Samsung B-die - how do you keep yours cool?

Sumanji · 22 Feb 2021 at 05:32

Hey,

I have 2 x 16GB of G.Skill Trident Z Neo PC3600, and before this had 2 x 8GB of Trident Z PC3200 - both are Sammy B-die. Both sets of RAM are totally stable under Memtest86+ and Prime95 with 1.35V, using 1usmus' Safe settings for the relevant speed. However, I get crashes / hard locks when gaming. Long story short I've narrowed this down to the RAM being at fault, and specifically think it's RAM over-heating.

I've read B-die likes being 20-40C to keep stable. Mine is mid-30s while idle, and goes to mid/high-50s when gaming. There are three 120mm intakes blowing into the case, which ramp up when the CPU temp goes up, so the RAM is getting pretty decent airflow, and I am not sure how else I can optimize this.

I've ordered some Crucial Ballistix to replace it, as it seems Micron E-Die is a lot less picky about temperature, but wondering if there is anything else I can do to optimize the Samsung sticks before I give up on them...

Cheers,

Su

clone · 22 Feb 2021 at 07:51

Got 16gb 3600 8pack ripped runs fine with no extra cooling.
You could buy a ram cooler but would need to make sure it will clear the ram.

Sumanji · 22 Feb 2021 at 13:41

What RAM temps do you get on load while gaming?

Su

OC2000 · 22 Feb 2021 at 17:56

What GPU are you using? I switched to water cooling the Ram after experiencing high Ram temps with a 3090 and have now dropped from 48C to 35C while gaming with tighter timings too.

Sumanji · 22 Feb 2021 at 23:27

Vega 64... so it definitely vomits a lot of heat into the case!!

Hotwired · 23 Feb 2021 at 00:18

Sumanji said:
Hey,

I have 2 x 16GB of G.Skill Trident Z Neo PC3600, and before this had 2 x 8GB of Trident Z PC3200 - both are Sammy B-die. Both sets of RAM are totally stable under Memtest86+ and Prime95 with 1.35V, using 1usmus' Safe settings for the relevant speed. However, I get crashes / hard locks when gaming. Long story short I've narrowed this down to the RAM being at fault, and specifically think it's RAM over-heating.

I've read B-die likes being 20-40C to keep stable. Mine is mid-30s while idle, and goes to mid/high-50s when gaming. There are three 120mm intakes blowing into the case, which ramp up when the CPU temp goes up, so the RAM is getting pretty decent airflow, and I am not sure how else I can optimize this.

I've ordered some Crucial Ballistix to replace it, as it seems Micron E-Die is a lot less picky about temperature, but wondering if there is anything else I can do to optimize the Samsung sticks before I give up on them...

Cheers,

Su

G.Skill does not support this idea of their memory overheating at such low temperatures.

https://www.gskill.us/forum/forum/product-discussion/ddr4/163842-what-is-normal-memory-temperature

https://www.gskill.us/forum/forum/product-discussion/ddr4/166098-temperatures-of-tridentz

But G.Skill will tell you on all their product pages that the overclock/XMP is not guaranteed:

Rated XMP frequency & stability depends on MB & CPU capability.

Both Samsung and Micron rate their DDR4 chips for at least 95 degrees.

Samsung: https://www.samsung.com/semiconductor/dram/ddr4/
Micron: https://www.micron.com/products/dram/ddr4-sdram

Also rated for 95 degrees is the Micron GDDR6X memory on a 3090 which is frequently seen toasting its nuts beyond that temperature: https://www.micron.com/products/ultra-bandwidth-solutions/gddr6x

On the other hand your memory controller can fail to support that RAM with that overclock/XMP at any temperature. Which is what the G.Skill warning is about.

Grim5 · 23 Feb 2021 at 02:40

I've got my B die at 3600 14

gets up to 55c while I'm gaming

takesnit like a champ no issues

Sumanji · 23 Feb 2021 at 20:18

Hotwired said:
G.Skill does not support this idea of their memory overheating at such low temperatures...

But G.Skill will tell you on all their product pages that the overclock/XMP is not guaranteed...

Hmm, ok this makes sense... seems unlikely I got two "bad sets" of memory.

Now I am double guessing my assertion it's the RAM. I'm fairly sure it's not the mobo or the PSU, since I don't have any issues loading the system with something like Prime 95. I initially thought the issue was a weak IMC on my old 2700X, but I've still had a couple of crashes on my new 5900X... which leads me back to the RAM.

