Overclocking the 6800GS by EVGA.

[vtohthree]

Hello, I recently purchased a 6800GS(dec 05'), I kind of regret it because of all these newer cards coming out. I wanted to get a 7800GT CO, but I figure I'd save my money, I was going to get a 6600GT, wait it out then upgrade, but instead I went for the middle guy, the 6800GS.

In reality, I'm probably stuck with it(nothing will probably change significantly in price by february for me to "step-up" to the next card). So I figure I mind as well overclock it, and get as many fps safely as possible.

Again, I already did a search for this topic so please refrain from pointing me to ones like these:

http://forums.anandtech.com/messageview...atid=31&threadid=1751866&enterthread=y

...which started out exciting but closed prematurely. Sorry didn't mean to bring that thread out, but yeah I'm interested in some decent overclocking with my 6800GS, so I'd llike to hear from some experienced people about the subject.

I also read through this article:

http://www.xbitlabs.com/articles/video/display/geforce-6800gs_9.html

(6800GS oc'd to 500/1360)..some interesting results, but didn't blow me away.

I just wanted to reiterate my goals, if feasible:
*highest possible fps results, yet safe/stable/practical enough for everyday
*if possible, on stock cooler(I have a 120mm case fan blowing towards it, might help slightly)

So without further a due, let the discussion begin!

 

[zagood]

I don't own a GS, but basic VGA logic is that you'll get much better overclocking performance with a good aftermarket heatsink. That's just the way it is. 10c can make a huge difference when overclocking.

If you're looking for scores above that 500/1360 you're 99% going to have to invest.

My 2c.

-z

 

[mindless1]

Well there's no real secrets here, just keep ramping up the clocks without ceasing until you start seeing instability, all the while watching temps. You won't know what the card's ceiling is till you hit it.

So far as practical for everyday, depends on what everyday means. You can of course keep a lower 2D speed, and after you'd been benching or gaming with the card be sure to check it's temps. If you have a thermal sensor that's great but touch-testing the power fets and capacitors is a very good start, expecting the caps to (need) be quite a bit cooler, no more than mildly warmish for best lifespan.

The prior suggestion of a better heatsink might be necessary for best results but you'll have to try the card and see how it goes. I don't know what it's fan noise is like but fan noise is a common reason I'd replace a 'sink, but then I do a bit of metalwork as a hobby and may come up with something myself rather than 3rd party 'sinks, plus I avoid the passive 'sinks because they really aren't very good for long-term use, rather than only keeping card cool enough not to crash in the short term. That is, unless you have a fan on a bracket pointed at the card too, not just a front intake fan unless it's fairly strong (and loud, which I also avoid).

It depends what you need though, you might be happy with the performance as-is, but you have the benefit of others' o'c scores as a guide so have fun with it.

 

[vtohthree]

Give me a little bit of time, I'm going to report results.

 

[rmed64]

Dont know about the Evga 6800GS, but I have a Gigabyte version. I hit a ceiling at 495 on the core and I have 1200 on the memory. At 500 core I see some artifacting, so I keep it at 490

 

[russki]

I bought the XFX GS clocked at 485/1100 and a Zalman V700 or whatever its called copper cooler. I can easily reach 515/1300 and have benched it at 540/1350.. YMMV

 

[vtohthree]

Well, I'm still "tuning" to find a good spot, but with the stock cooler and heatsink fan on there i was able to get 535mhzgpu/1.22ghzram for detecting the optimal frequencies.

My idle temps are at 41*C. And my load temps(running rthdrible), slowly got up to 70*C after 5 minutes, and I'm assuming it probably would have kept climbing(slowly) before I stopped the program. So I'm going to tone it down a bit to be on the safe side. What temps should I begin to worry at?

Any suggestions?

edit: I'm running at 77*C, while running rthdrible. I can put my finger on the capicitor, its pretty hot, but not enough to scald me(except the one closest to the chipset fan on the mobo), current setting for this is only 485/1.11.

 

[mindless1]

There really isn't much point to the "detecting optimal frequencies part", you're really best off never clicking that button and completely ignoring whatever it suggested.

If one flashes another bios to their card, for example, that alone can change what the supposed "detected frequency" is. I'd be very interested to know exactly what that button really does but it seems to have as much to do with product-tiering as overclocking.

