Printable View
Hmm, anyone dare to try their card without a pad on the PLX chip? Well, perhaps not without anything at all, but at least something less effective than a thermal pad? Just struck me that those chips had some issues with cold on older cards, not @ 30C, but they use a new rev on the 5970, so might be worth a try if anyone feels brave enough..
So if my Tri-Crossfire set-up 5970 + 5870 with EK waterblocks on both are idling at 23-24 celsius and never going over 28 celsius when gaming, can I also "confirm" there is no cold-bug? :)
Asus 5970 + XFX 5870. I just played Batman AA for 1 hour, and the 5970 cores were at 28 celsius, and the 5870 core was at 27 celsius.
@levesque: clearly you have no coldbug
anyway, just dropped in for a quick update, got new card today and under water:
http://i787.photobucket.com/albums/y...E/new_card.png
About to try some bfbc2, usually hard locks my whole system in ~2 minutes, *crosses fingers*
Also quick note: not doing the profile to bring up 2d clocks and not getting any weirdness in windows so maybe this card doesn't freak out 2d wise. Here goes..
Edit: thank goodness, should put to rest most 3d/coldbug related issues, it's the card, try and trick your vendor into an rma (sapphire wouldn't claiming card is fine on air, so think of something more clever than I did :)
played bc2 for awhile and ran furmark for 5 minutes, both of those wouldn't last 30 seconds before, this is new card same vendor same block same drivers didn't change 1 thing, put on the ek block exactly the same as first card:
http://i787.photobucket.com/albums/y...no_coldbug.png
Am I correct in assuming this is a hardware bug, and not something they could write a driver for to fix?.. like one that simply ignores temperatures?
Or is it some hardware glitch that no amount of driver revisions is going to fix?
I am guesstimating from fact that people reported bugs for same cards having it or not and fact that after all this time since 5970 was introduced in market no software/bios fix exists (apart from hack of overclocking @2D to artificially rise temps), that it's no driver or software bug, but rather hardware one. It's wild guess of course.
Someone on amd forums said switching to asus bios completely fixed his cb problem. That's only 1 person but it is sort of good news and makes me think possibly bios update can fix it. I would have tried this had I not already ordered/recieved/installed a replacement.
Honestly though, who knows. I wish good luck to all those with the problem, my suggestion is to try and rma.
Anyone got any idea how to detect these problems while the 5970 original heatsink is still on?
Seems to me that most people having problems only after they changed to water cooling.
With my own EK waterblock coming in another 2-3 weeks, just thought I should test my GFX board first so I can RMA right away should there be problems.
I already got 1 295 gfx card that doesn't seem to want to do any 3d apps.
And since I had bought it overseas... there's no easy way to get warranty claim.
Having another defect gfx card would be a nightmare.
It's simply because with original heatsink on air cooling card never has that low temps to trigger bug, only on water :)Quote:
Originally Posted by washu9
- If you lived in some country with cold winters, i'd put case outside apartments, i beleave you might trigger it that way too :)
Who knows, maybe something can be made by some air conditioner aswell.
Freaky:eek:
Well, i just tried this. I can't give you any results now as it's kinda warm here now and i can't get the card to cool down to the usual 18º boot temperatures. Tomorrow morning i will test this as soon as the computer boots but i don't have that much faith it'll get fixed :shakes:Quote:
Originally Posted by mbreslin
Like i said, not a reason for me to RMA it, but knowing that it's there is annoying enough :rolleyes:
EDIT: UPDATE:
I let the PC shut for 3h a while back as it's night already and temperature dropped considerably, PC booted with the card at 22º on each core, ran Crysis benchmark immediately after i got to the desktop and it went through just fine. If i did this yesterday, i would be greeted with an instant crash. This is a good sign, but i'll confirm tomorrow after the PC has been off for over 8 hours.
hmmm, well see if this works then. :shrug:
Unfortunately it doesn't. I'm keeping it as it is but it's still a shame.
Turned on the PC this morning, ran Crysis benchmark just for a test and it hung after 20-30 seconds. I shut it down, restarted and used the PC again for ~15 mins and tested again, no issues whatsoever.
Imho ati/amd can easily implement fix in drivers for it. On 3d app launch check if it's one of problematic coldbugged gpu ID and if temps are too low, if it is, automatically momentarily rise clocks to "preheat" gpu, and only after few secs allow resources for launched 3d app. I'm shure that they'll never do it this way though :/
I doubt they know which GPUs are troublesome. I am not considering about buying a new card and selling this one with the original cooler because, just imagine if the new card still has the same problem... All the hassle for nothing and wasted money on top :(
If only i knew that the new cards being produced don't have this problem anymore...
