RAM temperatures (with testing)

Soldato
Joined
6 May 2009
Posts
19,885
What are usual idle / load and maximum RAM temperatures for DDR4 RAM?

There is the following post on Reddit, which says upto 40c before stability issues occur and 50c before many errors / crash. I would rather have info from Ram Guru's. My personal testing below shows no stability issues when RAM is at 46c

https://www.reddit.com/r/overclocking/comments/9yx88h/ddr4_and_temperature_related_stability/
This is more of a PSA for all of the beginning ddr4 Ram overclockers. Ddr4 is rated for up to 1.5 volts 24/7 and temperatures of 81c before damage can occur. However temperatures above 40 degrees cause stability issues and random errors. And temps above 50 degrees Celsius can cause system crashes and huge errors on otherwise stable ram overclocks. Running your RAM at the lowest needed voltages to boot will often give better stability then just sitting 1.5 volts and trying to get it to stabilize.

This is a pretty well-known fact for ddr4 and can be reproduced yourself pretty easily. Doing some casual testing and observing I showed that random intermittent errors do occur much more frequently above 40° C. When pushed above 50° C Ram throws huge amounts of errors almost immediately once 50c is hit.

It probably depends on the specific RAM and heatspreader(s) used but have the following kit running at 1.5V. HWiNFO64 sensor status shows DIMM voltage at 1.528V (RAM running DDR3800, CAS14)

https://www.overclockers.co.uk/team...00c16-3600mhz-dual-channel-kit-my-002-8p.html

It does say "Tested Voltages: 1.20 - 1.40 V" So, here's me trying to give a bit back to the memory community since i've been helped a lot over the past two weeks.

As RAM doesnt (to my knowledge) have a temperature sensor I took a few of readings with a temperature laser, laser pointed just below the 'T' (on T-FORCE) on the heatspreader

Power on from cold - 20c (cold to touch)
10 minutes - 25c (warm / same as idle cpu temp)
20 minutes - 28c (warm / same as idle cpu temp)
30 minutes - 30c (warm / 1c higher than idle cpu temp)

TestMem5 v0.12 using [email protected]
1 minute - 32c
5 minutes - 38c
20 minutes - 41c / 43c
30 minutes - 42c / 42c
1 hour - 41c / 42c

Max temperature on side of RAM stick, 46c after 1 hour.

No errors throughout.
No cooling. No intake fans at all, only push pull on CPU and one exhaust fan.
Ambient room temperature 17c, no case side panel
 
Associate
Joined
10 Dec 2007
Posts
1,677
You know what would be interesting is to do this with 3 other setups.

1: XMP
2: 1.40v at semi optimised timings
3: 1.45v at semi optimised timings

You should hopefully be able to see a linear scaling in temperature.

What you might be able to do to get an even better reading is take the heatspreader off the top of the sticks which will get a reading much closer to the value of the chips themselves. Try and get the laser inbetween the heatspreader type of deal.
 
Associate
Joined
1 Apr 2019
Posts
1,523
What you might be able to do to get an even better reading is take the heatspreader off the top of the sticks which will get a reading much closer to the value of the chips themselves. Try and get the laser inbetween the heatspreader type of deal.

Contactless thermometers are nowhere near precise enough, the laser itself does not measure temperature, it is simply there so you know where you're pointing the sensor. The target area for the sensor is much larger then the point of the laser.

Only way to do it accurately would be with a thermocouple.
 
Soldato
OP
Joined
6 May 2009
Posts
19,885
I was thinking XMP / different timings tests and with / without heatspreader tests would be good but does take time.
It started out as me just worrying about running 1.5V and wondered if there was a max temp for DDR4 RAM in general.

I'll test XMP when I get some time to revert settings :)
 
Associate
Joined
1 Apr 2019
Posts
1,523
You're running it open/no side panel, you're never going to see silly temperatures unless you are also running silly voltages. This will only become a concern when pushing RAM voltages in a case with poor/limited airflow.
 
Soldato
OP
Joined
6 May 2009
Posts
19,885
You're running it open/no side panel, you're never going to see silly temperatures unless you are also running silly voltages. This will only become a concern when pushing RAM voltages in a case with poor/limited airflow.
Hence my question about maximum RAM temperatures for DDR4 RAM. Reddit quote says 40c before stability issues occur, I didnt have any at 46c. "silly temperatures and silly voltages" I don't know what is classed as being "silly" :D Some may say 1.5V is silly, others maybe more or less. Same with temperatures
 
Last edited:
Associate
Joined
1 Apr 2019
Posts
1,523
Reddit says a lot of things. Proof is in the pudding as they say - temps into the low / mid 40C range and you saw no errors at all.

