HD4870 Furmark black screen crash.

[dnottis]

i sinked mine. but i've decide fur just dont play nice. sometimes i can complete a run sometimes not. so i've given up on fur. however i can game at 830mhz core for hours with no plroblems. i'll dl everest in a minute and check vrm temp.

Anxiously waiting to see your VRM results...

I was hoping that was the spot to sink - argh.

 

[rangerone766]

having trouble finding a version of everest that will read vrm temp. still trying tho

 

[dnottis]

having trouble finding a version of everest that will read vrm temp. still trying tho

http://www.9down.com/EVEREST-Ultimate-Edition-4-50-1455-Beta-29037/
http://middown.com/news_3945_Everest-Ultimate-Edition--4.50.1455-Beta.html

This one does at least for a 4870.

 

[rangerone766]

that one didnt show vrm on it. getting version 4.50.1340 now but its dling slooow.

 

[dnottis]

that one didnt show vrm on it. getting version 4.50.1340 now but its dling slooow.

4850 or 4870?

 

[rangerone766]

4850 or 4870?

4870 srry

 

[dnottis]

4870 srry

I dunno how far back you can see VRM temps... here are some builds, 1455 is the latest.


Build 1455 -
http://middown.com/news_3945_Everest-Ultimate-Edition--4.50.1455-Beta.html

Build 1454 -
http://soft.softoogle.com/ap/everest-ultimate-edition-build-download-369.shtml

Build 1448 -
http://www.majorgeeks.com/download4876.html

Build 1426 -
http://www.lavalys.com/beta/everestultimate_build_1426_zyc4xlkn0gs.zip

 

[rangerone766]

1430 is what the op shows in his screen shot, i'm getting it now but the torrent is running slow, only 1 seeder.

By rangerone7669 at 2008-07-25

 

[dnottis]

I've changed the sinks around, sinked some other parts that I thought might be part of the issue. --- just noticed this - with the latest build of Everest GPU VRM - showing 72A drawn before it crapped out. (check bottom of the last picture showing Everest)

I actually set this 80mm fan right on top of the VRM area. This was pushing about 30 CFM right over the card.

VRM hit 125c and shut down. Clearly it's not a temperature issue because every heatsink was cool to the touch even at the point it went to a black screen.

 

[nafets]

I'm no guru at this, but here's my best guess...

I'd venture to say that the VRM sensor (and/or a combination of other sensors) is registering incorrect temperature values. There must be a set maximum sensor value, that when hit, shuts off power to the GPU/memory/VRMs.

So when running Furmark (or other intense 3D loads) this value is being hit, prematurely, even though the actual circuitry isn't even close to the maximum temperature.

With my HD4870 + the maximum cooling setup (4+ 120mm fans), doing the simple finger test, none of my heatsinks (on the front or back of the card) are at the point where they are burning hot to the touch, leading up to a crash in Furmark.

 

[dnottis]

Here is what I believe and what we need to test. The latest build of Everest shows the GPU VRM amperage draw (last reading) for the HD4870s. I need some others to run Furmark and watch the amperage draw. I dont think it's a temperature issue anymore, nothing is really hitting 125C / 250F as there is nothing on the card that hot. However, I see the GPU VRM amperage draw hit 72amps before I shutdown and the D601 light is a critical power trip on the 4870.

So I need you guys to run Furmark with latest Everest (build 1455) and see if your amperage draw is maxing your PSU rails. I have a TT1000 with quad rails. Rail 3 and 4 are dedicated to the PCI-e connectors. Each rail is 36a. My PC is going to a black screen when the HD4870 hit 72a draw, the maximum amperage I have available.


Links for latest Everest are here -
http://www.9down.com/EVEREST-Ultimat...55-Beta-29037/
http://middown.com/news_3945_Everest...1455-Beta.html

GPU VRM is the bottom reading in Everest -

 

[rangerone766]

i'm using the same version of everest as you, and not getting a reading on vrm's. my card is a 4870, ati branded. not a built by ati for saphire or any other 3rd party.

with so many 47xx users having troubles with fur, i wouldnt worry too much. i can game at insane core speeds without crashes, even run 3d06 at high core speeds. but fur and aquamark crash out.

screen cap showing mine doesnt read vrms.

By rangerone7669 at 2008-07-26

this is the exact card i bought if it helps.
http://www.microcenter.com/single_product_results.phtml?product_id=0292287

i guess my card just doesnt have vrm sensors. i tried many versions of everest and never got 1 to read anything on vrms. several did show incorrect values fore core temp tho. reading -214c on core diod.

