*GTX 260/280 Unofficial TPU! Thread*

[SplitBURST]

Thought I'd share how my 260gtx is doing....

First you forgot to reload GPU-Z after OC the gfx the values are wrong for that kind of OC. And what are those temp on the CPU full load at that insanely high OC?

 

[Tatty_Two]

Actually I think the core goes from 756 to 783 with no stops in between. I've seen where watercooled cards have done 783 but I don't want to watercool for various reasons. The current rumor about the 280+ is around 3 weeks from now.

This website has the 280+ and the 350 listed. I'm hoping to get a 280+ and stepup to a
350.

http://austin.net.au/ProductList/tabid/103/Default.aspx?Category=GRAPHICS+CARDS

I was just reading what you said again.....mines not watercooled?

 

[Tatty_Two]

Paul, have you tried pushing further yet? Interested to see what unlinked core/shader combination you can get!

 

[johnspack]

Just some fyi, I was doing some gaming because I just picked up a g5 mouse, and noticed at a particular point in infernal which uses phsyx, I'm now getting 160fps @ 1920x1200 whereas I was getting under 70fps @ 1600x1200 on my 9800gtx. I think the cuda engine really is better on these cards!

 

[dadi_oh]

I got my new D-Tek Fuzion V2 waterblock and Unisink today so my GTX260 is now under water. I'm hitting about 40C under load now with the clocks at 783 core, 1566 shaders, 2484 memory. If I try to bump the shaders up the next step is 1620 and I can bump my core to 810. At those settings Furmark stability test seemed fine but 3DMark06 crashed so I guess it is unstable up there still (even with temps only hitting 42C).

Before the water I had tried bumping extra voltage to 1.12V (mine is 1.06V default) but that seemed to activate some sort of voltage protection on the card. I may try it again now that it is watercooled and the unisink is on the power section.

So here is a 3DMark06 run of 18166. This was with the NVidia image settings on "Quality". I might squeeze a bit more by setting it to "high performance" but that is not what I game at.

 

[Kursah]

Good stuff guys! I can't wait to get my rig loaded back up...just a couple more days and this recent build I have completed is good to ship off. It's decent, but it's not my rig by any means!

 

[waxking1]

I was just reading what you said again.....mines not watercooled?

I don't think I mentioned your card. I was talking about someone on the EVGA forum. You have some kind of freak card, as I have never seen another one run the speeds yours does on air. Most of the shader clocks over 1512 in this thread are on water.

http://www.evga.com/forums/tm.asp?m=560911&mpage=1&key=&#560911

 

[Tatty_Two]

I don't think I mentioned your card. I was talking about someone on the EVGA forum. You have some kind of freak card, as I have never seen another one run the speeds yours does on air. Most of the shader clocks over 1512 in this thread are on water.

http://www.evga.com/forums/tm.asp?m=560911&mpage=1&key=

Ahh right, didnt realise you were talking about anyone in particular, just though it was a general comment......hmmmmm freak card, maybe not, two friends have also bought Gainwards from the same shop as I did (since I bought mine) and they get similar results......obviously that dont mean all Gainwards will do these speeds, but hey, it could mean thats some do more!

 

[waxking1]

I was under the impression that Nvidia made all these cards and they were just rebranded but I could be wrong. Do the companies such as Gainward, EVGA, BFG etc actually change anything?

 

[ZenZimZaliben]

Well I got my MSI GTX280 installed, and my adapter for my MCW60 just came in today, so that will be the project tonight getting it on my water loop.

I have no screen shots, yet, try to get some tonight, but with the card on stock settings with my below specs my 3dmark06 is 20K. I managed to push the card to 700/1400/1200 for a mild OC. That pushed me just over 21k.

It has more room to go even on air and once on water I would expect 750Mhz on the core, not sure if the shaders will make it linked that high. The RAM has plenty of OC head room.

Still in testing mode. It may not go any higher at all. I havent really pushed it as far as it will go on stock air.

 

[dadi_oh]

Ahh right, didnt realise you were talking about anyone in particular, just though it was a general comment......hmmmmm freak card, maybe not, two friends have also bought Gainwards from the same shop as I did (since I bought mine) and they get similar results......obviously that dont mean all Gainwards will do these speeds, but hey, it could mean thats some do more!

