• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

x264 stress test

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
post 171, looks like here I stated my temps, and posted after pic, not before....must have been ocn where I posted before/after....back in july though, if I find that post, will link it.

And my delid pic is post 144 and 146 And post 108 has pic 87C temps with similar settings with prime 27.9, which is 10c cooler than prime 28.5, and post 121. I know I posted my 98C temps somewhere, cause someone chimed in about the temps.

If you do delid, just use vice, much safer than razor.
 
post 171, looks like here I stated my temps, and posted after pic, not before....must have been ocn where I posted before/after....back in july though, if I find that post, will link it.

And my delid pic is post 144 and 146 And post 108 has pic 87C temps with similar settings with prime 27.9, which is 10c cooler than prime 28.5, and post 121. I know I posted my 98C temps somewhere, cause someone chimed in about the temps.

If you do delid, just use vice, much safer than razor.

Did you use any chemical agents to get the old glue and TIM off the lid and the CPU/PCB?
 
took me about 15-20 minutes to gently scratch off the glue with a credit card...credit card works great for avoiding any damage to pcb. Then just wipe tim off ihs/and die, and I used isopropyl alcohol to clean both die/ihs. I used nail polish to cover the vrms beside the die, most use either nail polish or liquid electrical tape since using liquid metal as tim.
 
Okay, I delidded the thing and didn't bork it. Used Liquid Ultra for the TIM. Now, keep in mind that I did not clean off the black RTV material from the PCB. I decided to leave it on for the time being as it serves as a centering guide for the lid. I didn't want the lid sliding around as I lowered and seated the socket latch. Leaving that bead of RTV may prevent the lid from making optimum contact with the die. I can always remove it later if I decide to. Having said all that, some preliminary testing seems to have resulted in temps that are around 10c lower in the IntelBurnTest. Variance between one core and another is also somewhat improved, being about only about 7c between the hottest and the coolest instead of 9c. Late. Bedtime.
 
It's pretty normal that 4770K/4790K hit 90*C under load in new Prime95 when you raise voltage ( even after delidding ). For some reason I have lower temps at higher voltage on 4930K or 5820K than I had on 4770K/4790K.

No real refresh on the "Refresh" then, huh?

Differences are mainly in internal vrm ( at least in theory as we can't check it :) ), durability and general OC potential. Early Haswells were sometimes dying at 1.7V+ while new don't have this issue. Also new chips usually need lower voltage and are scalling better with voltage.

Delidding is usually lowering temps by about 10*C but it depends how good IHS was touching CPU core. If there is like 12-15*C difference between cores while CPU is under full load then you can count on even 20*C improvements but sometimes difference between cores is ~2-4*C and then delidding won't help much but still can notice ~5-10*C improvements ( depends from used TIM ).
 
Okay, I delidded the thing and didn't bork it. Used Liquid Ultra for the TIM. Now, keep in mind that I did not clean off the black RTV material from the PCB. I decided to leave it on for the time being as it serves as a centering guide for the lid. I didn't want the lid sliding around as I lowered and seated the socket latch. Leaving that bead of RTV may prevent the lid from making optimum contact with the die. I can always remove it later if I decide to. Having said all that, some preliminary testing seems to have resulted in temps that are around 10c lower in the IntelBurnTest. Variance between one core and another is also somewhat improved, being about only about 7c between the hottest and the coolest instead of 9c. Late. Bedtime.

You need to clean the black adhesive off, that is the primary factor causing the high temps, as it leaves too large a gap, meaning tim1 will be too thick of a bondline. And the drop in temps will be directly proportional to wattage tested, since gradient through cpu depends on wattage, and ~1/3 gradient is through tim in paste tim cpus.

At idle C states on, 1-2W, you likely wont see any temp difference or 1C max, since gradient is only 4-5C through the chip, and ~2C through tim. I tested at 185W cpu dissipated, 4.7ghz 1.3V, prime 28.5 small ffts. At 130W dissipated my temp drop was about 10-12C. Intel burn test is about 30W lower than prime small ffts, because it has old linpack libraries that dont use fma3. So if you tested with less than 1.3w and/or less stressful test, like IBT, the wattage will be less, the gradient less, and the decrease in temps less.

And like Woomack said, there is variability chip to chip. But still largest variability in temp drop is what vcore and loading program was used to test. You can usually guess the wattage tested by looking at decrease in temps. But you need to clean the adhesive off to make sure you reduced the gap.
 
Yeah, little or no difference at idle as you say, rge, but significant difference under load. Right now I'm stressing with Prime95. Last night I did a quick and dirty stress with IBT. It's one of the tools I used to stress the chip before delidding so it was a quick point of reference.

It occurs to me that the black adhesive may or may not be much of an issue. It looked to me that the adhesive was gripping mostly on the inside vertical surface of the lid walls and that the sharp edges of the walls were pressed down trough the adhesive. Perhaps there was little or no adhesive actually between the lid wall edges and the PCB.

Besides, I'm out of Liquid Ultra right now as the little remnant of it I used last night is now gone. I'll order some more just in case I decide to redo the thing or in case I decide to use it on the top side. I've used it on the top side of the lid on another build with a 2600k and got significantly lower temps than with other types of TIM.

Okay, my Prime95 v.27.9 temps seem to be about 17c lower than before delidding. This is at 4.6 ghz/1.255 vcore.
 
Last edited by a moderator:
As long as the ihs rests on the die and not elevated off the die by sitting on the adhesive, that is what matters.
 
Well, I cleaned the black adhesive off and didn't see any real change in temps. I was a little short on Liquid Ultra so will try that again when more arrives from my Amazon order. In the meantime, I cleaned the LU off and put MX2 on just for comparison but that too is in waiting as I had to RMA that wonky MSI board again. I also took a mini hot glue gun and put a thin bead around the perimeter of the lid to keep it from sliding around on the PCB. The hot glue probably won't hold very long with heating and cooling cycles but it's easy to peel off if it doesn't hold up.
 
Back