Help 5870 running really hot

Hey everyone!
I received my second attempt at a 5870m today.
This time it actually booted up! I installed the dell drivers and vbios just in case....

Right now I am idle at about 62c but when I run a game like world of warcraft I am going all the way up to 95c!

I attached a gpu-z ss for reference. Bad paste job or something else going on?
When when I scroll on my notebook display sometimes I am getting a strange flickering on my external vga monitor.

 

View attachment 105114

Doing the catalyst render test I am up to 109c but feeing the air out the back of the pc it doesn't feel any hotter than when my 4870 was at 70ish.
Can it be that the temp is being misreported?

The keyboard is warm but I would expect a true 109c to be damn hot.

 

When you received your 5870M, did it already have the heatsink on it, and did you just plug the card in without taking off the heatsink and doing a re-paste?

If you did, these temps are most likely from the thermal paste, if you don't use your computer for long periods of time then the thermal paste ends up hardening and will not conduct heat properly. If you bought the card with the heatsink, it was probably laying around unused for some time, so this has happened. Just remove the heatsink and do a re-paste, IC-Diamond is my favorite thermal paste and I recommend using that to lower temps, but really a re-paste with any thermal paste will lower your temps.

 

no.. the heatsink and the xbracket were missing.
I repasted with artic silver 5 (rice grain in the center not smeared the whole die)
and used frozencpu thermal pads on the ram.

On one section of the ram where the heatsink met there is a sticker "void if removed" with an aleinware head on it and a barcode. I put the thermal pads over that.

Also there was not a translucent orange piece of plastic around the die like on the 4870 if that means anything.....

 

Stop whatever you are doing, the card has been tampered with!!!, the speeds should be 700/1000 ur's are 866/1232.
It looks like it's been flashed to those speeds, that explains the increase in temperatures, as for the idle temp, i guess the card doesn't go to the 100/150 profile or the paste job was bad (more likely the latter)

The flickering might be because of too high speeds or because of switching to 2D clocks.

Mine idles at 40°C and at full load barely reaches 70°C at stock speeds.

 

I suspected something like an overclock flash but I reflashed it with the vBIOS offered on dells site for the 5870.
As reported the BIOS is 012.020.000.027.037640

But it was this before and after the dell flash.
Assuming for a moment that the card was reflashed to run at a higher speed.... is it possible to restore it via software or is this card a loss?

strike 2

 

It is possibile that the GPU-Z is reading clock wrong, what is the version of GPU-Z?
Also try HWinfo32/64 and check what clocks does it report.

Here is my gpu.

 

its GPUZ 0.7.4- the latest version.
My 4870 it was reporting idle at 56 and 71ish under load.

After trying wow I accept it is probably running very hot. at minimum settings I was at 7fps.
I will check HWinfo when I get out of work-

 

Ok, let us know what how it goes.

 

Yeah I see the increased clock speed, memory speed, texture filter, and the bus is 2.0 instead of 1.1
I would think this all would be controlled by the vBIOS but we are running the same version.

very strange. Will see what I can find out

EDIT:
I guess I should also check CCC to see if it is overclocking in software.
I don't know why on a fresh install it would come out of the box with over drive enabled but will check that too.......

 

I didn't even notice the overclocks . With a new version of CCC you should have AMD Overdrive available for you to underclock the card back to stock. If you don't have Overdrive in CCC, then install MSI Afterburner and from there you should be able to bring it down to 700/1000. If the temps go back to normal at 700/1000 then you can try overclocking the card a bit if you're up to it.

 

I would tell you to download and install Msi Afterburner to see if you can bring down the clock speed because MSI afterburner also has a built in temperature monitor so you can tell during the game how hot your card is or how it is getting during a game . Plus you can set the card back to stock and save the profile with MSI afterburner also .

 

HWiNFO confirms the same readings.
The dell drivers don't seem to have overdrive- I tried downloading the 5xxxm series drivers from ati but they wont install.

