Workstation computer SSD with heatsink issue_Warranty cover?

[paul8f]

Hi,

My Z440 video-editing workstation has intermittent power problems. PSU was replaced 2 days ago, but now the problem looks to be a motherboard issue pulling down the PSU into protection mode.

While investigating, I discovered an unrelated issue:    The PCIe adapter board (plugged into PCIe slot #4) for the 256GB Samsung NVMe SSD has some sort of solder residue on it. I also found that this heatsink mounting hardware for the PCIe adapter board is not making proper contact with all of the ICs on the SSD. (See photos).

It's a lovely machine and thankfully still under its 3-year warranty. The field-Engineer should hopefully have the MB swapped out next week to get it up and running again, but could I ask for a replacement SSD, as surely the lack of heat dissipating ability has greatly reduced the useful lifespan of the SSD? (The machine is just over 2 and a half years old, and the heatsink air-gap is probably there since manufacture.)

Thanks. 

[coromonadalix]

Not solder residues, but flux residues to a badly cleaned board   ,   you can take a "horse hair  tooth brush" to clean this residue yourself.

For the heatsink   wow       If under warranty do an exchange, if not   you can add an thermal pad between the two to cover / fill  the space and help the dissipation


If you dont use the lan PXE  system, enter in the bios and put it to disabled ... will help to start faster

[paul8f]

Hi coromonadalix, I think I just quoted you in the 2000's car airbag thread!

Yes, I was suspecting that it may not have been solder on the PCB. (One sure way to see if it's conductive or not is to do an Ohms test using a DMM and needle-tipped probes I suppose.) I have an ESD-safe cleaning brush, but I'm not going to clean it off, as that would be destroying the evidence of the problem, which wouldn't help my cause in pleading with HP for a new SSD. Same goes for adding a thermal pad myself. (Plus there are different pads with varying properties such as thickness, temperature ranges, and thermal transfer rates etc.) Anyway, at this stage I've lost my confidence in the reliability of the SSD, as it may have been 'working hot' for a quite a while now.... Closing the barn door after the horse has bolted and all that!)

Cheers for the heads up on the PXE boot option in the BIOS/UEFI menu. The system starts pretty quick with the DDR4's and the Z-Turbo PCIe card, but every second helps!   

These issues were not something I expected to see on a high-end €2.5k system

Lovely machine though, so hopefully I can get it back to its former glory.

[bdunham7]

Run a monitoring program like crystaldiskinfo and see how warm in gets in operation.  That photo is blurry--can you see air all the way through to the other side?

[Mr. Scram]

Are you sure that's the actual chip it's supposed to cool? I've seen more than one heatsink which has a bump in the place of the chip it's supposed to cool and which protrudes over other chips. It's not unlikely there are other components that heatsink does serve. The "magnetic lines" look like droplet or bubble residue.

[paul8f]

Run a monitoring program like crystaldiskinfo and see how warm in gets in operation.

When I first bought the machine, I ran a few benchmark tests. In the Windows Experience Index the SSD got the maximum possible score of 7.9 for Win7. When the machine is up and running again after motherboard replacement, it will be interesting to see if the SSD performance shows any degradation. For some reason CrystalDiscInfo v5.6.2 would only recognize the 1TB bulk storage SATA (platter HDD), so I used a program called CPUID HWMonitor to view the temps of the NVMe SSD:

With system idle, the SSD Assembly was 27°C (80°F),   min. was 23°C (73°F),   max was 27° (80°F).
With system idle, the SSD Air flow was ... exactly the same as SSD Assembly.

Don't suppose you know how software like this reports the actual temperature? - Does it take a temp reading from each IC on the SSD and then calculate the average, or does it simply poll just the memory banks?

That photo is blurry--can you see air all the way through to the other side?

I placed some white paper as a background and got a better angle on it (see images). The two large ICs, both to the left have great contact with the thermal pad (thermal sheet.. or whatever it's called). But, with no finger pressure applied to the top of the heatsink, you can clearly see an air gap for the two smaller BGAs on the right. The gap in the groove/slotted end piece, moves from point B to point A with pressure applied down on the heatsink, which closes up the airgap on the BGA to the far right. (The BGA second from the far right still shows a very slight gap with finger pressure applied, so maybe this chip was never intended to have thermal contact).

I'm suspecting that the black shim (PCB isolaton strip...whatever that thing is) should have been placed on the 'solder side' rather than the 'component side' of the PCIe adapter-card PCB, thus pulling down the heatsink towards the top of the SSD. This strip is very thin, so it's location may not make enough of a difference to get BGA and pad contact.

[mariush]

Only the actual controller chip and maybe the DRAM chip (if any) needs to be heatsinked, usually to keep the temperature below around 60-70 degrees Celsius.
The NAND flash memory usually doesn't heat much except during sustained writes for very long periods of times... and the nand chips don't mind a bit of heat... they can safely stay at up to 50-60 degrees celsius.
It's actually often better for the heatsink that touches the controller ic and ram chip (if any) to NOT touch the nand chip.
 
