GTX 260 clocks keep dropping REALLY low?

[Rob_B]

I've noticed recently (as I was paying more attention as I started folding) that sometimes my clocks on my GTX 260 (700core, 1536 shader, 1221 mem) keep dropping to 400/297/799 & nothing but a reboot puts them back?

I noticed it after trying to change my shader clocks during furmark & my driver crashed/recovered. I've now tried to play a little HL2 DM (crashed as it sometimes does) and L4D2 (froze but logo still spins, ALT-TABd out) while F@H runs in the background and noticed they've dropped again.

I'm using EVGA Precision, CPU-z also reports the lowered clocks. 260.99 drivers.

I'm stumped, any ideas short of trying other, older drivers?

Edit: Hmm...Furmark reports my higher clocks (on mem/core at least), while EVGA & CPUZ report lower ones?

Cheers!

Rob

[Terbinator]

Sounds like its too much of an overclock if its failing/dropping during gaming/F@H.

Try reducing them to stock and seeing if they still drop back further when you do something intensive.

[cptwhite_uk]

You sure this isn't the just the auto-downclocking while at 2D on desktop. It's completely normal if you're only observing this at desktop. Fire up a 3D app and check the speeds then.

[Rob_B]

It's not really failing, L4D2 is just some bug, HL2 DM sometimes craps out on me, other games run fine. They won't then clock back up (using EVGA) though, even though the temp is only 45C

You sure this isn't the just the auto-downclocking while at 2D on desktop. It's completely normal if you're only observing this at desktop. Fire up a 3D app and check the speeds then.

Nah, even with furmark running EVGA & CPUZ report the low clocks, they're not even possible to set that low in the app

2D 400/799/297
3D 702/1548/1224
3D (after driver crash/recovery) 400/799/297

Looks like it isn't going back into 3D mode after a crash, might drop the clocks a little & see what happens but I'm getting no other artifacts/errors in anything else But why is furmark reporting the 'coorect' higher clocks?

[Domestic_Ginger]

It downclocks after failing an OC and only a restart sorts it out like you said. If something is telling you your drivers failing it won't crank back up. Good OC though; mine wont go much higher than 1440ish.

Whats GPUz saying? just to be difficult.

With precision you may want an daggressive fan profile too.

[Rob_B]

It downclocks after failing an OC and only a restart sorts it out like you said. If something is telling you your drivers failing it won't crank back up. Good OC though; mine wont go much higher than 1440ish.

Whats GPUz saying? just to be difficult.

With precision you may want an daggressive fan profile too.

GPUZ/CPUZ reporting my stock clocks (not actual) - not tried GPUZ with a crashed setting.

Seems that furmark is mistaken as its a worse score on the lower clocks. Clocked down to 700/1502/1202 (actual 702/1512/1215) now, will have to see how it goes Hovers around 70C on load with standard profile

Think it would be shaders crapping out? My clocks did seem a tad high compared to the average, don't think it's the mem as it doesn't seem to show any errors?

[Domestic_Ginger]

Furmark homepage has examples of failed OCs with little vids wrt to the bit. Mine just protests and locks up on both core and shader overclocks. Could try them for stability one at a time or just drop everything a little.

Interestingly L4D2 was using about 800mb vram the other day which is much more than furmark.