I downloaded MSI afterburner and can set 700/1000mhz but everytime I hit the "apply" button it jumps back to 866/1232!

 

Ok some progress.
As you can see... the default clocks are still 866/1232. I have been unable to bring the default into spec no matter what I do!
I ended up using a program called AMDGPUClockTool_v0.9.26.0
This allowed me to set 700/1000. Strangely enough, after that stuck, afterburner also stuck.

After rebooting it still boots up 700/1000 default 866/1232

Wow will spike it from 63c idle to 79/80c but it holds steady at that temperature now.
I will try to repaste at this point.

 

Have you tried the custom vBios for the 5870 that maxslo has made? A link to them are here: http://forum.notebookreview.com/alienware-17-m17x/734549-should-i-buy-m17x-r2-5.html#post9421159

Only put the REV02 and REV12 onto the boot usb, because the REV0 and REV1 are just the original ones.

I use this vBios and my 5870 runs a lot better than it did with stock vBios - so there's a chance these may completely restore your card. Where did you get this card btw, I've never seen a 5870 with default overclocked settings

 

this is a rev02 btw >.>
This was an ebay card.... it was only 100bux and has a 30day warranty or so they say. "seller refurbished". I dont know what that means though. Maybe the vbios chip swapped? Maybe something melted and had to be replaced? a missing xbracket would normally be a sign that it melted off like the bad card in the oven.

it is kinda strange how I cant downgrade it with the dell vbios on their support page specifically for this. I will try the other vbios to see.

I am about 79c now with medium gaming. i think all the new paste and pads need to cure a little anyway.

 

The bus 2.0 is ok, that's because you have i7 and i have i5 CPU

You should try my Vbios as akazeth said, and then update to the lastest drivers directly from AMD, that should make sure that this is not a software error.

Also this Vbios will enable you Overdrive in CCC, so you can set it to default from there and see if it defaults to 700/1000 or to the faulty speeds.

What are the load temperatures at full load now with 700/1000? Anything below 90°C is an improvement

 

I followed the linbk- the expand and reboot zip was a bad link but I downloaded the REV12.rom from an earlier post.
I used ATIFlash to load it (though the new and old version matched exactly).

I tried to download the beta drivers but the ati site wont allow me to do it and if I download the auto detect program, the driver I download says it is incompatible- I attached a picture. I don't really get it, I am using win7 64bit
I am able, however, to download the beta so I am currently downloading catalyst 13.11 beta 9.2.

Lets see how this install goes-

 

Ok so here is an update. The REV12.rom seems to have fixed the card. The default clock\memory is 700/1000
On bootup GPUz is reporting I am running at 430/615

The voltage going into the card is .9v

I installed the beta drivers but it seemed like strange things were happening on the VGA second display. When I have a window opened on it and mouse over I got a strange flickering or ghost boxes flickering at the bottom of the screen.

I ran Display Driver Uninstaller and cleared everything out.

I think reinstalled the OEM dell drivers from the site and rebooted.
The system still came up running 430/615 and still had a strange flickering. MSI afterburner, still being installed seemed to be holding the card to 430/615 so I cranked it up to 700/1000 and the change was reflected in gpuz. At this point the flickering also stopped. The voltage going into the card is still .9v

Under the beta drivers I was running at 54c idle. With the dell drivers I am running at 64c idle.

I am tempted to run the display driver uninstaller again and try the beta again (knowing it might have been afterburning keeping me down->causing the flickering) but I am not sure if I should hunt for updated drivers. I don't know why ATI is saying there are no drivers I an download.

Where should I go from here??
Long story short I think the card is salvageable.

 

I too can't install drivers off the AMD website for my 5870, however, I can install the latest AMD drivers if I download them off Guru3D, here's a link: Videocards - ATI Catalyst Windows 7 8 and Vista

I think it's just AMD's auto updater that conflicts with the card.