The maximum temperatures for Samsung B-Die are here - https://www.samsung.com/semiconduct...11/8G_B_DDR4_Samsung_Spec_Rev2_1_Feb_17-0.pdf

Its important to note that these temperatures only apply when operating at the rated speeds, any manual OC or OC kit may not work at the higher end of those temperature limits.
 
Soldato
OP
Joined
6 May 2009
Posts
19,885
Reddit says a lot of things. Proof is in the pudding as they say - temps into the low / mid 40C range and you saw no errors at all.

The maximum temperatures for Samsung B-Die are here - https://www.samsung.com/semiconduct...11/8G_B_DDR4_Samsung_Spec_Rev2_1_Feb_17-0.pdf

Its important to note that these temperatures only apply when operating at the rated speeds, any manual OC or OC kit may not work at the higher end of those temperature limits.

Thanks, that's a lotof useful (yet complicated) info :)
"Temp. 0 ~ 85 °C" Certainly wouldnt want them running at 85c!
 
Associate
Joined
29 Jun 2009
Posts
549
Temperature at which degradation of timings and speed occurs changes between chips. I would define 48c+ as the zone at which you are risking instability. I would definitely not recommend running high voltage (1.45v+) in combination with those high temperatures as well. Personally I run up to 1.55v daily but my normal usage memory temps are mid 30s.
I have RipjawsV which have temperature sensors - in fact im pretty sure all G.Skill B-Die sticks do. Also Crucial Ballistix Max should have temp sensors as well. As for other manufacturers - Patriot Steel Vipers don't have temp sensors, but my old corsair 3333mhz 16gb b-die sticks do.
 
Associate
Joined
1 Apr 2019
Posts
1,523
Thanks, that's a lotof useful (yet complicated) info :)
"Temp. 0 ~ 85 °C" Certainly wouldnt want them running at 85c!

No neither would I, and those maximums go out the window when you're increasing voltage, running at higher than 2666MHz, and tighter timings. As with everything overclocking, there is a delicate balance to be had, and that balance is very system / specific silicon dependent.
 
Associate
Joined
4 Jun 2020
Posts
2,401
My Micron E die has been running 24/7 at 1.53v 4200 Mhz CL16 for months now without issue, and previously 1.48v 4000 CL18 on Z390.

I used a Black and Decker infra red thermometer thing to check it and gets up to around 55c after hours of use. Maybe the chips are hotter than that but as mentioned above Micron E die seems to not care about temps.

Unfortunately several people have now found that the new 3600 CL16 kits aren't anything close to these old 3200 LT Sport kits which is a shame, and stuff thats rated for better costs more than twice as much and isn't worth it as you only get like a 2% performance boost max.

And actually anything that can manage 3600 CL14 is more or less the same latency. It takes +300 Mhz per higher Cas step for any minor improvement, but my kit still needs 1.5v for 3600 CL14 and doesnt manage 4000 CL15.

Someone else also showed a kit of this stuff managing 4400 CL16 on a 2x16 Gb kit as well which is astonishingly rare and beyond golden for 16 Gb sticks, but the new kits are rarely even pulling off 4000 Mhz anymore, which I suspect is because they started binning them for Ballistix Max (£450+ for 2x16 4400 CL19 yea no thanks, and theres never any stock).
 
Last edited:
Soldato
Joined
29 May 2005
Posts
4,896
Correct. My recent 3600 E die just wont clock much at all and hates voltages. If I yank it higher than 1.4V with tested and stable clocks, it won’t post.
 
Associate
Joined
4 Jun 2020
Posts
2,401
I was thinking of previously trying to get a 3600 CL16 kit based off how much gain I got from from going from my previous 2666 to 3200 LT sport kits, but they're worse.

Back when 3200 kits were the highest, Crucial were doing zero binning and didn't have any clue themselves about what the E die chips could do. Now that they do they are binning them for Ballistix Max so hardly any chance of getting golden chips on cheap ram now.
 
Back
Top Bottom