I can't tell from the pictures if heatsink touches the controller of flash memory ... the controller chip is usually closest to the edge connector, along with ram cache near it.

[Mr. Scram]

Only the actual controller chip and maybe the DRAM chip (if any) needs to be heatsinked, usually to keep the temperature below around 60-70 degrees Celsius.
The NAND flash memory usually doesn't heat much except during sustained writes for very long periods of times... and the nand chips don't mind a bit of heat... they can safely stay at up to 50-60 degrees celsius.
It's actually often better for the heatsink that touches the controller ic and ram chip (if any) to NOT touch the nand chip.
 
I can't tell from the pictures if heatsink touches the controller of flash memory ... the controller chip is usually closest to the edge connector, along with ram cache near it.

This is what I was getting at in my previous post. Not all the chips are heatsinked, generally only those requiring it. You don't want to unnecessarily heat up chips that don't get hot themselves and it's generally the controller which gets hottest.

[paul8f]

Are you sure that's the actual chip it's supposed to cool? I've seen more than one heatsink which has a bump in the place of the chip it's supposed to cool and which protrudes over other chips. It's not unlikely there are other components that heatsink does serve. The "magnetic lines" look like droplet or bubble residue.

See my previous reply to bdunham7. I think I kinda answered your question here. But investigating this further, I removed the heatsink from the adapter card, and found that the SSD card itself is a Samsung MZVPV256HDGL-000H1 (SM951 NVMe). I then compared my SSD to an identical SSD for sale on eBay. (Thankfully the seller had the sticker label removed, so I could see what type of ICs I was dealing with.

The two large BGAs to the left are both K90KGY8S7C - CCK0 chips. Each of these is a 128GB TLC (Triple Level Cell flash) NAND chips (32 layer 3D V-NAND). These two BGAs thankfully had proper thermal dissipation. So the IC I'm more interested in would be the controller BGA on the far right. This is the SAMSUNG S4LN058AO1-8030 (dubbed as UBX, it's a PCIe 3.0 x4 controller that apparently supports both AHCI and NVMe driver stacks). I can't seem to find any datasheet or thermal requirements for this Samsung Controller. Anybody know where I could find any info on the S4LN058AO1-8030?

The "magnetic lines" look like droplet or bubble residue.

I don't know what droplet/bubble residue actually is. Is it something to do with the soldering process, and is it a defect or just cosmetic? This area which I highlighted as 'Magnetic Lines' just so happened to be aligned directly across from the hub of the cooling fan in the adjacent GPU card. Coincidence maybe? I used needle tipped probes to try ohming out this residue, but I just got open-cct readings.

I also saw some green type of oxidation on some of the PCIe card fingers (see images attached). Would these green dots explain why the machine would re-boot when the SSD adapter card was gently wiggled side to side. (It may also help explain the 928-Fatal PCIe error - surprise link down error on slot 4 error I got the next time the machine went to POST (guess what.... the SSD comms are through the PCIe slot-4......). The green contamination might also explain the other errors I got: PXE-E61 and PXE-M0F which are related to 'NO BOOT DEVICE FOUND'.

[coromonadalix]

take a white eraser and try to rub the pcie pins, it should be enough, a pink eraser will scratch the pins ...

[paul8f]

I can't tell from the pictures if heatsink touches the controller of flash memory ... the controller chip is usually closest to the edge connector, along with ram cache near it.

The heatsink is not making any contact with the controller chip (yes, the one nearest the SSD edge connector). I was able to push a 0.05mm feeler gauge into the gap. (I couldn't push the gauge all the way through due to the tacky nature of the thermal pad.)

I've done a bit of reading on the SM951 NVMe, and it turns out that:
(a) when used in a heat-sinked adapter card like this, the SSD should have its label removed for better heat transfer (although that voids the warranty on the drive)
(b) this controller chip may or may not have a thermal limiter enabled. This depends on the firmware version employed by the vendor.
(c) during long write operations, data throttling may occur due to excessive heat build-up. (Which I definitely experienced during my video-editing work).

I'm going to highlight these findings to the HP Engineer, on his next call-out. My drive has been heat-stressed over a long period of time, thorough no fault of my own. It's reliability is now compromised.

Samsung SM951 Review:
https://www.legitreviews.com/samsung-sm951-512gb-m-2-pcie-ssd-review_161689/3

Similar kit, but without thermal foam:
https://www.servethehome.com/ngff-m2-pcie-x4-adapter-heatsink/

Samsung (XP941) benchtest video:


The HP Heatsink Kit:
https://www.newegg.com/p/14G-000J-00264

[coromonadalix]

I think you should use an regular sata drive, ie: Maybe an server grade one for intensive workloads

Using a SSD for video editing will be the fastest way to damage it from constant re-writes.

Check in the task manager the actual HDD usage during video encoding - probably it's 1/10 of what that HDD could do ?