 

Thank you for the link:
13.9 redirects to AMD (cannot be downloaded)
13.4 does not work very well... flicking
13.11 redirects to AMD (cannot be downloaded)
13.11 beta 9.26 (which I ended up grabbing off the ati site after being redirected) installed and seems to be stable. No screen flickering.

Uninstalled afterburner and ati tools.

At the moment I am holding steady idle at 62c 700/1000 gpu/memory .9v
under load 1.05v looks like I am spiking to 91c but dramatic changes from 75ish to 91ish with the average in the mid to low 80s

 

With my VBios and 13.11 beta 9.26 do you have stock clocks 700/1000 or did you have to set them in Overdrive?
Does everything work in CCC as it should?
For the dramatic changes it's either the thermal diode dying, program reading wrong, or heatsink not making good contact

I think it's the heatsink because the GPU idles at 20°C higher than mine.
What is your ambient temp?
And at idle are you sure the clocks are 700/1000 when voltage it 0.9v in the sensors tab?
The clocks should be 100/150 when voltage is 0.9v, because i haven't modified the voltages for stability purposes.

Can we have a SS of:
1. GPU-Z summary tab
2. GPU-Z sensors tab, when laptop is idle 10minutes (showing minimum readings)
3. GPU-Z sensors tab after 1 run of 3DMark Vantage (showing maximum temperatures)

But i'm pretty sure you will just have to do a repaste

 

Yeah it was getting way to hot yesterday. Not 111c hot but low 90s and most of the graphic settings were medium-good. I don't think it was a program\sensor misread... the thing felt very hot after a few hours.
An answer to some of the questions:

Everything appears to work in ccc now. and the card starts up in both 700/1000 as well as listing those values as default. I uninstalled afterburner and have not made any adjustments in overdrive so 700/1000 is really what the card is running at now.

The ambient temp is about 78F
I will get the screen shots as soon as I am back in front of the PC again.

I figured I would have to do a repaste. The thermal pads I have on the card ad the moment are .5mm clay-like substance from frozencpu

Fujipoly Premium System Builder Thermal Pad - 1/4 Sheet - 150 x 100 x 0.5 - Thermal Conductivity 6.0 W/mK - FrozenCPU.com

I have this over the ram, and any other component on the card that appears to make contact with the heatsink. On the die itself I am using artic silver 5. I ordered a new thermal pad 1.0mm thick that can be compressed to .5mm I believe it was this stuff:

Phobya Thermal Pad XT 120mm x 20mm x 1.0mm - 7W/mk (V-Regs, RAM, Ramplex, Koolance, MIPs, Innovatek) (19100) - FrozenCPU.com

It will sit higher and hopefully make better contact as it compresses. On my m15x there was really spongy thermal pads on everything but I could not tell what material it was.

 

Thats great, so now only thing left to solve now is overheating I forgot do you have M17x R1 or R2?
High grade thermal material they have 17W/mK but the problem is it doesnt compres very well :/
I've neve swaped Thermal pads beore so can't realy tell how thick they should be :/
I use MX-4 or ICD-7 for the GPU, i've been getting the best results with those.

 

Alright, so it looks like at idle the card runs at 405MHz at .9v and jumps between that and 700 depending on what I am doing. The fan seems to always be 30% but in HWiNFO64 I set the fan as high as 5500rpm before and the card still stays around 90ish c in game...

This is an R2

I have attached pictures of GPU-z after idling for a while both summary and sensor. I am in the process of downloading 3dmark to run those tests.
EDIT: I did not have MIN temp selected- it is 53c
EDIT2: I had GPU-z opened on vga monitor when I ran 3dmark. the picture I attached is a shot of it a minute into the test

 

Oh i'm sooo dumb xD ofcourse it won't go to 100/150.... it's because you have another display hooked up
Unplug the second display and let it cool off 10min, then you should be more towards 50°C idle

Have you repasted yet?