Wonder what is happening exactly, if it's the MOSFETs too cold, the PLX chip, the GPUs... Maybe some chip that doesn't like going from "too cold" to "too hot" that fast :shrug:
Raising the idle clocks is not a solution for me either because i like my PC to use the less amount of power possible when idling.
Well, been thinking what can be done here.
There's a few options, and depending on your loop configuration, may/maynot be applicable for your system.
1. Use something like bigNG, Alphacool, etc to control your fan/pump so the temp doesn't get too cold/hot. (best way I could think of at the moment)
2. If you happen to only have the problem during startup, you can use a relay + timer to set an automatic delay for the pump/fan to turn on. Recommend turning off fan instead of pump, cause the temp would drop like a stone the second it's turned on. You may also vary the speed of the fan to ramp up slowly.
3. If you overclock your CPU, may consider putting all the loop you got into 1. You may have to reduce the overclock this way, but the water would be warm in no time. Unless of course you got huge radiator(s).
I really doubt its the actual GPU's, as 5870's seem to handle the cold very well. My bet is on the PLX atm, but seems weird nonethelessQuote:
Originally Posted by NaMcO
I got a dedicated loop for the 5970 with a FESER 240 radiator. Keeps the card under 28º idle (and it's not really cold here now, about 20º) and reaches high 40ºs when stressed. This loop is already controlled by the aquaero - the fans idle until the water rises above 25º but the card will already be under "operating" temperature by then, that wouldn't solve it.
Anyway, i lost the love for €150,00 of mine and sold the XFX 5970 BE to a friend who doesn't watercool ever and gonna try my luck on either ANOTHER XFX 5970BE or a PowerColor... Still have to decide which.
Wow, I figured I was pretty much alone with this issue. Guess not..
As much as I love my HD5870 (that card LOVES cold), I have to say I am pretty disappointed in ATI. WTF were they thinking sending out cards that coldbug at ~30C positive.
Epic QA fail...
Can any conclusions be drawn if any vendor's 5970s are more problematic than other vendors?Quote:
Originally Posted by jcool
So Far I see:
Warm Bug problems:
- Diamond
- XFX and BE
- Powercooler
- Sapphire
No Problems:
- Diamond
- Asus
- XFX BE
- Powercooler
- Gigabyte
What a mess.... Looks like a total coin toss, no clear trends.
Mine's a XFX BE and its got the ColdBug. I found an ASUS dealer, gonna try ASUS instead of PowerColor.
Definitely looks like a coin toss... can just hope i hit the spot now on a 50/50 chance :shrug:
hmm got a xfx 5970 be coming tomorow
so no way to tell witch ones do it or a fix :eek:
what about the powercolor 5970 lcs any of these pop up with it, if not what did they do nothing in the one review on it
http://www.overclockersclub.com/revi...970_lcs/17.htm
My bet is that they aren't even aware of the problem. They assemble the cooler and ship it as is probably only testing for power up. Once users return their cards due to crashing, the problem will begin to be known.Quote:
Originally Posted by bill_d
I could be completely wrong though...
i would think anyone that payed 800 for a 5970 lcs card with a cold bug would post something somewhere and i can't find nadaQuote:
Originally Posted by NaMcO
how hot does the PCIe bridge chip get
and has anyone tried to add maybe tin foil or thin paper on top,between heat pad and block to add just a little insulation to it :shrug:
what else does a 5970 have different to 5870
Mine with the bug was sapphire, didn't want them left off the bugged list, they all just slap stickers on reference ati boards anyway clearly it's an ati problem. I see it popping up with more and more watercoolers they will have to take notice eventually.Quote:
Originally Posted by trn
Replaced my bugged sapphire for a new sapphire, new card works perfectly no bug, I can see not wanting to go through the hassle but I took a chance and it paid off, if I had got another bugged card I would have sold both and went dual 5870's or 480's. Glad this issue is over for me.
So summing it all up most sell rebadged reference (and one needs reference for full cover for reference PCB to fit), and after that it's game of luck no matter what the vendor? (i'm guessing that few vendors that hadn't yet reported coldbugged cards simply might have a bit more extensive QA tests or by same game of luck such as particular vendor not as often used for LC :) or LC-ers got more lucky with particular cards. Afterall, noone has extensive statistics of behaviour of at least 100 gpus of each vendor/each revision/each stepping under liquid cooling :/ ).