Removed E2160 IHS and temps went UP!

[palisade]

First, I have to thank you guys for all the helpful replies/advice to my first post. Now for the second...

I got my E2160 to 2.9GHz (325x9) stable on my P5K-VM, which seems to be lower than average compared to others here. Temps were peaking 64-66C using a Zalman 7000B-AlCu. I suspected heat was keeping me from getting stable above 2.9GHz. I had already lapped the bottom of the HS so my options were (1) lap the IHS on the E2160 or (2) remove the IHS and put the HS directly on the core. Since it had the potential for a larger temp drop, I opted for #2.

So, following the path others had blazed I removed the IHS....and managed to score the top of the PCB in 3 places with the razor blade. Believe it or not, the thing still works.

Anyhow, the IHS adds ~1.5mm in height to the CPU. This is important because when you remove the IHS the top of the core sits below the top of the load plate that holds the CPU down. The Zalman's wide base hits the load plate instead of sitting in the recess on the core. Fortunately(?) the base of the stock HS was small enough to clear the load plate and sit on the core.

Here's where things get funky. Just running at stock speed (1.8GHz) the E2160 idles ~31C. Under load it peaks ~60C. When I bump it back up to 2.9GHz it peaks ~72C! WTF?

Couple of notes:

• I used AS5.
• There is a delta between the 2 core temps of about 4-5C where it was maybe 1-2C before.
• Because the core is lower relative to the MB the stock HS does not press down as hard on the CPU.
• I disabled "CPU Q-Fan Control" in the BIOS to keep the HS fan running at full blast.

Any ideas on what I can try to get the temps down? (I can't use a large HS because I'm using a SSF case.) I thought about getting a decent copper core HS, but for the money I'd spend I could add another $20 and get another E2160 and hope for one that OCs better than the first one (without having to remove the IHS).

[NickS]

This is because the stock HS/F doesn't apply enough pressure to the CPU when you remove the IHS. Try putting the case on its side so the board is horizontal, or try pushing on the HS/F when your watching your temps. See if theres a change.

Try and see if you can find an original Thermaltake Big Typhoon. That might work, but be warned, it's huge Mounting them on naked AMD's was no problem though.

[Bail_w]

You need to remove the plastic holders for the stock heatsink then use your own customize mounting screws to apply enough pressure on it. btw, i am using stock heatsink with my own mounting on my IHS less E2160.

[palisade]

T...try pushing on the HS/F when your watching your temps. See if theres a change.

For grins I tried pushing down on one of the pin posts on the stock HS and one of the core temps shot up to 81C! I'm thinking that narrows the problem down to pressure of the HS onto the core.

[palisade]

btw, i am using stock heatsink with my own mounting on my IHS less E2160.

I hate to ask, but do you have a pic of the HS mount? I'm trying to figure out what I need to adjust on the mounting bracket. From what I can tell I need to take off ~1.5mm (the thickness of the IHS) from the bottom of each HS post. That would lower the bottom of the HS to put more pressure on the core. Does that sound right?

[ixtapalapaquetl]

I cut off the plastic mounting mechanisms completely. I then inserted some long screws through the mobo and HS and topped off with some thumb nuts. You should probably use plastic washers against the back of the mobo, or even better, use a backplate. I waited until I had a backplate. I can now get as much pressure as I want.

By the way, it is a very simple matter to remove the retention bracket. After a little twisting, it comes right off. You could then add four layers of electrical tape on either side of the die to form a shim before mounting for some peace of mind - less worry about smashing a corner while mounting.

[hollo]

if i were you i would -
1) remove the top of the CPU locking mechanism so you can use the zalman again (you just need to bend a small tab around the "hinge" area by about 30 degrees and it'll come off cleanly and easily)
2) carefully scrape down the top of the CPU with whatever instrument or sandpaper you need to so all the solder is gone, because it forms a big curved hump which is naturally terrible for heat transfer (you might have already done this to your satisfaction :P)
edit: oh, the 2160s don't seem to be soldered to the IHS, hah
3) cut out a thick steel backplate for your motherboard with 4 holes 72mm apart (which is the distance between the holes on LGA775 mobos). drill a hole through the very centre of the thick steel plate and put a small bolt with a curved head through it so the thread will be pointing away from the motherboard, chop off the excess protruding bolt - this is the spacer that will transfer the mounting pressure directly through the centre of the CPU-die to the HS on the other side. but any small hard object in the correct position will do the same thing, a ball-bearing sitting in a hole smaller than the diameter of the ball-bearing would be an ideal spacer

cut out a smaller steel square about 50mm by 50mm, maybe a little larger. put a decent protective layer of foam or rubber on the 50mm square and place it against the back of the CPU socket, then mount the zalman with bolts to the motherboard backplate you have made. place the bolts so they point up through the motherboard, do them up TIGHT and measure the length of the protruding thread on each bolt (or even better - count the number of exposed threads on each bolt) and get all 4 bolts the same for nice even mounting pressure