 

I have not repasted yet because I am going to wait for those thermal pads to arrive. They should be here in a few days. I pasted twice to date: once on the initial install where I only used a rice grain dot in the center of the die and once after the bios fix when I was running at 700/1000 where I did a thin layer over the entire die. In both cases it was the same thermal pad material used and nearly the same temperatures: 97c vs 111c (though the latter was probably due to having it OCed). I would have pulled the tabs from the m15x but I replaced those too with this new material! :O

I will see how this runs with one monitor. It will be tricky to get a screen shot of the GPU-Z with 3dmark because as soon as I tab out the temperature could drop 20-30 degrees! I guess that is why you suggested I set it to "max" so I can see the highest temperature without it having to be in real time. Let me try this out and I will report back.

 

Ok I tried the tests again only using the primary display (vga unplugged).
The new idle image is after 10mins of nothing- 45c
The 3dmark image is after 1 minute of run time. I switched everything to show max temp. Looks like I am all over the map:
45c idle to 95c back down to 49c a few mins after the test.

I am not sure why at 95c my fan is only staying a max of 30% either.... that might be something I suppose

 

Ok, the idle temps are much better now ,
About the fan, AFAIK the reading always shows 30%, use HWInfo to see the real RPM of the GPU fan, and let us know if it increases with the temperature

 

Ok so as an update:
Last night I opened up my m15x and took all the stock thermal pads off the circuitry of the GTX460M. I transferred them to the 5870M of the m17x. I repasted. The vram chips I still had to use the same compound material I posted earlier as the replacement is still on its way. This time I did not use Artic Silver 5, instead I used T-G2.

Still getting up to 98ish c but idle in the high 40s and with just streaming video am at about 65.

So I did a little more research and I am running out of ideas. I placed another order to frozencpu for ICD7. The pads should arrive Saturday and the paste Monday so the next repaste attempt will be sometime on Monday. I am wondering if it is possible that the .05mm compound material on the ram now is actually lifting the heatsink off the gpu a little. Is that possible?! I would think it would get much hotter if that were really the case.

 

Wich circuitry?
I only use what looks like 0.3mm Thermal pads on VRAM, and Paste on the GPU die,
it's probably something lifting the GPU die away, just as you said... did you check if the fan speed is increasing with temperatures?

try removing the VRAM T pads to see if you are getting more reasonable temperatures then...
Just for 1 3DMark Vantage run to measure max temp and rule out Thermal pads as the cause, the card will be fine, because Vram chips don't warm up that fast and don't produce that much heat anyway....

 

good progress!!!

Ok once again I opened her up. This time I did 3 things.

1. I repasted with Artic Silver 5 instead of TG-2 with the method described to used with ICD7: I did a bead about 4mm long and did not spread it. I let the pressure from the heatsink distribute the paste and I am sure there is run-over but who cares at this point.

2. I totally removed the .5mm thermal pads from the ram chips so it is metal on ram contact but this ensures that the heatsink makes firm contact with the GPU.

3. I removed all the 1mm pads I carried over from the GTX460M. (in answer to the previous question, Alienware used this all over the place under the heatsink. They seemed to put it on all the resistor clusters. When I took the heatsink off the GTX460M nearly every component under the heatsink had a thermal pad on it).

Anyway, the results. After 10mins of idle and no VGA plugged in I was at 40-41c. I ran 3dmark vantage the entire way through: 2 video tests and 2 CPU tests. I posted the results but the max temperature was 74c!!!!! now it is back to 46-47c.

I am not totally sure what the ultimate fix here was but I suspect the .5mm is too high and the heatsink was off the gpu die. There was like no compression with that clay too. So for reference .5mm is probably too thick to be between the heatsink and the ram. I wonder what the correct size is!! or if I should be rushing to buy it right now.

All in all big changes here! thank you for all the help!!!! between the over clocking rom and this problem I was about to say this board was no good....!

