What are the tell-tale signs of bad ihs seating?

I know when I look at the ihs that it is very concave. But, are there any tell-tale signs that I am not getting a good heat transfer fom the core with the ihs? I have a bip.3 with 6 47cfm Loons in push/pull with a storm WB and my 1900xtx is in the loop too. I am able to dual prime (with a 0550 upmw) for only 10-40 minutes at 2.61ghz. I always errors out on Core 0 with a short string error. My temps never get above 37deg c(27 idle).

Thanks!!

Richard Jackson

well a concave ihs is very bad i would say if the ihs is concve either lap it flat or cool it naked (no ihs)

Oh yeh I know its bad. I have heard quite a few stories of these things being concave. I just dont know if I am getting a good interface between the storm and the ihs then to the core. I really want to pop the top but im a weenie. Might do it later on today though. Isn't it 2mm thick foam you need for the pads?

Yea if it is concave then you probably will want to lap it, or suck it up and pop the IHS off of it.

Ok just closed dual prime at 2.5 and its perfect at 2.5.... when I cut both off at the same time the temps INSTANTLY drop back down to 27. Is that a sign of a bad mating of the two? Thanks again.

"instant" temp dropping would NOT point to bad contact (IMO)

"BIG" increases in CPU temp without "similar" changes in water temp would be a sign of bad WB to IHS contact, but without IHS temps (and/or separate/multiple DIE temp sensors) its almost impossible to diagnose bad IHS to Die contact

you may just have found the "stability line" for that CPU.

here is about whats under your IHS:
(a pic of my X2 3800)




GOOD LUCK!!

Those temps are really rather good, but i would be a LITTLE concerned about a 10ºC difference with that watercooling system. Try lapping your core to see, but popping your top is the best way to get results. storms work beautifully on smaller dia area. dont worry too much about crushing it. I used some vinyl tape on the resistors to protect them from my rbx.
In my opinion, your heatspreader is doing its job well.

That pic really makes me want to pop the top off mine... must resist

1stOVERCLOCKER said:

well a concave ihs is very bad i would say if the ihs is concve either lap it flat or cool it naked (no ihs)

Click to expand...

I'd actually say, RMA it...even though he's prolly OCing and is gonna burn the warrenty anyway, I don't think lapping or ripping the IHS off is as great an idea as getting a correctly shaped IHS.

pay close attention to the lower right corner of the "CPU 1/2"...see that chunk missing?
now look over @ the lower left corner of the IHS...see that spec of green?

i was DAMN close to killing this thing (but didnt) and this is the 3rd A64 IHS ive removed (see topless link in sig)

IF ANYONE tries to pop-the-top on their (AMD) A64 CPU....BE (FREAKING) CAREFUL!!!



edit:
White Runner, here are pics of a Venice core:
http://www.ocforums.com/showthread.php?t=399740&highlight=naked

how is amd RMA wise? I got it from newegg on the tenth of last month so I am past my new egg rma time I think. I mean should I try to sell this or is there anything that I can try other than more voltage(already at 1.51volts)? Thanks guys!! I mean should I try to go to 2T to see if its just a bad memory controller or what do you think?

Thanks again!

Richard Jackson

PS... I would just like to hit 2.6 stable if not higher with a 165 or 170 on water.

PS^2... I am erroring out in prime95 with short strings in the small fft tests if that helps a diagnosis. BTW... what do short versus long strings mean?

jcw122 said:

I'd actually say, RMA it...even though he's prolly OCing and is gonna burn the warrenty anyway, I don't think lapping or ripping the IHS off is as great an idea as getting a correctly shaped IHS.

Click to expand...

RMA'ing that chip would be wrong. It works perfectly fine at 2ghz. The speed it was intended for. You are never guaranteed anything when overclocking so anything over stock is just a plus. RMA'ing a bad o/cing chip is morally/ethically wrong imo.
It sounds like he just got a chip that doesnt like a high o/c. Taking off the IHS is really the only option left to try and get a higher o/c.

If the IHS has a physical defect, ie. concave/warped IHS, I really don't think that its an ethics or moral issue; unless he is the reason why the ihs is distorted, but again, I don't believe he is...

Haste,
I disagree with you. If this proc had a perfect ihs and it didnt oc well then I would be wrong rma'ing it on account of a bad oc'ing proc. However if it has a defect physically then I am in the right being able to RMA something with a defect.

