Frustration with the repaste

I repasted my CPU(2720qm) and GPU(6970m) today in my M17x R3. The temperatures are great after. They are 7 C lower at least. However, when I try to overclock(750/1000) the GPU. It is not as stable as it was before, although the temperature never goes beyond 77C. When I check GPUZ, I saw the clock was maxed at 400/900 when the slowdown occurred(about 10 minutes after playing). The slowdown only happens during intensive gameplay. It never happened while running 3D vantage or anything else. I also replaced the thermal pads. Did I do something wrong? I am sure that the power setting is on high performance and powerplay is enabled. Dell stock drivers. Thanks for the help.

 

what bios are you running? try switching to igp and then back to discrete.

 

A04. Switching IGP does free the clock but it slows down to 400/900 after another 10 mins. Acutally its not just Metro 2033. Also happens in Crysis 2. But Benchmarking was fine.

 

This makes me wonder how reliable these temperature sensors are?

 

hmmm strange, have you tried using another driver???

 

I tried M18x's driver. Same thing.

 

i had this problem before, when switching from igp back to discrete but A04 fixed it for me.

can you make a gpu-z and throttle stop log files, so i can see what going on during the time it throttles back.

 

I took apart the laptop and repasted again. Same problem.
Here is the log from GPUZ.
Check the bottom where the clock down clocked to 400/900.Thanks alot.

 

hmm looks like the 0% load triggered it and it never jumped back.

try a non-dell driver, if that doesn't help try re-flashing your bios.

 

Tried both but no luck. I just hope I have not damaged anything when I repasted it.

 

Anyone can help me? It also seems to slow down during Furmark stability test

 

I don't understand the problem you're having here. Did you have good (and constant) results before you repasted? If yes, then how would the BIOS, VBIOS, software, drivers, IGP vs. GPU, etc. etc. have anything to do with your post-repasting results, if all you did was simply replace the paste?

Please advise. I'm very curious

 

Temperature seems to be greatly improved. But it is throttling for no reason. I couldnt OC the GPU at all now. I could play games for hours with OC before. So i think there is something wrong with the GPU.

 

Maybe the memory is overheating but the main gpu die is cooler?

 

Good guess. How am I supposed to know the temp on vram then? I have replaced the thermal pads on memory to Fujipoly 0.5 mm.
Thanks.

 

Gpu-z reports all temps zones , but doubt its getting that hot the Fujipoly is good stuff.

 

The Fujipoly is GREAT stuff. I recently re-pasted 3 video cards, 1 doesn't get any hotter than 57C when gaming and another reaches 62C and holds there. I'm very pleased. I haven't noticed any throttling

 

Hmm, thats more weird then. I've been wondering for 2 days but still have not yet found a solution. It seems to be temperature or power related, cuz furmarkd does make it slow down. And the clock is locked at 400/900. Another note: if my room temp is lower, it throttles slower. ????

 

Update on the issue:
With much lower environment temp(22C). It does not throttle in games anymore, however it still cant run Furmark with OC(775/1000) for more than 5 minutes without throttling. Can anyone run Furmark 1.8.2 with that clock speed for more than 10 mins? Thanks.

 

Hope someone can help me. Thanks.

 

Fan Speed [RPM]
"2" then "0"

GPU Temp. #3
Does it always trigger at 87.5 C? Did you test it again?

 

GPUZ doesnt read the Fan speed correctly.
Temp #3 could hit 95C before repaste but still no throttling.

 

is it only furmark now that causing throttling??? what settings are you using??? if it no longer happening in games then i wouldn't worry about it.

 

Right now it is. Its running @ 775/1000 with Furmark 1.8.2 stability test default window mode. Thanks for reading my post inap. +rep

 

Summary of my recent research:
GPUZ or Hwinfo32 do NOT show VRAM temperature. Temp 1,2,3 are all on die temp. So possible VRAM overheating causes my throttling. 6970m doesnt have sensors on Vram.

Possible causes:
1. Heatsink is not contacting with Vram very well.
2. Thermal pads are contaminated with my fingerprints while I was doing the replacement.
3. This is an ATI driver protection enabled specifically for Furmark burn in test.
4. The fan is not running fast enough as its adjusted according to GPU die temperature which is greatly reduced after repaste.

 

Yea I used a tweezer when I replaced my pads. Pain in the butt but didn't want finger grease on it.

 

I think I finally found the problem.
I examined the dell stock thermal pad. I found it was 1.0mm. However, I replaced it by Fujipoly 0.5mm which people on the forum suggested. This could be a big problem. Can someone else confirm this??