 

nice temps again !! good job man !!

i have to dis-agree with u about the last of part 1.
the paste u used (aarctic silver 5) is actually conductive..
so u DONT want this paste on anything other than just the core..
i used the same on my gpu and cpu of my alien, its a good paste, but the risk is shorting out components, again, because its conductive..

i hope u didnt put 2 much on it, but only time will tell... its not a big problem with ic diamond or any other (as a basic paste i alwas use CM high performance thermal compound) because they arent conductive

 

Ok so looks like I am really fixed for now!
usual settings in game for 4 hours: 68c this is with a vga plugged in 60fps but the card was clocked at like 500 probably due to the monitor.
Then I unplugged the vga and turned all the settings up and took off vertical sync (so was at 100ish fps) I didn't go over 76c gpu clocked at 700


Big difference from 111c!!!!
So the fix for this was:
ROM
Drivers
Repasting
Heatsink making good contact

At this rate I may not even repaste this card with ICD7, I hope nothing shorts out if there was over flow from Artic5 but if there was nothing I can really do anyway... might do more damage cleaning it than what I would be preventing

On to project phase 2!!!! card 2 and a crossfire cable otw

 

Awesome , but do try to get correct size thermal pads later on
I've once on a G73JH cut the thermal pads horiziontally (from 0.5mm to 0.2mm) to increase the contact between GPU/HS with a cutter, brought temps way down like in your case (This requires a LOT of patience though xD)

IS5 is not conductive FFS xD why do people keep saying that, it's only slightly capacitive artic silver 5 IS NOT ELECTRICALLY CONDUCTIVE!!!

Glad to be able to help , if you are looking for a second 5870M i still have 1 with heatsink on that i have no real use for, since my mobo i clearly not capable of running a CF, (that or the CF cable is bad, really dont have the time and money to find out wich it is)

 

I already did the ebay thing last night as soon as I saw this worked :/ but if his doesn't work I would be on the market for sure!!!
I am not sure what prices are even good... this current card I got for $100 shipped- but it didn't have a heatsink or xbracket. The next one has a heatsink and all but it was $150.

But wow this has been some adventure. I would not have been able to get this far without this forum or the support! I am totally grateful for everything!

 

Nah too bad xD, looks like i'm stuck with it for a while longer
That prices seem about right (I think 5870M should go for about 80$ wihout HS and 110$ with it
You are very welcome just remember to share the love now

 

wow I love to play around.....
Ok so as a further update-
Both my frozencpu orders arrived today despite being placed days apart:
The ICD7 and the Phobya 1mm thermal vram pads.

I opened the system up again and as suspected I got Artic Silver 5 all over the place. The die was completely covered but it also really overflowed on all sides and was quite pooled up.
I cleaned it all up, applied the phobia pads and a bead of ICD7 (I am using a bead not a "pea" because the die is not a square- still following what they recommend on their site).

Closed it all up and was idle at 44c. Then I ran 3dmark and I maxed at 83c! .... Note: before this I maxed at 74c.
So I opened it all up again, cleaned up and reapplied the ICD7 and totally removed the thermal vram pads. Closed it up and idled at 43c. 3dmark: 72c

So definitely for reference beware anyone who may be in my same boat in the future.... vram pads! 1mm is way too thick and these phobia pads compress up to 50% so consider them really .5mm in this application. Still way too thick.
They sell .5mm pads that can compress to .25mm which might be ok but I do not even want to try at this point. Running without thermal pads on the vram seem to be acceptable. I only went the route of trying to put them on because that was what alienware had on my 460m. But that material was much more spongy despite looking like it was 1mm. also I guess I should take into account it is an NVidia card and a heatsink from a different system. Maybe there is a fundamental difference in how the m15x and m17x heatsink sits on the card (obviously there is a difference with the heatpipes and all but I assumed that they both sat the same).

It will be interesting to see what temperatures are achieved once this paste cures. I didn't wait long enough with Artic Silver 5, I think it needs 200 hours. ICD7 needs something small like 2 hours from what I have read.

 

Well the temps should drop by max 1-2deg in load, so don't expect too much

But i stil wouldnt run completly wihout thermal pads xD, just to be safe