CPU Overheating HSF Questions

[planetnine]

Hey guys i have a new system with a q9550 cpu. I bought this heat sink:

http://www.xsreviews.co.uk/reviews/cpu-coolers/asus-triton-79-amazing/

q9550 @ 2.83ghz
Vista 64
8gb ram
veloc HD
4870x2
Stacker 830 case with 8 Fans not including the one included in my Heatsink.
1 standard 120MM Fan
2 Delta 200CFM fans
5 Scythe 150ish CFM fans

So the Air is really pumping through this case at a high rate and its really loud.

Here is the problem: Cpu runs at about 60-65C idle.
If i run a game the system shuts off in about 1-5 minutes.
I installed Real Temp to monitor my temperatures and ran Prime95. The temperature immediately start going up from 65-130+ within 10 seconds of running prime95 then the system crashes.

I think the Heatsink is decent enough that this should def not be happening.
I applied thermal 5 in a small thin line across the middle of the cpu and then pushed the heat sink on top of that. Rotated it a bit then pushed down the locks. Did i install it incorrectly? Im guessing i need to redo it but i didnt think i did it wrong from the beggining.

The point of contact for the heatsink is black so i guess its painted and its also a bit rough. I would think this is not good for a heatsink so im thinking about trying to lap it but it will be my first time to do so.

Any tips or help with my current problem is appreciated.


thank you,

 

[Conumdrum]

I think the locks aren't on there right. Did you inspect the pins from the bottom of the mobo to ensure they are on right?

I'd do that before you taost something, idle should be much less.

 

[Superhal]

It's a flawed design. The fan in the middle design just doesn't work well. Try mounting the fan outside one of the heatsinks so that it blows through both, rather than pushes one and pulls the other.

From frostytech.com: http://www.frostytech.com/articleview.cfm?articleID=2291
"While the Triton 79 Amazing heatsink runs fairly quietly, it's unable to break out from the crowd. It offers good, but plainly average thermal performance for anywhere in the neighborhood of $70... far too expensive for this class of heatsink. By comparison the Xigmatek HDT-S1283 is half the price, runs 10°C cooler and just a bit louder. "

Also, almost everybody here says the pushpins are worthless. try using a backplate. i'd lay 10:1 odds that one of the pushpins popped out.

if you decide to return it, i wrote a little buyer's guide:
http://forums.amd.com/forum/messageview.cfm?catid=11&threadid=100223&enterthread=y

 

[planetnine]

Hi Superhal,


Thank you very much for the helpful info. I read your article, great stuff. I think im going to go ahead and purchase the sunbeam corecontact or the therm ISF. I am going to try and reapply the HS and Thermal soon though and see if that helps. I will also look at moving the fan but it might not be worth the effort to deconstruct the HS to remove that fan. It is really well placed in there.



thanks again!

 

[iluvgillgill]

change the fan from stock see if it improves as 1300rpm is barely spinning when considering its between 2 bunch of fins!><

 

[iluvgillgill]

ah!sorry just read the fan cant be swap but you can try ziptie a fan in the front if to improve airflow throught the fins see if temp drops.

but i think maybe its the thick black lacquer that stop heat from transfer heats to the heatsink

 

[Superhal]

you can see tests with fan placement with that design with the thermalright ifx-14.

you can also check with push-pull multifan setups on the Silenx ixc-120ha2. pulling only affects it by 1c.

i want to see the silenx with push-push.

some square designs also allow side mounts...i wonder what that would do.

i'm sure that dual fans should give at least 10-20% better performance than single fans (akin to taking off a side panel and blowing air in with a desk fan or using a 30k rpm fan), but nobody's come up with a good mounting system yet.

but, the severe problems the OP posted are very likely to be caused by malfunctioning pushpins.

the thermalright ultra 120 extreme is also lacquered.

 

[planetnine]

so is lacquered a bad thing im guessing because its one more layer that heat has to go through? Why do HS companies do it then?

Im going to check the pins again tonight and i will post again later.

 

[Superhal]

i don't think the lacquer affects performance that much (+30-50C.)

These are what I would consider "small effects" that affect performance by 1-2C:
direct touch heatpipes vs covered heatpipes
fin surface area
lacquering
2nd fan

These are medium, maybe 2-5C difference
TIM quality
90mm fan vs 120mm fan
perpendicular mounting vs parallel mounting
low cfm fan vs high cfm fan
lapping (huge variance here though, +3 to -10, so average ~5c)

Large, 5C+:
Size
Case airflow (probably one component's exhaust is another component's intake)
pressure
TIM quantity

Severe, 20C+ or thermal throttling/shutdowns:
Fan failure
Overvolting/Overclocking past hsf capabilities
MOSFET failure
improper installation
mounting failure
Blocked case venting (creating a closed loop, usually by putting the case too close to a wall or in an enclosed CPU desk.)

At 60C+, you're looking at no hsf installed or something melted.

 

[planetnine]

ok here is an update:

I removed the HS. The paste was in a circle the size of a quarter on the HS and CPU. It wasnt touching the entire surface area. 1 of the pins seemed loose too. So i thoroughly cleaned the HS and CPU off with alcohol swipes and qtips. I then proceed to take the ASUS HS apart to see if i can replace the fan or move it. I took the top gold plate off and yes i can remove the fan. However after looking at my layout in the case, i can not place the fan anywhere else. The fins only shoot air from right to left-left to right. The fins are closed on the north and south if that makes sense. There is no point on putting the fan on the left side of the fins because that makes it only 1 inch away from the back exhaust fan. It wont fit on the right side of the HS because my RAM is right next to the CPU.

So this time i reapplied the paste evenly across the entire cpu in a thin layer. I then attach the HS and applied close to scary force to the entire structure while pushing down on the pins. I locked them all in place while still continuing to press scary hard. I let go after about a minute and it seems more solid than before. Everything put back together i boot into vista and use 3 programs.

RealTemp 2.7
CoreTemp
and Prime95

First readings IDLE:

CoreTemp
core1: 53
Core2: 47
Core3: 43
Core4: 40

RealTemp:
Core1: 48
Core2: 43
Core3: 39
Core4: 36

Not sure which program to believe? Real Temp runs 4-5 degrees cooler but im more inclined to believe the higher readings.

Prime 95 on max 100% full load on all cores:
Coretemp
Core1: 65-68
Core2:59
Core3:55
Core4:51

And Real Temp read about 5 degrees less than that for each core.

This still seems too high? Is this dangerously high that i shouldnt be running any games until i get a better solution?
Superhal what are your thoughts? This is still 10-15 degrees better than before and i couldnt even run a game that long or prime95 for longer than 5-10 seconds before the cpu temps hit 130Celcius and my system shut down.



Thanks again!

 

[Superhal]

yep you look like you are maxxed out for that cooler. i'm not too familiar with intel chips, but who is? (LOLOL) but you seem to have acceptable temps. wait a few days, the temps should lower as the paste cures.

did you check which direction the fan is blowing in? make sure the fan is blowing the same direction as your front case fans (blowing towards the back of the cpu.) The best way to tell is to get a small piece of a paper towel and hold it in front (or back, if the case fans get in the way) and see which direction it's sucked. tissue paper is lighter, but seems to be dusty. usually, it's hard to tell fan direction by holding your hand there.

with that many fans, you should get a lower temperature.

hopefully, somebody should chime in about intel chip temps. usually, for amd, there's only one temp reading for all cores. but, from what i've read, under 70C (without bsod/freeze/etc.) is fine.