Samsung PM983a M.2 22110 SSD NVMe PCIe 3.0x4 1.88TB - open box - $125 OBO + free ship

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.

zac1

Well-Known Member
Oct 1, 2022
432
358
63
in the end i got 2 toshiba's and 2 samsungs.
I've done full disk write test on it, they appear to be new; it would take few months to kill them with writes to verify if they were ever used.
Additionally i've looked up similar nvme's on ebay and seem the prices are only bit higher - not in significant way.


The samsung gets so god damn hot - heatsink in picture struggles under load. (single 80mm noctua at 2000rpm blows on pcie slots).
View attachment 25711
A faster 4-12k fan should be used instead to keep it down with that heatsink for 24x7 load.

(green is samsung / cyan is toshiba)
View attachment 25712
I've been looking into heatsinks in anticipation of the new drives. Are those copper ones the most effective?

Probably not going to quad these up on a double-sided card like I planned initially... but it is a tight 1.5U chassis with 3x 13.5k rpm fans with one pointed right at the card...
 

CyklonDX

Well-Known Member
Nov 8, 2022
784
255
63
Are those copper ones the most effective?
Yes, and no.
The more area you have to transfer heat is going to be the most effective one. Those aren't that tall so the area for heat transfer is quite limited, and i think the idea was for fan to be over them or very near them.

There's very few heatsinks for that would fit on 22110's.

Potentially if you have space you should use copper heatsinks with bigger fin stacks. They will be able to cool it down much better, and faster.
(as long as size is right - and you have enough space, and mount them on every component on the nvme.)

1669174396788.png
1669174648805.png
1669174665409.png
 
  • Like
Reactions: zac1

tozmo

Active Member
Feb 1, 2017
142
102
43
74
The holy grail is an adapter that supports 22110 without bifurcation on the motherboard. So 22110 and plx switching on the card
 

TRACKER

Active Member
Jan 14, 2019
169
48
28
this is what i have in my system: click
it has massive heat sink and temps are (960 evo 1TB, 970 evo plus 1TB): 40-43°C idle, 48-49°C under load :)
 

zac1

Well-Known Member
Oct 1, 2022
432
358
63
Yes, and no.
The more area you have to transfer heat is going to be the most effective one. Those aren't that tall so the area for heat transfer is quite limited, and i think the idea was for fan to be over them or very near them.

There's very few heatsinks for that would fit on 22110's.

Potentially if you have space you should use copper heatsinks with bigger fin stacks. They will be able to cool it down much better, and faster.
(as long as size is right - and you have enough space, and mount them on every component on the nvme.)

View attachment 25715
View attachment 25716
View attachment 25717
Apologies for bumping old thread. Just wanted to follow up on the thermal convo and thank you for the pointers.

I ended up ordering some of these 15mm x 15mm heatsinks to slap on the controllers a la this post I found while researching. Results pending.
 
  • Like
Reactions: CyklonDX

zac1

Well-Known Member
Oct 1, 2022
432
358
63
for testing i've just ordered those 2 heatsinks (in near future - i will post results.)
View attachment 25905
I have to avoid the full-length heatsinks due to space constraints. From what I understand, the primary heat source is the controller, so I'm hoping the targeted heatsink + direct fans will perform well enough. The test results in the forum thread I linked previously suggest this will work swimmingly in a tight space — and cost-effective to boot.
 

zac1

Well-Known Member
Oct 1, 2022
432
358
63
Apologies for bumping old thread. Just wanted to follow up on the thermal convo and thank you for the pointers.

I ended up ordering some of these 15mm x 15mm heatsinks to slap on the controllers a la this post I found while researching. Results pending.
These worked swimmingly indeed. Tested with the PM983a 1.92TB on a quad M.2 NVMe card in a 721TQ chassis (5028D-TN4T) with the one stock fan. I expect the results will be even better in a tighter case with more airflow.

Without the heatsink, I can't finish the benchmark for fear of starting a fire. After applying the heatsink to the controllers, max temp reached is 71C. Ambient temp is about 19C. I would probably add a small fan in this chassis for longer-term deployments.

Edit: tested a different drive on the card, and it hit 78C. Ordered fans and will re-test with better airflow.

Edit2: there are three temp sensors. Not sure which is which. Need to figure that out.

Edit3: bumped case fan to max and re-ran benchmarks on each of the four drives. A little airflow goes a long way! These should work like a charm with a fan on them. Not bad for 50 cents each.
1670205698495.png
 
Last edited:

CyklonDX

Well-Known Member
Nov 8, 2022
784
255
63
After applying the heatsink to the controllers, max temp reached is 71C.
On my box i've noticed that after reaching 71'C the sensor actually stops working (the polling stops, and once below 70'C it comes back up - you can still write/read during that time)
 

zac1

Well-Known Member
Oct 1, 2022
432
358
63
On my box i've noticed that after reaching 71'C the sensor actually stops working (the polling stops, and once below 70'C it comes back up - you can still write/read during that time)
Interesting. Mine had no problem polling all the way up to 88C.
 

VMman

Active Member
Jun 26, 2013
125
45
28
These worked swimmingly indeed. Tested with the PM983a 1.92TB on a quad M.2 NVMe card in a 721TQ chassis (5028D-TN4T) with the one stock fan. I expect the results will be even better in a tighter case with more airflow.

Without the heatsink, I can't finish the benchmark for fear of starting a fire. After applying the heatsink to the controllers, max temp reached is 71C. Ambient temp is about 19C. I would probably add a small fan in this chassis for longer-term deployments.

Edit: tested a different drive on the card, and it hit 78C. Ordered fans and will re-test with better airflow.

Edit2: there are three temp sensors. Not sure which is which. Need to figure that out.

Edit3: bumped case fan to max and re-ran benchmarks on each of the four drives. A little airflow goes a long way! These should work like a charm with a fan on them. Not bad for 50 cents each.
View attachment 25955
thanks for the detailed feedback, I have the same card 4 x 4 using 16x PCIe.

Out of interest have you tried to run concurrent benchmarks on each drive, I found some odd results where the card drops some of the NVMe's during write loads on more than a single card?

My tests would only yield 1100MB/s (write) on 1 of the 4 cards with the others less than 200MB/s

I checked the slot was rated at 16x and since I was seeing 4 cards inside the OS the bifurcation was working.
 

zac1

Well-Known Member
Oct 1, 2022
432
358
63
thanks for the detailed feedback, I have the same card 4 x 4 using 16x PCIe.

Out of interest have you tried to run concurrent benchmarks on each drive, I found some odd results where the card drops some of the NVMe's during write loads on more than a single card?

My tests would only yield 1100MB/s (write) on 1 of the 4 cards with the others less than 200MB/s

I checked the slot was rated at 16x and since I was seeing 4 cards inside the OS the bifurcation was working.
You know, that crossed my mind... what do you use for concurrent benchmarks?
 

VMman

Active Member
Jun 26, 2013
125
45
28
no expert but this is my params

diskspd.exe -d600 -c32G -b4k -t8 -si -w100 c:\test1.txt d:\test1.txt e:\test1.txt f:\test1.txt
 
  • Like
Reactions: zac1

zac1

Well-Known Member
Oct 1, 2022
432
358
63
no expert but this is my params

diskspd.exe -d600 -c32G -b4k -t8 -si -w100 c:\test1.txt d:\test1.txt e:\test1.txt f:\test1.txt
Ditto, just want to make sure we're comparing apples to apples. I'll run this sometime this week and report back.