overkill? maybe
that's just an idea to get the force vectors of the mounting pressure pointing where you need them (through the die, not around it)

but as you know from pushing down one of the push-pins... the forces are very poorly applied on a standard LGA775 heatsink mount when you're dealing with a thumb-nail sized CPU-die. this is less of a problem when there's a IHS with a large surface area as the contact surface, or on AM2 mobos where the heat-sink mounting mechanism has a V-shape that pushes directly onto the centre of the heat-sink

ok, so you wanna get another degree or two temperature drop for cheap or free ?
http://www.xtremesystems.org/forums/...d.php?t=173091
imo the way to do this with a bare-die would be to use a big blob of TIM and put a foam cutout that'll sit around the edges of the CPU-package and hold it all in - i used silicon TIM in ^ thread because metal-based compounds are slightly conductive and there is componentry on the surface of bare Athlon 64 CPUs, however with a 2160 you won't have that problem

and here's a pic ~

GL HF

[palisade]

Okay, guys, many many thanks for the advice. Being the meticulous type, I measured the thickness of the plastic standoff under the feet on the stock HS (0.120") and the thickness of the IHS (0.079"). Without the IHS I needed a 0.041" standoff for the stock HS instead of 0.120" to get the same amount of pressure on the core as would be on the IHS if it were installed. Hope that makes sense.

Anyhow, I removed the plastic retaining pin assemblies from the HS and fabricated some cardboard standoffs to replace them (just three thin pieces of cardboard glued together - ~0.070" thick). I reused the knurled thumbscrews from the Zalman bracket to attach the HS to the MB. (Pic attached showing one of the feet with the cardboard standoff.)

Without having time for the AS5 to break in, temps at 2.9GHz are ~30C (idle) and ~62C (under load). I figure by the time the AS5 settles it will be about 5C lower under load than what it was with the IHS + the Zalman - not quite as low as I was hoping for, but at least lower than what it was before I started. If I can get it to 3.0GHz stable I'll be a happy camper. Oh, and the cores are now back within 1C of each other, too.

Again, many thanks for the help.

[ziddey]

you sure vcore is where you think it is?

As everyone said, after you remove the IHS, you shouldn't be using the retention mechanism anymore. It pops off easily and can be put back on again later. Hardmount is looking good Keep the pressure even across all 4 and you should be in business

[palisade]

you sure vcore is where you think it is?

I have it set to Auto in the BIOS. CPUZ reports 1.312V idle and 1.320V under load (I did the VDrop mod and somehow ended up with a VBump ).

As everyone said, after you remove the IHS, you shouldn't be using the retention mechanism anymore. It pops off easily and can be put back on again later.

The retention mechanism (I think it's called the "load plate") wasn't getting in the way of the stock HS so I didn't take it off. I don't think it's affecting the performance....but I could be missing something.

Just an update: I bumped it up to 3.0GHz (333x9) and got a spontaneous reboot after about 15 min with ORTHOS Small FFTs. CoreTemp log said the cores never went over 62C.

Guess I'll be fiddling with the VCore now after all.

[Corsa]

Time to set vcore manually, start with 1.320v and keep lowering till ur unstable...less voltage = less heat.

[palisade]

This things gonna drive me nuts. First, I noticed a difference of -0.062V on average between the VCore setting in BIOS and the VCore detected in the hardware monitor section of the BIOS. For example, a VCore setting of 1.4000V in the BIOS is detected as 1.3360V in the BIOS hardware monitor. (CPU-Z reports the hw monitor value.) FWIW, I've also done the VDrop (VDroop?) mod for the P5K-VM (which has the odd side-effect of bumping my VCore up about 0.008V under load).

At any rate, I had to bump the VCore setting in BIOS up to 1.4250V (1.360V detected, 1.368V under load) to get the E2160 stable at 334x9. (I say "stable" meaning I ran ORTHOS Small FFTs for almost 24hrs before I got an error.) Temps averaged 66C and peaked at 68C at some points. I'm hoping the AS5 settles in a little better and gets it down to about 65C avg. 65C was sort of my comfort threshold for heat.

At this point I'm guessing I'm hitting the limits of my CPU. This is rather disappointing as others are getting 3.2-3.4GHz without having the IHS removed. I suppose I could upgrade the cooling but, again, it's a $70 CPU - at this point I'm thinking it would make more sense to roll the dice and get another E2160 that may overclock better.

[Leeghoofd]

Man don't waste ya time on the stock heatsink, these things run bloody hot once you start pushing them... I had an E2140 under WC with 100% OC and it went 65°C on the cores ( that's with a triple 120.3 rad , Dtek Fuzion and a Laing Ulra pump)...