Sometimes that button will show a given frequency that isn't stable later as card gets hotter, or other times there's a bunch of headroom left.

70C is not too hot, going higher is of course dependant on how much higher but in that range it's more a matter of how the increase in temp effects a stability threshold. IE- the faster the card is running, the less stability margin at any "n" MHz speed. More important is the capacitor temp mentioned earlier, the GPU can run 24/7 in a safe zone for years but caps, for each 10C temp rise their lifespan is cut in half. Fortunately the initial lifespan is usually several years so 1/4 lifespan may not effect use within the useful gaming life of the card, and if 3 years from now the card can't o'c to the last few %, no big deal as it was easy enought o'c and easy enough to o'c a little less.


Unless you'll be continually cleaning out your heatsink, leave at least 5-10C between your max and your max-comfort-zone to allow for dust interfering with best airflow. You will need to leave the card running for hours though to get the top temp, presuming 70C then "might" rise higher isnt'really determinant. In some games, it may not even get that hot, depends on what's being stressed.

Even if you want to keep using the stock 'sink you might get a few degrees improvement through basic things such as pulling it off, cleaning off all original thermal compound from GPU and 'sink, lapping the 'sink some and rubbing some thermal compound deep into the 'sink base and carefully remounting it. Sometimes it may not seem to effect overall temps much but if the GPU wasn't making same degree of contact with entire 'sink, the temp report may not reflect it. Further, use of good quality synthetic based compound (such as Arctic Silver or Alumina, or others), is expected to provide longer lasting interface than any/all silicone based grease.

Generally speaking, for some test runs you should just keep going, push the core higher till you find the max o'c regardless of temp. If you're already pretty close to that now it will seem unnecessary to get a different 'sink, but if there's a whole lot of room left (if only it were cool enough), you have a better gauge on investment return if you were to buy an aftermarket 'sink. If it's wintertime there, an easy test would be moving the system over towards a cracked window so ambient temp is lower- that's expected to directly effect resultant GPU temp so you could see what kind of gains, say 10 or 20C lower might yield.

 

[vtohthree]

Thanks for that informative post mindless1. Yes, I edited my post before, and I can feel the capicitors at a scalding level, definately not desirable. This has "scared" me into less extreme overclocking with my gpu for now, it's currently clocked at a frugal 451/1.09, load temps seem to reach around 77*C now no matter what. When it was sitting for nearly 20minutes on much higher settings(around 500 +/-) it seemed to have "peaked" out as well, again this is only brief observation, not 24/7.

I can really see the need of good airflow on this card. From the heatseank to the capacitors on this card, the air flow only becomes a faint drift that blows hot/warm air onto the capacitors, can't be good. What I need to do is put a 80mm fan or so to reinforce the draft.

 

[mindless1]

Originally posted by: vtohthree... I can feel the capicitors at a scalding level, definately not desirable. This has "scared" me into less extreme overclocking with my gpu for now, it's currently clocked at a frugal 451/1.09, load temps seem to reach around 77*C now no matter what. When it was sitting for nearly 20minutes on much higher settings(around 500 +/-) it seemed to have "peaked" out as well, again this is only brief observation, not 24/7.

It may also help to leave the case slot cover off on a slot or two under the card, allowing more pass-by flow from other case fans. Of course this means leaving at least one slot (or that much space is already "empty" on some boards) space empty below the card but it goes without saying that putting another card next to it will interfere with optimal cooling.

Don't even get me started about swapping in better capacitors... it will usually resolve cap overheating issues but that's another topic for another day and until someone suggests they want to take a soldering iron to their card i assume they don't.

A fan on a bracket would help cool the card more, especially the caps. Need not be high powered fan, even 2000 RPM 80mm would be enough and it need not even be mounted on the case side-panel, sometime i'll just take a couple of 4cm long strips of aluminum, drill holes at each end and bend the end so it screws into the slot bracket holes and mount the two nearer corners of the fan to those. Has to be a well-balanced fan though so it doesn't vibrate. You may need longer screws for that, 6-32 1/4" to 3/8" (thread length) are about $1 a pack at some hardware stores.