Do you know if the Gainwards are reference designs? I believe that Gainward/Palit are often among the first to introduce non-reference boards but I don't know if the GTX 200 series is available in non-reference yet.

Back when I got my first 9600GT it was a reference design and ran into power issues (OCP) if overvolted. The non-reference Palit cards had a better power section design and could be over-volted without this issue.

If this is a non-reference board then it is entirely possible that they beefed up the power design in keeping with their past boards.

 

[dadi_oh]

Well I got my MSI GTX280 installed, and my adapter for my MCW60 just came in today, so that will be the project tonight getting it on my water loop.

I have no screen shots, yet, try to get some tonight, but with the card on stock settings with my below specs my 3dmark06 is 20K. I managed to push the card to 700/1400/1200 for a mild OC. That pushed me just over 21k.

It has more room to go even on air and once on water I would expect 750Mhz on the core, not sure if the shaders will make it linked that high. The RAM has plenty of OC head room.

Still in testing mode. It may not go any higher at all. I havent really pushed it as far as it will go on stock air.

Sounds great. Keep us posted.

I just got my 260 on water last night (hitting 40C when folding overnight) and plan to do some more tweaking tonight. So far I only managed to get it 1 "step" up on the shaders (1512 to 1566) with 100% stability. The next shader step is 1620 and I can't get stability there yet. Funny thing is I can get the Furmark stability test to run fine even at 1620 shaders and 810 core but when I run 3DMark06 it will crash the drivers. hmmm... I may try some Crysis bench to see if the same thing happens there.

 

[ZenZimZaliben]

I have been checking out other forums and have found several users with the core above 750. Highest I have seen so far is 786.

 

[dadi_oh]

I have been checking out other forums and have found several users with the core above 750. Highest I have seen so far is 786.

With my voltage boosted to 1.12V on extra I fail at 810 on the core. But I just did a run on 3DMark06 at 783 core and 1620 shaders. Funny thing was the score was marginally worse than at 756 core 1620 Maybe every run has some variation to it or possible the extra core speed doesn't make a difference at the 1280X1024 resolution. I will try some Crysis benches at higher resolutions and see if I can see a difference there.

 

[Tatty_Two]

I was under the impression that Nvidia made all these cards and they were just rebranded but I could be wrong. Do the companies such as Gainward, EVGA, BFG etc actually change anything?

Well I think the Gainward, and therefore at a guess, Xpertvision and Palit are non reference cards so yes, maybe an extra phase (my Palit 8800GT's have 3 phases as opposed to the standard 2) that makes for better and more stable voltage delivery at high speeds.....but I cannot speak for all non reference cards.

 

[dadi_oh]

With my voltage boosted to 1.12V on extra I fail at 810 on the core. But I just did a run on 3DMark06 at 783 core and 1620 shaders. Funny thing was the score was marginally worse than at 756 core 1620 Maybe every run has some variation to it or possible the extra core speed doesn't make a difference at the 1280X1024 resolution. I will try some Crysis benches at higher resolutions and see if I can see a difference there.

Well it seems to have resolved itself. I managed to set core to 810 and shaders to 1620. Ran a bunch of benchmarks and then Furmark for about 1/2 hour. Then played COD4 for about an hour so I assume this is stable now. In Furmark I can get to 48C on the core so the higher voltage definitely generates more heat.

This card rocks... Now at 810 core / 1620 shaders / 2500 memory.

 

[Tatty_Two]

Well it seems to have resolved itself. I managed to set core to 810 and shaders to 1620. Ran a bunch of benchmarks and then Furmark for about 1/2 hour. Then played COD4 for about an hour so I assume this is stable now. In Furmark I can get to 48C on the core so the higher voltage definitely generates more heat.

This card rocks... Now at 810 core / 1620 shaders / 2500 memory.

Show us a 2006 run.....with those clocks you might just beat my score, I can run those clocks in most games but they crap out on 2006, best I can do on that is 800/1600 I think......nice clocks though!

 

[mrw1986]

I should have my EVGA GTX280 in soon! I can't wait!

 

[dadi_oh]

Show us a 2006 run.....with those clocks you might just beat my score, I can run those clocks in most games but they crap out on 2006, best I can do on that is 800/1600 I think......nice clocks though!

