Hi! I've had a stock evga 470 for a while and oc'ed it to (725/1450/1674) @1.00v. It passed several furmark burn tests so temperature wise i was pretty sure it's ok. Fan profile helped keep it at 89c flat at the extremes of furmark and crysis and mw2 are the only ones that bring it up to 82-83c. I also play tf2, l4d2 and several other games occasionally.
The only problem is that although it solidly passed all those stress tests, it occasionally funks out while playing crysis or mw2. It's a "once every 2 or 3 weeks" driver crash kind of thing. Not terrible but it sets the core clock back to 405 mhz so I do need to reboot and it does make me wonder if there's something wrong w/ my overclock. What's odd though is the crashes happened at temperatures like 75 or 80, nothing nearly close to mid 90's or higher.....so maybe voltage or something else non-core temperature related?
What do you guys think? I've heard it might be inadequate vrm cooling but I'm not really sure . Thanks in advance
Drop your memory speeds. You get little performance gain from pushing the VRAM. Aside from that I'm not really sure. I once had an issue where ater 5-10 minutes of GTA 4 my GTX 260 SLI would just go back to 2D clock speeds, and I found out it was the new 250.xx drivers causing it. I went back to 197.45 and now have perfectly stable gameplay (with better fps than new drivers)
To test VRAM in FurMark, run it with 8xAA or extreme burn mode. If you don't use much AA it won't stress the VRAM much so you can have a stable core but unstable memory.
Memory speed does make a difference tho, a huge one. But, it also has the most stability issues since you can't increase the voltage.
Oh right one other thing, if you're only reading GPU core temperature, temps could still crash it. Use GPUZ sensors to see the VRAM and VRM temps. I've had my core stable at 80C but the VRM went up and up to 120C and it crashed.
Hmm, well 1674 is the stock memory speed but I guess dropping it 10 mhz can't hurt. It is intermittent but I guess we'll see how dropping memory, the new 260.99 drivers, and bumping up core voltage a tad goes (other advice from evga forum).
VRM temps are from the "PCB temps" in GPU-Z or msi afterburner right?
I'm wondering if zalman's vf3000f might solve that, their pcb heatsink looks a lot nicer than the stock one
Hard to say, I have ATI so sensors and such are labelled differently.
And aftermarket cooler might help but they're usually focused on core temp. However, with two fans the vf3000 should provide good airflow onto the PCB as well so it might work nicely for you. I'd check out as many reviews on it as possible first.
If the memory speed is stock, then maybe it's not the issue. Could just need a tad more voltage. Otherwise yeah could also be a simple driver malfunction and updating might fix it. At least it's intermitten!
add a bit more voltage and see how that works out for ya....i have a GTX460 overclocked WAY past the GTX470 almost to 480 specs but not their yet and its the other way around...3Dmark Vantage doesn't crash it just goes a lil flakey but l4d2, new medal of honor, crysis all those games and i never see my display driver crash.