6700K fails Prime 95 blend at stock speed (Update: Resolved. Now limited by temps.)

[dank69]

Chip manufacturers think that the safety of the chip is worth a couple degrees, and I partially agree, but they could offer a desktop model without the lid. They could even charge a premium for it.

I thought you get the 12-13C even after putting the IHS back on? I think they just do a lousy job of putting paste in there.

 

[Deders]

I thought you get the 12-13C even after putting the IHS back on? I think they just do a lousy job of putting paste in there.

It's not so much the paste as it is the gap the paste has to fill. You have to remove all the adhesive under the edges of the IHS to lessen the gap. I still think they could do a better job.

 

[dank69]

It's not so much the paste as it is the gap the paste has to fill. You have to remove all the adhesive under the edges of the IHS to lessen the gap. I still think they could do a better job.

Yeah, it's pretty crazy. Engineer one of the most technical gadgets on the planet and then cheap out on the assembly.

 

[dank69]

I had some more time to play around last night. I was able to run a few passes of Prime max heat at 4.6GHz vcore 1.3v and RAM at 2133. I decided to boost the RAM up to its 2800 rating (14CAS) but it failed Prime blend during the first pass. Dialed the system down to 4.4 @ 1.3v and ran Prime blend all night with 2800-14 so these settings are quite stable. Let blend continue running and will check it after a full 24 hours. Temps kissed 90C in one core, 88-89 in the other 3. Sucks that I am limited by temps. Will probably need to delid if I want a better overclock. If I increase even to 1.35, actual voltage in CPU-Z increases to 1.41 during stress and I start hitting TJmax. Not happy about this. I'll obviously need to double check my own paste application before I go the delidding route.

Is it normal that setting the RAM to 2800 would destabilize a 4.6 OC? Anyone know the technical details of why this happens? I would have thought that they are somewhat independent but that clearly isn't the case.

 

[Deders]

You might need to raise VCCIO voltage for better stability at higher clockspeeds with fast ram.

 

[dank69]

You might need to raise VCCIO voltage for better stability at higher clockspeeds with fast ram.

Great, thanks for the tip. I'll give that a whirl next.

 

[Geforce man]

I had some more time to play around last night. I was able to run a few passes of Prime max heat at 4.6GHz vcore 1.3v and RAM at 2133. I decided to boost the RAM up to its 2800 rating (14CAS) but it failed Prime blend during the first pass. Dialed the system down to 4.4 @ 1.3v and ran Prime blend all night with 2800-14 so these settings are quite stable. Let blend continue running and will check it after a full 24 hours. Temps kissed 90C in one core, 88-89 in the other 3. Sucks that I am limited by temps. Will probably need to delid if I want a better overclock. If I increase even to 1.35, actual voltage in CPU-Z increases to 1.41 during stress and I start hitting TJmax. Not happy about this. I'll obviously need to double check my own paste application before I go the delidding route.

Is it normal that setting the RAM to 2800 would destabilize a 4.6 OC? Anyone know the technical details of why this happens? I would have thought that they are somewhat independent but that clearly isn't the case.

Is it normal? Yep. For example, my 2700k in my p67 board has no problem running the ram @ 1866 c9. If I try to up it to 2133, which the ram can easily do, I lose stability, at any Mhz. That would be down to the IMC of the chip. Last I checked, Intel's IMC isn't rated for 2800Mhz officially (on the 6700k), so its possible that yours doesn't like that speed. It could also need a bump to one of the other voltages. It could also be your motherboard! There are a ton of variables.

 

[dank69]

Is it normal? Yep. For example, my 2700k in my p67 board has no problem running the ram @ 1866 c9. If I try to up it to 2133, which the ram can easily do, I lose stability, at any Mhz. That would be down to the IMC of the chip. Last I checked, Intel's IMC isn't rated for 2800Mhz officially (on the 6700k), so its possible that yours doesn't like that speed. It could also need a bump to one of the other voltages. It could also be your motherboard! There are a ton of variables.

I was just surprised that this RAM runs 2800-14 fine when the CPU is at 4.4 but then the RAM fails at 2800-14 when the CPU is at 4.6. I didn't change any voltages. I am curious how all this works and why bumping up the CPU speed would make the RAM fail.

 

[dank69]

You might need to raise VCCIO voltage for better stability at higher clockspeeds with fast ram.

Great, thanks for the tip. I'll give that a whirl next.

Bumping the VCCIO to 1.2v didn't help.

I need to figure out which setting on the Asrock is system agent. I think I know but don't want to guess, haha.

 

[Deders]

bumping the vccio to 1.2v didn't help.

i need to figure out which setting on the asrock is system agent. I think i know but don't want to guess, haha.

vccsa?

 

[wanderica]

Did you delid yourself? It kind of pisses me off that delidding can drop temps 12-13C. Consumers shouldn't need to assemble the chips properly.

I did delid myself, and I was super nervous about it. It turns out it wasn't all that difficult as long as you take your time. I didn't get into lapping the IHS or anything too advanced. I simply wanted to get rid of the glue (silicon sealer?) and replace Intel's crappy thermal paste. I haven't hit thermal limits yet, but I'll admit, I have no intention of pushing the VCore any further, so there wasn't really a point to trying to hit them. I'm satisfied with my 4.7 at ~70 C.

I agree that seems like a QA issue. I don't pretend to know what goes into developing these things, but I can't help but think that Intel could come up with a better solution than letting this fat guy with a razor blade fix their oversight in development.

 

[ShintaiDK]

I think the current IHS setup is fine. People have to remember all the situations it needs to handle. And the change at stock settings between that and delid is irrelevant.

However, I would like to see Intel shipping special SKUs of the K/X chips without the IHS. These would have to have 117 warnings on the box with neon lights on saying its only for the most extreme. And any chipping/crack of the die results in no warranty covering. Specially the HEDT chips is a nightmare to delid. Tho they use solder.