 

i thought u already had fuji poly on it. normally with gpu-z temp #3 should be the vram temps.

have u tried a different version of furmark. try running 1.9.0 @1080p full screen burn in test.

also make sure your power play settings are set to max performance.

 

I do have fujipoly 0.5mm on it right now. But the stock pad is 1.0mm. So I think 0.5 mm is the wrong thickness suggested by people.

 

i use the .5mm too and have gotten much better temps then the stock 1mm dell ones.

 

lol don't worry about that man.
I "replaced" my pads with paste which is way less than .5mm

Just call dell and get a new 6970 sent out due to heat and throttling. It comes disassembled, which is perfect for applying your own pads and paste the correct way from the start.

 

Yeah. I know what you r saying man. But the problem is you replaced all the pads so the heatsink is evenly contacting every component. I only replaced the pads on vram and left other ones alone. This may cause the headsink being uneven on different components and making bad contact.

 

Yeah, my temperature is much better. But as I said GPUZ doesnt show VRAM temperature. All 3 temp sensors are on die. The vram temperature is "VDDC Temp" in GPUZ and it is missing. Could u please try Furmark with 775/1000 OC for 10 mins? If there is no throttling then I need to start over my research.
I had a conversation with rafix73 about throttling here:
http://forum.notebookreview.com/alienware-m17x/570528-m17x-r3-6970m-overclock-8.html
An example of desktop 5870 showing VRAM temps in GPUZ:

Thanks alot.

 

thats for desktops, on the laptop cards the #3 sensor is the vram temps. gave me a few mins and i'll run a furmark for ya.

 

Much appreciated.

 

so what settings do you want for the furmark run???

 

GPU 775/1000 OC. Everything else in Furmark is set to default, window mode, stability test. The version I used was 1.8.2.
I couldnt get through without throttling 400/900.

 

well about 7 min into the run right now, but it looks like its something to do with the stabilty test, after 3 or so mins it drop to 400/900 then after a min or so it goes back up, then down again it 3 mins then back up again.

i suggest using the newer version 1.9.0 and doing the burn in test. it prolly won't throttle back at all.

 

I think I tried it before and it was the same result for me. I also tried OCCT, and it also makes 6970 throttle.
My card is also not stable in Metro 2033 very high settings with 800/1000 OC.
After playing for about 25 mins, it downclocks to 400/900. So I do not think it is a Furmark specific thing.
The weird thing is that whenever I experience throttling, I never had high temp reading in GPUZ. That makes me believe it is not showing VRAM temp.
I have tried many drivers. I was running dell stock before and 11.6 now. No difference.
Thanks for your testing tho.

 

yeah i'm outta ideas, prolly best to ask for a new 6970. to test if #3 is the vram temps, just take off the pads and see if the #3 temps skyrocket.

 

That would be too dangerous...lol.
Btw, What was your temp during the test? Mine never goes over 75C.
The throttling is triggered by something is not shown in GPUZ. If it happens in Furmark, it could happen in games. I still do not think its a bad card. Prob the thermal pad thickness is wrong as I said before.
I mean it wouldn't make sense that you get worse OC stability after the repaste(lower temps). I am pretty sure I've done the pasting job correctly since it wasnt my first time. Something else is wrong.
It wasnt throttling for you in Furmark 1.9.0?

 

i was hitting 82 on #3, 1-2 were about 73-75.

i don't have the throttling problem at all, and i tried about every driver out there. i'm pretty much outta ideas.

 

So you mean it doesnt downclock to 400/900?
Thats exactly the same temp I get then it downclocks

 

nope, no down clocking. my 6970 is flashed to 750/1000.

are you sure you have max performance set in the power play options, and high performance set in your power options.

 

Yeah, its on max performance.
Sorry I made a mistake. I didnt mean VRAM as memory.
I meant VRM (Voltage Regulation Module) overheating. 6970m doesnt have sensor on these.
Source:
http://www.geeks3d.com/20090925/ati...on-against-power-virus-like-furmark-and-occt/

 

As far as I know there are no diodes near the RAM, this is only the MemIO, the memory controller, which is on the die as well. I'm pretty sure about this.

 

then i have failed, jumping off cliff now, hahaha

 

Nooooo, Inap!!! Who's going to beat the benchmark scores of Dane when not you?

@FallenwingsMiku: What VBIOS version do you have? Is it the same as Inaps?

 

Hmm... idk. I have never flashed my Vbios

 

i'm just using the stock dell bios, with modified clocks.