Two of same CPU: one runs a lot hotter?

[NAC]

I have a dual CPU machine with Xeon e2670 CPUs. One is cooled by a Corsair Hydro Series H80i V2. The other one by Enermax Liqmax II 240. I often run Sheepit in the background with a few cores on each CPU, and I noticed something strange. For example: each CPU can render with 12 threads. According to HNWMonitor - the CPU cooled by the Enermax will run hotter - just above 70C. The Corsair cooled CPU will typically be below 60C. But what is strange: the air out of the Enermax cooler is consistently hotter. I would think that either both streams of air would be about the same, or if anything - the air out of the Enermax would be cooler since it is removing less heat from the CPU.

They aren't overclocked, the voltage is the same, and I see this result day after day - so it isn't as if a particular render was more taxing than another.

So when doing the same work one CPU can consistently produces a lot more heat than an identical CPU?

 

[rickxross]

your setup is like this?
cpu1 @ 60C -> corsair rad hot air
cpu2 @ 70C -> enermax rad hotter air

let's assume a steady state where each cpu is producing a certain amount of heat, say X watts for cpu1 and Y watts for cpu2. I'm assuming each cpu's voltage/power supply is equal. With silicon variations I could see X and Y not being equal. Let's just assume that X=Y.

Each cooler can dissipate a certain number of watts, but I think that number is dependent on the cooler design and the delta between rad/coolant temperature and ambient. If you place a heat load on the cooler, the rad/coolant temperature increases until the amount of heat produced by the load is equal to the amount dissipated. Higher coolant temperature results in a higher cpu temperature. If X==Y, maybe the enermax needs a higher temperature delta to dissipate the same amount of heat compared to the corsair? are the ambient intake temperature on both coolers the same?

another thing to consider is cpu->waterblock interface. If we assume a constant heat production from the cpu, won't it take a higher temperature delta across the interface to get the same amount of heat transfer to the water loop if the interface is poor? a poor interface would manifest in an increased cpu temperature, but the radiator would be outputting the same amount of heat regardless of interface quality (assuming the cpu is not forced to throttle), so not an increased radiator temperature. so I don't think this matters.

If you want to check X==Y? you could use a power meter while loading up one cpu then the other and compare values.

 

[NAC]

your setup is like this?
cpu1 @ 60C -> corsair rad hot air
cpu2 @ 70C -> enermax rad hotter air

Yes!

let's assume a steady state where each cpu is producing a certain amount of heat, say X watts for cpu1 and Y watts for cpu2. I'm assuming each cpu's voltage/power supply is equal. With silicon variations I could see X and Y not being equal. Let's just assume that X=Y.

Each cooler can dissipate a certain number of watts, but I think that number is dependent on the cooler design and the delta between rad/coolant temperature and ambient. If you place a heat load on the cooler, the rad/coolant temperature increases until the amount of heat produced by the load is equal to the amount dissipated. Higher coolant temperature results in a higher cpu temperature. If X==Y, maybe the enermax needs a higher temperature delta to dissipate the same amount of heat compared to the corsair? are the ambient intake temperature on both coolers the same?

Basically, yes. They are next to each other on the top and back of the case.

another thing to consider is cpu->waterblock interface. If we assume a constant heat production from the cpu, won't it take a higher temperature delta across the interface to get the same amount of heat transfer to the water loop if the interface is poor? a poor interface would manifest in an increased cpu temperature, but the radiator would be outputting the same amount of heat regardless of interface quality (assuming the cpu is not forced to throttle), so not an increased radiator temperature. so I don't think this matters.

Agreed - I had thought about this and thought maybe I should re-attach the radiator with new thermal grease. But that wouldn't explain the hotter air.

If you want to check X==Y? you could use a power meter while loading up one cpu then the other and compare values.

According to HNW Monitor, the CPU is telling the motherboard that hotter CPU runs at 1.161v, and the cooler CPU runs at 1.151v. So there is a small difference, but only about 2%. I'm not sure how accurate that is. The motherboard CPU voltage shows as about 1.1 - it flutuates up and down a little. But only one value is given for both.

So to get back to this:

maybe the enermax needs a higher temperature delta to dissipate the same amount of heat compared to the corsair?

Indeed, it could be that one cooler performs that much better than the other. I think I will give it a good cleaning to ensure there is not too much dust on the Enermax, that has been in the case much longer.

 

[Markfw]

Yes!


Basically, yes. They are next to each other on the top and back of the case.


Agreed - I had thought about this and thought maybe I should re-attach the radiator with new thermal grease. But that wouldn't explain the hotter air.



According to HNW Monitor, the CPU is telling the motherboard that hotter CPU runs at 1.161v, and the cooler CPU runs at 1.151v. So there is a small difference, but only about 2%. I'm not sure how accurate that is. The motherboard CPU voltage shows as about 1.1 - it flutuates up and down a little. But only one value is given for both.

So to get back to this:

Indeed, it could be that one cooler performs that much better than the other. I think I will give it a good cleaning to ensure there is not too much dust on the Enermax, that has been in the case much longer.

That last point. Its possible the enermax has much slower fans. I know on my watercooling setup, I had one enermax and 2 nidac gentle typhoon's before the pump died. When I took it apart there was a lot more dust where the nidac fans were.

 

[Yeroon]

If the fan speeds between the two coolers is different, then slower air through the rad will come out hotter.

Edit: I did a quick google to see if i could find any info on fan differences - didnt find any yet - but it seems the liqmax has a fan switch somewhere.

 

[NAC]

Yes. The Enermax fans have a switch - 3 settings for high medium low I think. I believe mine are set at low. I also run all the fans through a fan controller, basically at at a setting that I just barely hear them. So the Corsair and Enermax fans are all throttled by the controller.

To my hand, the amount of airflow out of the single 120mm Corsair and the 240mm Enermax seem comparable. The Corsair flow is clearly faster, but the other one has twice the area through which air is moving. But it could be that the Corsair is actually moving more air, and my hand isn't making a good judgement.

If I turn up the Enermax fans, the CPU temperature will obviously go down. I'm not sure if I realized that the exhaust temperature will also go down, but it makes sense that it also would.

 

[aigomorla]

So when doing the same work one CPU can consistently produces a lot more heat than an identical CPU?

This is what we call silicon lottery.
It doesnt apply to just overclocking, it also applies to heat and power draw.

You can look at 2 identical cpu's however when cut from the waffer they are cut at different locations.
The odds of having 2 identical CPU's behaving like true twins, is astronomically small in regards that it had to be cut from the same iso wafter with mirror quality down to the litho.