Originally Posted by dragzi Problems
Signed up to post on this page.
Got an xfx 390x (first edition) a week after it was released. Regretted the purchase for ages.
Had many problems with it, mostly that it would hit 94'c as soon as anything intensive was fired up. This lead to the vrm temp hitting near 110'c and the system shutting down. It would overheat constantly, even in winter when my ambient temp was 19'c (live in london, so ambient in summer is only about 23'c). ASIC score was abou 78%
That was all with stock clocks (1050 core, 1600 mem? and 100% power target).
Ended up taking the side of my case off to get the card to sit at 85'c under gaming load. Really iritated me as I have a 5820k with a dh15 sitting ontop, overclocked to 4.2ghz, and it hits a max of 65'c with the side of the case on (nzxt noctis 450). Was preparing to spend £500 on watercooling gear just for the sake of this card.
Sent an email to XFX asking if changing the TIM would void my warranty, which I was told it would.The fix!
Got frustrated yesterday. Fired up furmark, and did some tweaking. Results are below Chronological. All "Stable for" results were furmark runs. Stock voltage is 1.1vSide of case off, 66'C idle
150% power target, stock voltage - Driver reset in under 20 seconds @ 95'c
100% power target, stock voltage - Driver reset in 60 seconds @ 95'c
100% power target, -25mv - Stable for 10 min, Driver reset @ 95'c
100% power target, -50mv - Stable for 30 min, No crash, max temp 84'c (Core clock was bouncing around the 950-980 mark)
150% power target, -50mv - Stable for 15 min, Driver reset @ 95'c
150% power target, -100mv (was hopefull) - Stable 30 min, No crash, max temp 72'C (Core clock was hitting 1050 throughout)
Time to switch from Afterburner to Trixx (as it supports > -100mv undervolts), and sort out a fan curve (no more 100% hoover for me!). Ended up setling for a steady ramp up from 20% < 50'c, 100% > 68'c
150% power target, -125mv - Stable for 20 min, Then artifacts
150% power target, -110mv - Stable for 50 min, No artifacts, Max temp 68'c
150% power target, -116mv - Stable for 60 min, No artifacts, Max temp 65'cSide of case on, 45'C idle
150% power target, -116mv - Stable for 60 min, No artifacts, Max temp 71'c
Threw a good 4-5 hours of cs:go and rocket league at it and saw it never hit more than about 65'c, and kept rock solid at 1050mhz (which at stock it would throttle down to 900ish). ended up with a voltage that was sitting at around about 0.984v, occasionally climbing to 1v. From what GPUz's sensor reading were telling me, I cut nearly 80W off the peak power usage. Unigenie heaven was giving me higher fps after the voltage drop, probably down to throttling.
TL;dr If your gpu is a toaster, see if you can undervolt without losing any performance.
Anything else I can do to ensure that its running stable? Was thinking maybe if the memory fills up it might not have enough to power the mem controller and core, but my logic may be off.