When I downclocked the previous 3200MHz RAM to 2933MHz and played games, the crashed no longer occurred, which is why I was convinced it an issue with the RAM during gaming only (i.e. GPU heat pushing it over the edge of stability).

Problems like this are so annoying to pinpoint! Do you have any advice on what else I can do to test?

Grim5 said:
takes it like a champ no issues

Hmm, so like pretty similar to mine... what voltage are you using?

VaderDSL · 23 Feb 2021 at 20:24

Now adays with the memory controller on the CPU running XMP or overclock can be more fiddly to pin down. It could be you happen to get a CPU that just doesn't like high RAM frequencies/low latencies at stock volts, then you might have to go down the route of looking into the volts on the CPU that control/impact the RAM. I think SOC, VDDP, VDDG voltages are three key ones to look at.

postmanfw · 23 Feb 2021 at 20:49

Get a small fan blowing directly over them....knocked nearly 10c off mine.

Sumanji · 23 Feb 2021 at 23:03

VaderDSL said:
I think SOC, VDDP, VDDG voltages are three key ones to look at.

Cool thanks man - one thing I noticed was that my VDDG CCD was a little lower than what 1usmus recommends (0.89V vs. 1.00V), so I just bumped that up. Also noticed a little droop on my VDIMM (set to 1.350V but HWMonitor reads 1.341V) - will try and bump that if the crashes continue.

Still not convinced that either of these are to blame though, as it's solid during Memtest86+ and Prime95, but willing to give anything a shot to sort this out

postmanfw said:
Get a small fan blowing directly over them....knocked nearly 10c off mine.

Any reccs mate? Can you post a pic of yours please? Noise is a concern, so if it's a 60mm or something it would need to spin pretty slowly. The rest of my case fans are at 600rpm right now, haha!

[Edit] Another question please - any good reccos for a 3D stress test I can run on a loop for hours that doesn't have any downtime please? I'd like to stress the CPU, RAM and GPU at the same time, to mimic what gaming would do. I've tried 3DMark on a loop, but the system load goes up and down whenever it reloads the next scene which isn't great to stress the system continuously.

muziqaz · 26 Feb 2021 at 14:48

Any moving air directly on top of the RAM sticks will help. However you cannot escape the GPU heat whatever you do, unless you move the GPU away from the RAM

Cyber-Mav · 1 Mar 2021 at 21:42

My ram is around 60c and no crashes. You sure its its actually stable or the imc?
Best to run proper testing on it or run it stock

Sumanji · 4 Mar 2021 at 20:34

Hmm, ok, just had a hard lock yesterday while playing Destiny 2 - was pretty damn cold, so convinced it's not the RAM at fault anymore!!

Any ideas where else I should hunt? Motherboard and PSU I haven't investigated yet, and they are the hardest to diagnose...

Could also try a full format of Windows I guess (still rocking the original install from summer 2018).

Cheers,

Su

Mr crowley · 7 Mar 2021 at 20:54

Hard luck, i'm going through the same. Make sure SOC is disabled in bios, that's the latest advice i've followed, hopefully it does the trick. my corsair vengeance 3600 has been running flawlessly at 3400mhz for a week, today after the pc has been idling for ages I try to launch forza horizon 4, it crashes literally on the intro video and event ID 1000 indicating unstable ram. Look at system file checker, it's something you can run through command prompt and rule out that it's a windows problem.

Stu · 10 Mar 2021 at 17:38

Isn't the first step to increase the VDIMM above 1.35V ? Don't most people run B-Die at 1.45V (or a bit higher)?

Cyber-Mav · 10 Mar 2021 at 17:41

Sumanji said:
Hmm, ok, just had a hard lock yesterday while playing Destiny 2 - was pretty damn cold, so convinced it's not the RAM at fault anymore!!

Any ideas where else I should hunt? Motherboard and PSU I haven't investigated yet, and they are the hardest to diagnose...

Could also try a full format of Windows I guess (still rocking the original install from summer 2018).

Cheers,

Su

Cpu imc?

TT158 · 10 Mar 2021 at 21:20

Cyber-Mav said:
My ram is around 60c and no crashes. You sure its its actually stable or the imc?
Best to run proper testing on it or run it stock

Is that b-die? And how do you measure it? Built in sensor or temp probe? 60c is very hot! I find my b-die sticks get less stable in the high 40s, when clocked fairly hard (as hard as my meagre skills can get them, at least)

Cyber-Mav · 10 Mar 2021 at 21:56

I used ir gun. Not sure what die it is. Dont think kingston uses Samsung chips