Well I only have a Q6600 so I might not beat your overall score. Here it is anyways.

I seem to get erratic results with 3DMark06? The score sometimes moves by 100 points or more in subsequent runs. I don't really take much care shutting down everything in my tray so it is possible something in the background interferes a bit.

Now are you sure you are actually running at 800/1600? When I watch the actual values in EVGA Precision it seems that the Shaders and Core clocks go up in certain increments with nothing in between.

For example the core increases 702 -> 720 -> 756 -> 783 -> 810

and shaders increase 1512 -> 1566 -> 1620

If I set 800/1600 it goes directly to 810/1620.

 

[johnspack]

Well, aim for this in vantage:
http://img.techpowerup.org/081003/Vantagegtx2.jpg
Remember, 06 is a cpu bench, vantage is a gpu bench...

 

[largon]

How come nobody has posted FurMark results for GTX 260 in the result compilation thread, there's only a few GTX 280s?

 

[Tatty_Two]

Well I only have a Q6600 so I might not beat your overall score. Here it is anyways.

I seem to get erratic results with 3DMark06? The score sometimes moves by 100 points or more in subsequent runs. I don't really take much care shutting down everything in my tray so it is possible something in the background interferes a bit.

Now are you sure you are actually running at 800/1600? When I watch the actual values in EVGA Precision it seems that the Shaders and Core clocks go up in certain increments with nothing in between.

For example the core increases 702 -> 720 -> 756 -> 783 -> 810

and shaders increase 1512 -> 1566 -> 1620

If I set 800/1600 it goes directly to 810/1620.

You need to do some tweaking there, your core and shaders are a little higher than mine but my SM2 and SM3 scores are a lot higher, my CPU speed will help with that but it still seems to be a lot of difference, my scores at 800/1600 are:

SM2: 8983
SM3: 9298
CPU: 7337

Sometimes like you have mentioned, you find that if you are too close to the wire and you are over pushing the speeds that performance suffers because of it. As for your point on incremental increases, yes, that has been the case ever since G80, all NVidia cards have incremental increases, as for mine, dont know, everything says 800/1600 including rivatuner.......... just double checked in EVGA and it is set at 810/1620 on the left although 800/1600 on right!

 

[Tatty_Two]

How come nobody has posted FurMark results for GTX 260 in the result compilation thread, there's only a few GTX 280s?

perhaps they were too busy tweaking! I dont do all the benches, I wouldnt have time to play games! I'll throw one in there this weekend though just for you.

 

[dadi_oh]

You need to do some tweaking there, your core and shaders are a little higher than mine but my SM2 and SM3 scores are a lot higher, my CPU speed will help with that but it still seems to be a lot of difference, my scores at 800/1600 are:

SM2: 8983
SM3: 9298
CPU: 7337

Sometimes like you have mentioned, you find that if you are too close to the wire and you are over pushing the speeds that performance suffers because of it. As for your point on incremental increases, yes, that has been the case ever since G80, all NVidia cards have incremental increases, as for mine, dont know, everything says 800/1600 including rivatuner.......... just double checked in EVGA and it is set at 810/1620 on the left although 800/1600 on right!

This inspired me to try to push my Q6600 a little harder just for a benchmark. I run my Q6600 at 440X8 = 3520 24/7 folding and it is Prime95 stable. However I can bench at higher clocks if I want. So I tried 3800MHz (475X8) but it made it all the way to Deep Freeze and reset the machine. So I backed down to 465X8 = 3720MHz. Attached is the screen shot of the 3DMark06 run. I finally broke 19K. I will see what sort of stability I have at this Q6600 clock but I suspect it will not be Prime95 stable.

So you can see that the increase of 200MHz on the Q6600 translates to big increases in both SM2 and SM3 scores as well as CPU score. I just need a better processor I guess. But I am cheap and I got this one for $150 shipped.

I tend to believe the EVGA Precision output on the graph at the left. So I think that you really are running at 810MHz even though you set it to 800MHz.

 

[dadi_oh]

How come nobody has posted FurMark results for GTX 260 in the result compilation thread, there's only a few GTX 280s?

Well here's a Furmark run for me 4319. Same score as when I ran at 756 core and 1512 shaders?

I only use Furmark for stability testing most of the time.