Chilled 7970
 

Just wanted to chime in with some of my recent results on my MSI 7970. I posted around a month ago about a ton of issues I was having with my MSI 6970 with regards to running cold temps. Well, all those are completely gone with my 7970.

I put a koolance block on yesterday and put it into my water chilled loop and was ecstatic with the results. We had extremely low dew points yesterday, right around -17C, and was getting idle temps of 2C and load temps ~13C. This allowed me to push my card up to 1300 MHZ with 1.25V. I almost got 3dmark to run at 1375 but it crapped out with about 2 seconds to go on the last test (probably could of pumped up the volts a tad).

One thing I would note is that I had to do several Windows installs to get the AMD drivers to work properly with afterburner. In the end I went with a clean install, drivers from the CD that came with the card (something like 8.91), then installed the 8.95 drivers on top of those. Once I did this I was able to overclock beyond the CCC limits but I have not been able to get the memory voltage to show up. Hence I have only been able to oc the core and not the memory yet.

My main reason for posting this is for those people who chill their loops and have issues with their gpu's like I was. Just wanted to put it out there that the MSI 7970 OC'd version will play nicely with really cold temps!!

Proof:

http://img52.imageshack.us/img52/2926/maxst.jpg

A little OT but what kind of waterchiller do you have and does it make alot of noise? Also is it in or outside the case?

Nice post and results :) ...I'm glad I am not the only one who had to do a clean install of Windows lol. I had nightmares for almost a week, before I gave up. I use Afterburner 2.2.0 Beta 12 and it seems to be holding its ground. I am currently at 1250/1575 @ 1.22v ...I got an average of 100 FPS on AVP Directx11 Benchmark...The most I got was 1250/1785 @ 1.25v ...However, the drivers didn't like that and now I am comfortable where I am at. Maybe down the road, I will push it again. The highest I can overclock my Sapphire 7950 OC Edition is 1300/1620 @ 1.3v , but I don't want to blow up my card lol , so I will see what the drivers will do down the road.

Well my chiller is made from a old window ac unit that I recycled. I took it apart and placed it underneath the crawlspace of our home so I don't have to deal with the noise. Also, my method of chilling my water is not quite the same way others do it. I have a box cooler that I placed the cold end of the ac unit in and I placed two swiftech 3.120 rads inside of that box. So basically I pump my fluid thru the rads and pull cold air across them to chill my water. I control my fans in pairs to help control condensation when I have high dew points by turning some off when necessary. I also have a 120 rad attached to the back of my computer case so that I can draw cold air into my case hence the reason my hard drives and mobo have nice low temps!

Thanks for the info..! Nice clocks!

Though most expect the card to work as expected above freezing...its BELOW freezing that cards have issues.

Cooling your hard drives will kill them faster.

Well, my 6970 had issues with the driver constantly crashing when I got anywhere near 30C. Which was really frustrating cause even without my chiller on I can run temps at start up below or near 30C. I would literally need to run occt for nearly 20 minutes before my fluid would be warm enough for me to game without any crashes. I was able to use that as an excuse to give my wife as to why I HAD to purchase a 7970.:D Especially since I only had the 6970 for less than 3 weeks before I sold it.

Wow, I wasn't aware of that issue. I don't actually cool them directly, but I blow cool air into my computer case. This helps keep my other components cool while helping me to keep my fluid temps above the dew point. In particular I like feeding my PSU cool air in hopes of keeping it from blowing up!!

I know what you mean about pushing the card. I was running mine at 1200/1375 at stock volts and was hitting temps ~85C before I got the block. This would be after about 4 - 5 hours of playing skyrim.

I wonder why my 6970 had no issues going cold (and I mean for real cold, not AC cold).

Beucase not al ahrdware is the same? I had two cards myself with cold bugs, and a coupel of 5-series cards, too. It was me that said "roomtemp coldbug" when respol23 first posted his thread. It's possible if he went colder colder it might have gone away, but htat's not what he was after.

It's all about the leakage.

No offense EarthDog, but did you really post that, or is your girl playing tricks agian??:laugh:

I am not sure why I was having issues either. I scoured the net but didn't find anyone with the same issues as myself, but I thought it may have been limited to my card or the components on the card itself.

What brand of 6970 did you have? Was it a reference design?

I was initially worried about my MSI 7970 having the same issues because it was the same brand as my 6970 but I hoped that having different components would fix my cold bug issues.

edit:

Didn't get your meaning at first regarding the "real" cold comment. But I think the real cold you applied was only to the core possibly. What I have read about cold issues is that it could be the vrm's, the ram, or anything else that is being chilled.

Reference and aftermarket...

When putting a card under DI or LN2, the entire board gets cold around the socket.

True, I didn't consider that! In your experience, would the vrm's also see cold temps or are they far enough away from the core that they don't really get that cold. Also, there is a small piece near the display ports that is also cooled, would that be far enough away to not be affected by really cold temps.

Just trying to see what else could have been issues before just in case I run into any issues further down the line with other upgrades.

VRMs, voltage controller, RAMDAC, memory, and the GPU itself can all be probable causes. Some caps and stuff don't like freezing either, but you weren't really at the temps that any of those should have really been the issue, except memory or the GPU itself.

I was told by XFX that this behavior you/I had was a room-temp coldbug, as I had the exact same issues s you did, plus running the card warmer made the problems go away.

If I shut down my rig since i got my 1100 there is something where I have to let it run on the black prepost screen for a minute and then restart before it will actually post and boot. Then its fine.

Worst 'coldbug' ever. :shadedshu

Repsol - What he said it about right. I did not have an XFX (that I know), so maybe they have different, perhaps cheaper parts as Cad described above. But that is just SILLY to have a 'room temp cold bug'. XFX FAIL. :shadedshu

Actually, i think it's the memory controller on certain chips, specifically the low-vid ones. And yeah, poor quality GPUs, for cold, anyway. You know how coldbugs work. :laugh:

The bug I had was room temp and below because anything ~30C and below became an issue. My card was an MSI so I don't think it is limited to brand but I also think it could have just been a bum card. I sold the card to someone within the department I work in and he hasn't had a single issue running it on the stock cooler. Go figure!!

i can get 1800mhz with stock volts on my mem :confused:

iirc, the GTX580, damn near if not all, do not like the memory getting very cold, but uhh room temp was ok. :rolleyes: :roll:

Huh, good info. I haven't played with 5-series NV cards at all.


When I reported the issue to XFX, it was the tech there who said coldbug, and instantly. He even seemed to know what temp the card needed to reach to stop artifacting, and he said it was common, not like a big percentage of chips, but enough that they knew of the issue, and how to deal with it.

And my card did it on stock cooling. I guess my house is cold. :laugh:

Wow!! It seems that every time I try to run 3dmark11 with higher clocks on mem it errors out on the last test. I even tried just 1400 Mhz and it crapped out. Tonight I was gonna scale back my core clock and up the memory clock to see what happens.

Frequently, when pushing the limits, that is exactly the trade off. With memory at stock, my core clocks are higher. When my memory is at its limits, I cant clock my core as high.

Thats so messed up, its almost hard to believe. But due to your innate awesomeness, I have no choice but to believe you. :p :rockout: :D

(in all seriousness, I do believe you, thats just so jacked up its a mind fuck for me)

That matrix I played with cold I had to blow dry the memory/board to warm it up......... bleh. Fine line between warm enough, and OH SHIT ITS MELTING AND RUNNING DOWN THE CARD!

saw your new run at 1350 makes me wanna put my rad in ice or get my lazy butt and put it on ln2 already :roll: