Do the core have bad contact with the IHS?

sablabra · 15 Aug 2006 at 17:32

Hi. I've been suspecting the IHS on my x2 3800+ to have bad contact with the core for a long time. Core 1 always fails before core 2 in prime95, so I now checked the temp in everst ultimate. It turned out core 1 is a lot warmer than core 2. Do you guys think the core have bad contact with the IHS? I've been wanting to remove it for some time, but don't have the guts to do it, nor do I have anything to train with

sablabra · 15 Aug 2006 at 18:09

I have a swiftech Apogee, do you know if it will fit nicely?

sablabra · 15 Aug 2006 at 18:18

Cyber-Mav said:
if it uses nuts and bolts to attach to the cpu then yes it will work. if it uses the pressure clip thingy then you will get problems.

Nah, it's screwed in place with a spring inbetween...

But I'm not sure if I want to remove it... Anyone else here who have removed their IHS from x2 3800+?(or similar cpus?)

sablabra · 15 Aug 2006 at 18:29

MikeTimbers said:
Big thread on this stickied in the CPU section here.

It really isn't too hard. I did mine with a razor blade and it took a couple of minutes. The trick is to cut slowly and without too much pressure. Start with the corners, then go down the sides.

however, before you do that, are you sure both cores were being equally idle or stressed? You need to run two stress test programs to make both cores run at load. if you install prime95 in two directories and run both, does the difference in core temps remain?

yes

sablabra · 15 Aug 2006 at 18:39

Hesky82 said:
each core has its own independent temp sensor??? and everst can monitor each. I did not know that!!!!

Me didn't know either, until now

Anyway, here are two more shots to show the difference:

And as I said, core 1 ALWAYS fail a lot before core 2, which is really annoying...

sablabra · 15 Aug 2006 at 18:51

Creidiki said:
Dont believe coretemp.

According to it one of my cores idles at temperatures below ambient (not to mention below coolant)

I know that, especially since I have a dfi expert

sablabra · 15 Aug 2006 at 19:13

matt100 said:
its highly unlikely you have a core idling at 22 degrees.. coretemp tells me much the same about my 3800, with a hastily fitted freezer64 pro it says one of them is idling now at 29 degrees which given the fact my house feels about level with my body temp right now is almost impossible.

I know it's wrong, but it's still a big difference, right?

sablabra · 15 Aug 2006 at 19:29

Hesky82 said:
Hum, according to everst, I'm getting a difference of 10C between cores. core one being the hotter running core and core one is always the 1st to fail for me to.
Funny thing is I can leave core 2 burning away and it still runs cooler than core 1
like Creidiki says, I'm not sure how accurate coretemp is.

either way I should have my ihs removed already lol

Hmm, could be some kind of bug then? :confused:

What mobo you got?

Actually I had added 10c to the core temp. I turned that off, so now I'm about 5c under ambient

sablabra · 15 Aug 2006 at 19:30

Cyber-Mav said:
remove the heatspreadder and then run the test. im sure the cores will be both level in temps and clock a silly amount higher.

Yeah, I want to do that, but I'm affraid I will destroy the cpu. Either by chopping off resistors, or cracking the core

sablabra · 15 Aug 2006 at 19:38

snowdog said:
RMA it, say its unstable and show them pictures of it (core 1) overheating (if needed)
And hope they will send u a correctly working one...

If impossible to RMA or you dont want to, then its either remove the ihs or live with one overheating core...

Well, it's not exactly unstable, it's 100% prime95 stable up to 2.5ghz at stock volts... So I don't think I will have a good chance in RMAing it

sablabra · 15 Aug 2006 at 21:27

I seriously can not believe the difference in temperature here! :eek:

In this pic I'm only stressing core 2(core 1 is called core 0 in prime95, core 2 is called core 1

), and still core 1 is 24c warmer!! Also, when the cpu is nice and heated like now, the second I start stress testing core 1, the computer hangs... The second I start it! So this means I can might gain up to 200mhz by removing the IHS?

I'm running 2.8ghz 24/7, although core 1 isn't prime95 stable... Core 2 seems to be very stable at 2.9ghz, haven't tested further(will certainly do now!)

sablabra · 15 Aug 2006 at 21:42

Creidiki said:
Or most probably youll find that core sensors are miles off the mark and you'll get 1, maybe 2c @ load from removing the IHS and gain basically nothing.

Keeping in mind your lower core idle is actually physically impossible...

I know the sensor there is wrong, as I idle below ambient. However, the fact that it shows 24c difference, and freezes my computer in a second(in contrast to core 2, which keeps on going) convinces me that the IHS is making bad contact. I can't see any other logical thing(?) :confused:

sablabra · 15 Aug 2006 at 21:51

Creidiki said:
Easy, that core cant cope with those speeds.

When a processor fails to run at a certain speed, it doesent all fail at once, its often just one bad gate.

In my case, my 170 will do 2.8GHz @ 1.375v. @ 1.35v, core 0 (the one with reported sub-ambient idles) fails, every single time.

Yeah, you could be right, I'm not really sure :confused:

sablabra · 15 Aug 2006 at 22:04

Ok, now I'm REALLY confused :confused:

I just tried stress testing core 1 alone, and not core 2 at the same time. And I was shocked! The temperature of core 1 is a lot lower when I'm only stress testing core 1, than when core 1 was idle and core 2 stressed... How can that be? core 1 is actually stable now, until I start core 2 at the same time...

sablabra · 15 Aug 2006 at 22:46

Creidiki said:
You dont get it do you?

YOU CANNOT TRUST THOSE "INTERNAL SENSOR" READINGS AT ALL.

Thats all there is to say about the matter.

Relax, I get it, but it still doesn't make sense though

sablabra · 27 Nov 2006 at 12:12

Bringing this thread back to live

. I just popped off the IHS, and guess what, max 3-4c difference between each core! :eek:

So it obviously was bad contact between the core and IHS. Unfortunately, it doesn't look as if I can overclock any further at all

. Well well, it was fun though

sablabra · 27 Nov 2006 at 18:46

Jokester said:
Even with more volts?

Jokester

Yeah, it doesn't seem to want to go any further at all

. Can't even get it 100% stable at 2.8ghz now...

What's a little odd, is that during idle there's still a big difference in temps between the cores, but as soon as I fire up some load it's more or less equal. However, I do have another x2 3800+ that might be worth removing the IHS from, but I have to use one of them at stock cooler all the time... So if I'm going to remove IHS from my second cpu, I have to reassemble the IHS on this one again...