Or maybe use an raid array storage ?


https://fstoppers.com/gear/can-super-fast-nvme-drive-help-your-video-editing-235852
https://www.4kshooters.net/2018/03/25/can-an-nvme-drive-really-enhance-your-video-editing-workflow/

[paul8f]

take a white eraser and try to rub the pcie pins, it should be enough, a pink eraser will scratch the pins ...

I'm not going to do this either, as this destroys any evidence of the problem, and also risks damaging the gold-plated finish. The HP Warranty-Engineer will hopefully address this.

A better solution to the eraser would probably be using IPA with a lint-free cloth (or lint free swabs). Gentle strokes should remove the contamination. When my motherboard gets replaced, popping in the SSD PCIe adapter card without cleaning the card fingers first, would probably cross-contaminate the new MB connector slot. This is a type of "hardware virus" I suppose. Just another reason why I feel the SSD and the adapter card should also be changed.

Bottom line... Professional Workstations and Business Server-grade equipment should be reliable and dependable, under a broad range of workloads. If I see something that may compromise this, I'm going to speak up. This Z440 was not a cheap machine.

[paul8f]

I think you should use an regular sata drive, ie: Maybe an server grade one for intensive workloads

I'll take your advice on board, but this Workstation was marketed (and optimized) especially for 4K video editing. This Samsung SSD broke a few records at the time of its release a few years back. Very good all-round performance especially when used through a PCIe slot with cooling, rather than an M.2 slot. I've read in a few places that for 4k editing, a Solid State Drive is a must. Even better is having multiple SSDs to split the OS from the cache and software. Then to use a large capacity SATA drive for archiving your footage, soundtracks and project files.

Anyway, for various reasons, the video editing business didn't work out for me, so I'll probably just end up using the workstation as an everyday PC and a gaming rig. My poor old laptop is on its last legs, and needs to be retired soon.

[coromonadalix]

I know technology evolve pretty fast, you have better tech in ssd / nvme drives nowadays with better error corrections and better lifespan,  maybe one day you'll re-think this ??

I do use kinda old ssd, i see they have begun to slower down a little and i use 50-75% of their capacity, never more than 75%

Last time i've played in video editing was with an Matrox rxt100 and Sony Vegas "marriage souvenirs"  loll

[Mr. Scram]

I'm not going to do this either, as this destroys any evidence of the problem, and also risks damaging the gold-plated finish. The HP Warranty-Engineer will hopefully address this.

A better solution to the eraser would probably be using IPA with a lint-free cloth (or lint free swabs). Gentle strokes should remove the contamination. When my motherboard gets replaced, popping in the SSD PCIe adapter card without cleaning the card fingers first, would probably cross-contaminate the new MB connector slot. This is a type of "hardware virus" I suppose. Just another reason why I feel the SSD and the adapter card should also be changed.

Bottom line... Professional Workstations and Business Server-grade equipment should be reliable and dependable, under a broad range of workloads. If I see something that may compromise this, I'm going to speak up. This Z440 was not a cheap machine.

If I'm totally honest you seem to be looking for issues until you've found something to be suitably unhappy about. Are there any actual issues or are you just looking for them by wiggling cards until they cause errors?

[TheAmmoniacal]

Most setups use these SSDs without any heatsinking, it should not be a problem. Even minor case air flow is good enough to prevent throttling.

[Nerull]

I think you should use an regular sata drive, ie: Maybe an server grade one for intensive workloads

Using a SSD for video editing will be the fastest way to damage it from constant re-writes.

That's just plain false.

[paul8f]

If I'm totally honest you seem to be looking for issues until you've found something to be suitably unhappy about. Are there any actual issues or are you just looking for them by wiggling cards until they cause errors?

Just trying to root cause the boot up issue, and this "wiggle test", while not very technical I admit, proved repeatable. With the upmost care, very gentle sideways pressure on the card forced the machine to re-boot. It re-booted three times in a row with this test, and then never again after I removed the card, applied a little compressed air to the fingers and MB pins, and re-seated the card. (Not conclusive I know, but still some sort of result.)

I'm not simply hunting around for issues to complain about. I bought an expensive machine that's now giving trouble, and I stumbled across a semi-related problem. I enjoy figuring out why things malfunction, and I'm on this forum looking for advice from like-minded people. Let's hope that what I've discovered wasn't a tactic of planned obsolescence!

[paul8f]

Most setups use these SSDs without any heatsinking, it should not be a problem. Even minor case air flow is good enough to prevent throttling.

Yeah, the fact that many SSD setups don't employ any heatsink did occur to me, but the problem here is that the big useless unconnected heatsink is blocking any airflow from getting to that controller BGA chip. This IC is probably slowly cooking away in it's own dissipated heat as well as the heat reflected back from the heatsink, while any air currents that would offer cooling are diverted away.

This SSD drive is worked hard during video editing, and an air gap like this is the last thing you need.