Is there an appreciable effect on CPU temps when running 4 vs 2 DIMMs?

[2is]

Per the subject line, I'm wondering with modern CPUs that have an IMC, if there is an appreciable temp difference between the two configurations?

I know on my LGA 775 systems, the Northbridge ran significantly warmer when I had all 4 DIMM slots populated. Also along the same subject, how about voltage? I know low voltage DIMMs themselves would run a bit cooler, but would it have any effect on the CPU?

I'll be able to test this out for myself later this week when I help a friend build his setup. I have 4x4GB 1.5v DIMMs and he will be running 2x8GB 1.35v DIMMs. (both 1600mhz) Going to run 10 passes of IBT and log temps, then swap his sticks in my machine to do another 10 passes and compare.

Anyone done a similar comparison already?

 

[Idontcare]

I am sure it does make a difference but the IMC is not in the CPU core and is nowhere close to the DTS (digital temperature sensor). So you probably have no way of measuring the difference caused by a more active IMC versus a less active IMC.

Now it might cause a knock-on effect in temperature within the core if the lowered memory bandwidth results in a lowered performance (lower effective IPC because of pipeline stalls from cache misses).

So you might be able to observe a lower operating temperature but it wouldn't be for the reasons you were thinking.

 

[sefsefsefsef]

Here are the things that I can think of that could/couldn't affect CPU temperature in this scenario:

1. Each DIMM is good for ~5W of power consumption, so that's an extra 5W per DIMM of heating inside your case, slightly raising the temperature of the air around the CPU.
2. The number of RAS/CAS/PRE commands being sent from the IMC to the DRAM would not change, so the IMC isn't going to be "more active" with 4 DIMMs populated instead of 2. In other words, the bandwidth requirement (as far as the IMC is concerned) would not change at all.
3. However, the capacitive load on each memory channel does increase when you add in the extra DIMMs, so each signal sent from the IMC to the DIMMs is going to use a little bit more energy, and this energy is consumed in the IMC's external drivers, so that *could* increase your CPU temperature, but I kinda doubt it makes a measurable difference.

IMO, the largest effect would be the increase in ambient case temperature from the extra ~10W of DIMMs, but I'd be surprised if you could measure this difference.

 

[Liquid_Static]

Just tested it and noticed a 1 degree increase going from 2-4 DIMMS.

 

[2is]

Load or idle?

 

[Diogynes]

Just tested it and noticed a 1 degree increase going from 2-4 DIMMS.

That may just be within the margin of error. I think sefsefsefsef gave a pretty good assessment.

 

[fixbsod]

While difft I noticed lower load cpu temps (3 deg c?) using 1600 mhz vs 1866 ram settings w/sandy and its imc.

 

[2is]

While difft I noticed lower load cpu temps (3 deg c?) using 1600 mhz vs 1866 ram settings w/sandy and its imc.

that's probably because the core themselves were doing less work (lower memory bandwidth) I'm not going to be changing speeds or timings, Just 3 of sticks and voltage.