Personal computing discussed
Moderators: renee, Starfalcon
Aranarth wrote:You want a large delta between coolant and heat source.
The cooler your coolant the more efficient it is at taking heat from the processor.
If you measure the coolant temp going into the processor and again coming out you should see the cooling performance of your block and radiator.
In creasing the speed of the coolant flow will improve the performance of both.
Sounds like things are working just perfect to me if you are really worried try reapplying your tim.
just brew it! wrote:Unfortunately, the most likely culprit is the interface between the die and the lid, which can only be fixed by de-lidding.
just brew it! wrote:Yes and no. While a high thermal gradient will result in higher rate of heat transfer, it is also a symptom of an awful lot of thermal resistance between the die and the waterblock. Unfortunately, the most likely culprit is the interface between the die and the lid, which can only be fixed by de-lidding.
Dposcorp wrote:just brew it! wrote:Unfortunately, the most likely culprit is the interface between the die and the lid, which can only be fixed by de-lidding.
100% true. I just returned a 6700K to Microcenter because after multiple heatsink testing it just ran WAY too hot for comfort even at stock, and that was in a open case.
Waco wrote:This. Your flow is too low if you can measure a serious delta anywhere in the loop.
I measured this back when I had:
A highly clocked Phenom II X4
Two 4870X2 with full cover blocks
That's nearly a kilowatt of heat output, and the highest delta I could measure with my pump on low was around 1 C between the exit of the radiators and the exit of all the heat (CPU -> 4870X2 -> 4870X2 -> temp sensor -> radiator -> temp sensor).
just brew it! wrote:Dposcorp wrote:just brew it! wrote:Unfortunately, the most likely culprit is the interface between the die and the lid, which can only be fixed by de-lidding.
100% true. I just returned a 6700K to Microcenter because after multiple heatsink testing it just ran WAY too hot for comfort even at stock, and that was in a open case.
This is definitely a hidden downside to Intel's process superiority. Their ability to shrink their CPUs using the latest process means the thermal interface between the die and lid becomes the weak link for cooling, since the surface area available for heat transfer is so small.Waco wrote:This. Your flow is too low if you can measure a serious delta anywhere in the loop.
I measured this back when I had:
A highly clocked Phenom II X4
Two 4870X2 with full cover blocks
That's nearly a kilowatt of heat output, and the highest delta I could measure with my pump on low was around 1 C between the exit of the radiators and the exit of all the heat (CPU -> 4870X2 -> 4870X2 -> temp sensor -> radiator -> temp sensor).
I don't think the OP said anything about deltas in the cooling loop. AFAIK the 50C gradient is between his die temp and the coolant temp (not sure where the coolant temp was measured, I was assuming at the outlet of the radiator).
Waco wrote:My point is that it shouldn't matter where you measure in a loop at all, as the in loop temperature should be consistent within a very small window. If there is an in loop delta, something is wrong.
arunphilip wrote:Waco wrote:My point is that it shouldn't matter where you measure in a loop at all, as the in loop temperature should be consistent within a very small window. If there is an in loop delta, something is wrong.
OT - I've never used an LC setup, so pardon the naïve question. If the difference in temperature is so low (e.g. entry and exit of radiator), then why do we have a radiator/fan setup, and why not just loop it around. If I had been asked to guess, I'd have just assumed (without any empirical evidence) that the radiator would have cooled the coolant by 10 degrees if not more. Maybe I'm just confusing the behaviour of automotive radiators/intercoolers with this.
Waco wrote:just brew it! wrote:I don't think the OP said anything about deltas in the cooling loop. AFAIK the 50C gradient is between his die temp and the coolant temp (not sure where the coolant temp was measured, I was assuming at the outlet of the radiator).
Sorry, I wasn't clear. I thought you were talking about an in-loop gradient. I should learn to read more carefully! I should have replied to Aranarth.
My point is that it shouldn't matter where you measure in a loop at all, as the in loop temperature should be consistent within a very small window. If there is an in loop delta, something is wrong.
Drewstre wrote:----------
So, what's the verdict? Bad lid?
Thanks again to all
Drewstre wrote:Aranarth- I get that cooler coolant is cool. But if the heat isn't getting transferred to the coolant, that ain't cool. I have no way of measuring the coolant on both sides of the block besides poking probes into the tubes. This is a closed loop, off-the-shelf cooler. It's like my Momma always said, "don't go poking probes into a sealed system". She was smart.
----------
So, what's the verdict? Bad lid?
Thanks again to all
synthtel2 wrote:1.35V is basically stock - are you sure about that?
just brew it! wrote:Are *you* sure about that? A lot of sources around the 'net seem to think stock for that chip is 1.2V.
Drewstre wrote:By the way, while looking for stress testers, I see some folks online saying that Linx (Linpack) is a better stressor for CPUs than Prime 95. Thoughts?
Drewstre wrote:Dposcorp- what temps were you getting that caused you to return your 6700K?