DDR3 R-DIMM vs LR-DIMM temperatures?

[BLinux]

This is a HW question, but couldn't quite fit it into any of the HW sub-forums, so I'll ask here.

Do LR-DIMMs typically run cooler than R-DIMMs?

I have a server that runs 24/7 in my home with PC3-12800R RDIMMs, and the temps are around 38~48C and some times goes up to 55C if I'm putting a load on RAM.

I just got a set of PC3L-12800L LR-DIMMs, and was testing them with memtest86+. After about 12 hours of continuous testing, the temp sensors only showed between 30-35C.

I know the 2 scenarios are different, but I was some how expecting to see the LR-DIMMs run hotter? So, question is, if/when I swap the R-DIMMs for the LR-DIMMs, what should I expect from the DIMM temp sensors? same? lower? higher?

 

[nthu9280]

I don't use LRDIMMs so take my comments with a grain of salt. I think following two factors are in play.

- You are comparing PC3 (1.5v) vs PC3L(1.35v).
- The LRDIMMs I've seen so far have heatspreaders.

 

[EffrafaxOfWug]

I dare say there's a fair chance of comparing apples to Sunny Delight here, I imagine as have I you've read that LRDIMMs are meant to consume more power than RDIMMs overall - whether this translates to higher temperatures depends on a lot of factors I think. Aside from what's already been said about the voltage and heatspreaders;

> Was the test with the LRDIMMs in the same motherboard (and chassis)?
> Are they both from the same manufacturer? I suspect there'll be a marked difference in temps between samsung vs. micron vs. SK
> Are the temperature sensors in the same place on the DIMMs? IIRC they're normally integrated on the EEPROM IC - usually in the middle of the board, but it can make a huge difference to apparent temperature if they're located at the bottom vs. the top of the module

 

[BLinux]

I don't use LRDIMMs so take my comments with a grain of salt. I think following two factors are in play.

- You are comparing PC3 (1.5v) vs PC3L(1.35v).
- The LRDIMMs I've seen so far have heatspreaders.

Good point about the 1.5v vs 1.35v.

I thought LR-DIMMs had heat spreaders because they typically run hotter?

I dare say there's a fair chance of comparing apples to Sunny Delight here, I imagine as have I you've read that LRDIMMs are meant to consume more power than RDIMMs overall - whether this translates to higher temperatures depends on a lot of factors I think. Aside from what's already been said about the voltage and heatspreaders;

> Was the test with the LRDIMMs in the same motherboard (and chassis)?
> Are they both from the same manufacturer? I suspect there'll be a marked difference in temps between samsung vs. micron vs. SK
> Are the temperature sensors in the same place on the DIMMs? IIRC they're normally integrated on the EEPROM IC - usually in the middle of the board, but it can make a huge difference to apparent temperature if they're located at the bottom vs. the top of the module

1. yes, same motherboard and chassis. the system used to run the memtest86+ on the LR-DIMMs is an offline backup spare machine to the one running 24/7 with the R-DIMMs, they are identical except that the spare machine has slower CPUs and no HDDs.

2. well, not sure brand matters between R-DIMMs vs LR-DIMMs, but it just so happens both are from SK Hynix.

3. I don't know where the temp sensors are located.

 

[Evan]

Temp sensor location maybe as mentioned.

Other thing, what chips ? It could be as the LRDIMM’s are higher density they are using a smaller process node, eg 22nm instead of the RDIMM’s which could be say 32nm, you can usually find out based on the chip part number.

Do either have heat spreaders ?

I would not have expected any appreciable difference in total, LRDIMM has extra chips to act as a buffer to allow the memory controller on the CPU/main board to use less power to drive the memory but behind that the memory should be roughly same i though. (Not I am not a memory engineer so this is really all speculation)

 

[EffrafaxOfWug]

1. yes, same motherboard and chassis. the system used to run the memtest86+ on the LR-DIMMs is an offline backup spare machine to the one running 24/7 with the R-DIMMs, they are identical except that the spare machine has slower CPUs and no HDDs.

Ah OK so it's not the exact same server you're using, but rather mostly-identical kit in a cold-spare configuration? If so it's highly likely that the case ambient temperature is much lower which can easily account for 5-10°C of difference. What's the temperature delta like on the CPUs vs. the production system (assuming comparable heatsinks)?

2. well, not sure brand matters between R-DIMMs vs LR-DIMMs, but it just so happens both are from SK Hynix.

Just curious as there's likely differences in specs between the temps on the chips that go in one module as opposed to another, and any such differences would likely be accentuated further in the event of different manufacturers.

3. I don't know where the temp sensors are located.

It's almost always included in the EEPROM which is normally a little chip that usually sits in the middle of the DIMM (typically near the notch); from a butchers at some pics of SK RAM their DDR3 some are at the bottom and some are at the top, but if there are heat spreaders on your DDR3L it's likely not going to be easy to pinpoint.