Joe Camel said:

pay close attention to the lower right corner of the "CPU 1/2"...see that chunk missing?
now look over @ the lower left corner of the IHS...see that spec of green?

Click to expand...

Looks like you dodged a bullet there.

On RMAing the CPU, the warranty covers defects in materials and workmanship. A concave IHS would be one of those even if the CPU can still do 2GHz.

As usual no one answered his questions, just went off on a side track rambling

Anyway, here are some signs the IHS has poor contact

1.High Load temps (especially on water): Obviously if you have high load temps with water the IHS is a problem. A good way to check this is load prime, and run it for about an hour. Then quickly shut down the PC, and remove the waterblock. If it's cool to the touch then there is little energy transfer to the waterblock.

2.Large Delta between Idle/Load (water): Most people with watercooling have large heatercores that have the capacity to disapate a much larger amount of energy than the cpu can source. Without the IHS temps are usually within 5-10C load/idle. If the delta is large (10C+) then the IHS is a problem.

3.Large Increase in Temps/Voltage (water): Again capacity of watercooling as sink is usually greater than source of CPU. When you scale voltage, and the temp rises fast with water, IHS is a problem. Example with my Opteron 170 temps between 1.30v and 1.58v is 5C at load.

4.One core overclocks significantly better than the other: Both cores are obviously manufactured closely together. Their quality should be about the same in terms of overclockability. If one overclocks a lot better than the other, the IHS contact with 1 core is probably poor. Example: One core on my 170 would do 3000, the other had trouble at 2900 with IHS. Removed IHS, then the 2nd core does 2980 steady.


Should you remove the IHS?
Well that depends on you. If you trust yourself, don't plan to sell the cpu soon, if you can afford to replace it,if you are on water, and the cpu seems to have more potential from removing the IHS, then I'd do it. With air it's probably a waste of time, but with water the results are usually worth it.

How to check if it will help you OC?
A good method to see if removing the IHS would be worthwhile, is to lower the temps so your load would be about what you expect from removing the IHS. This obviously varies by CPU and cooling systems. As an example my IHS was semi steady at 2900. I lowered the room temps so CPU load was around 38C, and it was steady at 2950 or so.

I removed the IHS, and it made it to 2980 with load around 38C with normal room temps.

Indyxc1 said:

1.High Load temps (especially on water)
2.Large Delta between Idle/Load (water)
3.Large Increase in Temps/Voltage (water)

4.One core overclock significantly better than the other: Both cores are obviously manufactured closely together. Their quality should be about the same in terms of overclockability. If one overclocks a lot better than the other, the IHS contact with 1 core is probably poor.

Click to expand...

EXCELLENT points !!!

i like your logic on point 4 A LOT!!

thanks for the info



PS i did bring up large temp delta B4 i sidetracked this onto IHS removal...

Ok... really weird Speedfan readings in dual prime. Does this sound as weird as it seems to me? I am dual priming and the temps go up and down by more than 5deg C. Also, processor usage never changes at 100% on both cores. I dont guess I have ever noticed it if it is ok. Anyone?

Thanks!
Richard

eightballrj said:

Ok... really weird Speedfan readings in dual prime. Does this sound as weird as it seems to me? I am dual priming and the temps go up and down by more than 5deg C. Also, processor usage never changes at 100% on both cores.

Click to expand...

My reported temps used to go down when I fired up the stability tests. At first the temps were rational with my Toldedo. Then I had a problem with the OS, and the reported CPU temp started going down when the CPU usage was at 100%. After overvolting it a bit, the CPU temp stayed about the same. My temps bounced around a 3-4C range. Not as much as yours, but more than they should.

It was on OS or driver problem of some sort, but I never figured it out. I was running win2k and lost an old data drive, the OS decided the surviving drives were new and insisted on "sharing" every file on them, even though the server service was turned off. This resulted in a hive overflow. That's easy enough to fix, but in the process I blew away some of my drivers. After that, and even after I reinstalled all the drivers (I know about), the reported CPU temp was wonky. I didn't try very hard to fix it because I was about to switch to 64-bit windows anyway, but another installation of win2k on the same computer reported rational temperatures, and now xpx2 does too.

Windows repair and upgrade installations didn't do me any good, hence you'll probably have to do a clean installation and reinstall all your software. If you have a lot of aps, try installing windows on another partition first to see if it's worth starting over. Or if you've got one, restore from a backup image.