View Full Version : Max. 24/7 HD4890 core voltage
trakslacker
08-06-2009, 08:11 AM
Been working with an XFX 4890 the past few days and have been tinkering with voltage to see how far it'll go. Currently I'm at 1050 core/1185 mem 3DMark and Furmark stable, with a volt bump to 1.38v VCore. Card is aircooled, but core and VRM temps are both maxing out in the mid-70s C.
Assuming temps stay safe, what's the consensus on max. core voltage for 24/7 use? Also, has anyone played around with memory voltage adjustments on the 4890? Thanks for the input.
SNiiPE_DoGG
08-06-2009, 08:58 AM
with good temps in order 1.45v is my 24/7 topend threshold - although I have pushed my 4870x2 on water to 1.5v 24/7 without any negative repercussions.
trakslacker
08-07-2009, 06:45 AM
Thank for the response Snipe Dogg. This is generally inline with what I've seen elsehwere. Are we really the only ones here that are bumping voltage on these cards?
SNiiPE_DoGG
08-07-2009, 08:52 AM
not sure, I think some people are just scared or something :rofl:
W1zzard
08-07-2009, 10:38 AM
i'm seeing cards from aibs with 1.5v on the gpu and more
G H Z
08-07-2009, 01:08 PM
1050 core Furmark stable on air is a damn good card :up:
trakslacker
08-07-2009, 01:22 PM
Thanks W1zzard. I saw the TPU write-up on the Powercolor PCS card that had 1.5v being pumped through it stock, but didn't know if this was a truly one-ff thing or not. From your experience with these cards do you think 1.5v is actually safe or are some AIBs pushing a bit past the limit?
As always, thanks for your continued efforts with all the TPU apps (AtiTool, GPU-Z, GPUTool, etc). They are extremely useful to the community.
Do have one quick question regarding the artifact detector in GPUTool. when I push much above 1050 core I start getting artifacts detected, but its minimal (~2 pixels missing, only happens once per minute or so). I know that if its artifacting at all it's not truly "stable" (at least in my book), but does that level of artifact detection translate into actual gaming issues? I only ask b/c my time for gaming is pretty limited right now, so don't have a huge amount of time for testing. Thanks.
1050 core Furmark stable on air is a damn good card :up:
Thanks, it does hit a wall right at 1050 though. I can push it to 1.45 vcore and it really doesn't go beyond 1050. Given this thread though, I may try at least a limited run @ 1.5v and see if brute force and break the wall. :cool:
W1zzard
08-07-2009, 02:21 PM
Thanks W1zzard. I saw the TPU write-up on the Powercolor PCS card that had 1.5v being pumped through it stock, but didn't know if this was a truly one-ff thing or not. From your experience with these cards do you think 1.5v is actually safe or are some AIBs pushing a bit past the limit?
i believe that to be safe enough for several years of use.
Do have one quick question regarding the artifact detector in GPUTool. when I push much above 1050 core I start getting artifacts detected, but its minimal (~2 pixels missing, only happens once per minute or so). I know that if its artifacting at all it's not truly "stable" (at least in my book), but does that level of artifact detection translate into actual gaming issues? I only ask b/c my time for gaming is pretty limited right now, so don't have a huge amount of time for testing. Thanks.
i wouldnt worry about it, but why not go to 1045 and be artifact free? there is no difference between 1045 and 1050 but in return you get 100% stability. i've seen games crash randomly (with very low probability) in such cases. wouldnt it be annoying if it crashes just at the worst time?
BeepBeep2
08-07-2009, 02:46 PM
i wouldnt worry about it, but why not go to 1045 and be artifact free? there is no difference between 1045 and 1050 but in return you get 100% stability. i've seen games crash randomly (with very low probability) in such cases. wouldnt it be annoying if it crashes just at the worst time?
I think it's an e- thing.
If someone crashed from heat in FurMark but didnt artifact it might be different...
so max for daily 1.45 v ??
SNiiPE_DoGG
08-08-2009, 02:06 AM
Hipro says that unless you use 50mhz clock increments in the 4890 then the clock is not true (or something along those lines) so only 850-900-950-1000-1050-1100 are legitemate
maybe I am misinterpreting him or he is wrong but something to explore definitely.
Hipro says that unless you use 50mhz clock increments in the 4890 then the clock is not true (or something along those lines) so only 850-900-950-1000-1050-1100 are legitemate
maybe I am misinterpreting him or he is wrong but something to explore definitely.
well easy to test , try 900 and 925 if no deffrent so hipro is right !
i'm in
http://i29.tinypic.com/10fqb1k.jpg
SNiiPE_DoGG
08-08-2009, 10:06 AM
as soon as I am done photographing my system (lol) I will definitely test