 benchmark program floating about that gives a more reliable stability test in my opinion. Or as already stated, the heaven benchmark, or maybe 3Dmark11.

In my experience with my 580 gtx, to approach anything like 900 on the core, you will almost certainly need to increase the voltage to the GPU. Obviously depends on what your stock voltage is and how good an OCer your card is (not all cards are made equal... as they say).

Good luck...
 
I used the 196 nVidia drivers and Furmark to overclock my GTX580. The earlier drivers don't throttle.
Anything over 900MHz should indeed need a voltage increase. I have 1125mV for 925MHz and 1150mV for 950MHz
 
Thanks for all your suggestions guys.

Using Unigine Heaven and stock voltages, I reached 825/2250. These clocks were stable for around 2 hours of gameplay so I imagine they are fairly stable.

I would like to get just slightly more but am apprehensive about increasing voltages as I really don't know what's safe. What is a good conservative voltage that won't damage my card or void the warranty? Will 1.1 - 1.15v as mentioned be OK?

I've also changed the fan curve slightly to give the fans slightly more power as temps approach 70°C.

Thanks again.
 
InvisiblePrawn said:
Thanks for all your suggestions guys.

Using Unigine Heaven and stock voltages, I reached 825/2250. These clocks were stable for around 2 hours of gameplay so I imagine they are fairly stable.

I would like to get just slightly more but am apprehensive about increasing voltages as I really don't know what's safe. What is a good conservative voltage that won't damage my card or void the warranty? Will 1.1 - 1.15v as mentioned be OK?

I've also changed the fan curve slightly to give the fans slightly more power as temps approach 70°C.

Thanks again.
Click to expand...

Anything up to 1150mV is fine, it's the max that tweaking utilities will give by default. Just make sure you keep it cool, aim to keep it under 80c.

Here's my fan "curve" for reference.

fancurve.png
 
Personally when OC'ing I keep my cards below 70 degrees. They may be rated to go up to 92 degrees at stock clocks and volts but this does not mean these temps are OK with increased speed and voltage.

I fell into the Furmark trap when I built my rig in December. It is useless, just uninstall it and spit on the CD! All it does is extract the most power out of a card without replicating real life situations.

Your best bet for testing is to use whatever made you crash first, whether it be Heaven, Unigine, 3dmark11, Battlefield, Metro etc. BF3 is what I play the most and it's very GPU intensive so it's what I use.

Keep doing what you're doing by bringing up clocks in 20-25Mhz increments and then when it crashes add .005 or .010 to your voltage and test again. Rinse and repeat until you hit the wall or are happy with your result.

By hitting the wall I mean some chips cannot go past a certain voltage or clock speed regardless of what others may do. Since you're new to this do not enable unofficial OC'ing as you can blow the card.
 
I would agree if we were talking about the 6 series but on the 5 series the max thermal threshold is officially 105c and some stock cards under toasty environments will see the 90s. 5 series cards are good up to 80c and maybe more when overclocked.
 
most i have put through my 580s to hit 1ghz is 1.213 V but this is only for benchmarking and im under water never seen my cards go above 56C

but in normal use im running them at 950 core 2400 mem @ 1.150 V rock solid stable
idle temps 31-33C max out 51C
 
hlennie said:
most i have put through my 580s to hit 1ghz is 1.213 V but this is only for benchmarking and im under water never seen my cards go above 56C

but in normal use im running them at 950 core 2400 mem @ 1.150 V rock solid stable
idle temps 31-33C max out 51C
Click to expand...

Hey hlennie. I'm only on 4400 for the memory (1100), what voltage are you running the memory? .
 
I've now got 900/2315 stable at 1100mV which I'm extremely happy with. Temps maxed at 78°C after 2 hours of Skyrim with an average of around 75°C -well within limits.

Gaming performance is also noticeably better. I'll do some tests comparing stock clocks to my OC at some point - I'm having too much fun with it at the moment!

Thanks for all the help and advice guys.
 
Rossi~ said:
I would agree if we were talking about the 6 series but on the 5 series the max thermal threshold is officially 105c and some stock cards under toasty environments will see the 90s. 5 series cards are good up to 80c and maybe more when overclocked.
Click to expand...

If anything you should be keeping your OC temperature lower than your stock temp not vice versa.
 
Rossi~ said:
I disagree. Never once considered this, in the many years of overclocking.
Click to expand...

Surely in general the lower you can go, the better?

I generally just set the GPU fan as loud as I can tolerate it whether or not I'm overclocking the GPU or not :)

My Asus 580 Matrix is happy to sit there at sub 50 degrees when undervolted and playing non-demanding games ;)
 
You must log in or register to reply here.
Back
Top Bottom