 

[dnottis]

Actively cooled - VRM temps not affected in Everest at all!!! VRMs are pretty cool to touch now when loaded. WAIT A FREAKIN' MINUTE... is that 900 seconds of Furmark I just passed?????? Why yes, it was....

I would call this one fixed. Gonna be looking at a D-Tek Unisink and mounting a 50mm fan on it over the VRM area.

 

[dnottis]

Gentlemen - here is the fix.

 

[mat9v]

And here is my Everest sensor....

 

[nafets]

Odd, how your last Furmark run shows the GPU VRM draw is only 64A, 8A less than before, while the GPU VRM temp is still at 126C.

Seems you've got a full version of Everest, which is why you can get the GPU VRM value. It's blocked out with [TRIAL VERSION], otherwise...

Still working on it...

 

[dnottis]

Heat would cause it to draw more though. Which is why with the fan right over the top its keeping the card from drawing more amperage. The cooler the VRMs run the more efficient it will run.

RTHDRIBL only draws about 51a max in Everest. Which is why the only app I was having issues with was Furmark - Furmark is causing the GPU to suck some major power.

I dont even believe the VRM temp is correct, nothing on my card is hitting 250F. 250F would probably melt the PCB.

 

[nafets]

Well I tried again, with a 120mm case fan right over the area you specified. Furmark lasted quite well for about 5 minutes, much longer than before, and then crashed as before. So I think you've hit the jackpot as to what the problem is. Obviously with a higher CFM fan, I could probably have complete stability in Furmark, but then noise would become a factor.

One of the reasons I've gone to aftermarket cooling is to decrease noise, by using quieter 120mm fans. Adding more fans or higher CFM ones just defeats the purpose, only to gain in running 1 or 2 extreme load 3D programs, with stability. So for now I'm just going to stick with one 120mm cooling fan, centered closer to the VRM sensor area, for improved stability.

For me, stress testing with Rthdribl, is fairly good at representing usual temps and GPU load during 3D games, and runs just fine, with no stability problems. So as before, I'll just have to bury this little Furmark problem in the back of my paranoid mind.

All in all, great job though Dnottis. It's nice to have such thorough and improvisational posters around to really take the time to figure out these nagging little problems.

 

[dnottis]

Well I tried again, with a 120mm case fan right over the area you specified. Furmark lasted quite well for about 5 minutes, much longer than before, and then crashed as before. So I think you've hit the jackpot as to what the problem is. Obviously with a higher CFM fan, I could probably have complete stability in Furmark, but then noise would become a factor.

One of the reasons I've gone to aftermarket cooling is to decrease noise, by using quieter 120mm fans. Adding more fans or higher CFM ones just defeats the purpose, only to gain in running 1 or 2 extreme load 3D programs, with stability. So for now I'm just going to stick with one 120mm cooling fan, centered closer to the VRM sensor area, for improved stability.

For me, stress testing with Rthdribl, is fairly good at representing usual temps and GPU load during 3D games, and runs just fine, with no stability problems. So as before, I'll just have to bury this little Furmark problem in the back of my paranoid mind.

All in all, great job though Dnottis. It's nice to have such thorough and improvisational posters around to really take the time to figure out these nagging little problems.

Yea even with an 80mm fan over it I had issues. But with that little 50MM fan right over the top thats when I finally made it past 1 minute and pushed on fro 24 mins at stock speed. I went back and re-ran Furmark @ 790/1000 without an issue for about 10 mins. Its definitely heat and I feel it's U601 which isnt sinked under the stock cooler. I think ATI should've sinked it, it's causing stability issues with the cards.

Thanks - I hate having a problem that I can't figure out, I will work it till I figure something out.

 

[echo75]

guys, i read on the net that it was a driver incompatibility problem. i have the same problem too though use stock fans.

 

[mat9v]

Was it from an official source or one of the theories like the one we have here?

 

[echo75]

Was it from an official source or one of the theories like the one we have here?

i think it was from this - http://www.rage3d.com/Board/showthread.php?t=33927044

 

[mat9v]

Forum speculations (no matter how probable) are only that, speculations. I was looking for official ATI information. While I agree that driver problem is most likely, it may also be application problem (adherence to standard?) or even some hardware problem with the card line. On the other hand while testing stability I ran Furmark for 2 hours with no problem, than closed it, started mplayer and my screen froze. Disk was still working, mplayer was playing movie soundtrack but screen was frozen. And all that @c300/m400. I really have no idea why

 

[rangerone766]

got a version of everest to finally read vrm temps.

By rangerone7669 at 2008-07-28

73c vrm temp at 98% gpu usage. with clocks at 830/1000

 

[dnottis]

my card died last night - RIP. So maybe these issues are a problem with the hardware...