A reason to stick with SATA SSDs, rather than PCI-E?

[VirtualLarry]

I mentioned this in a post in another thead, but I got to thinking, it's an interesting example of how SATA is inherently superior to consumer PCI-E SSDs, really. It's all due to the inherent hot-swappability of SATA.

I had a TLC 2.5" SATA SSD, plugged into a test rig, that hadn't been booted in nearly a year. Well, you know TLC, it degrades from being non-powered, and when I finally fired it up, it was having severe trouble. It couldn't really boot Windows, would hang at 7% at the disk check stage, etc. A real mess.

So I booted a Linux Mint distro off of USB, and had another donor working SATA SSD plugged in, and then I hot-swapped the "bad" drive in at the last minute before executing HDPARM commands, to Secure Erase the drive.

That apparently "took", and I was then able to re-install Windows 10 onto it. After installing Windows 10, some things were taking a "long time", but then eventually things smoothed over.


Oh, and at the disk partitioning scheme, I created partitions, formatted them, then deleted them and re-created them, which should have triggered a full TRIM pass over those partition areas too.

It works now.

If I had a PCI-E M.2 TLC SSD (I do own several of these), and it "died" from dis-use / being powered-down too long, then I'm not really sure how I would "revive" it. I mean, I don't believe that M.2 PCI-E slots are hot-swappable in consumer motherboards, and you can't use HDPARM's secure-erase command tools on an NVMe SSD anyways. So I would be stuck, throwing the drive away.

 

[eton975]

Would a simple reformat be an issue? Are you worried about the unpowered cells behaving unpredictably when written to again?

 

[VirtualLarry]

Would a simple reformat be an issue? Are you worried about the unpowered cells behaving unpredictably when written to again?

The BIOS wasn't always seeing the SSD, it wasn't internally initializing properly.

 

[Elixer]

I must say, I do love hot-swapping, can't really live without it IMO.

That said, they actually DO have hot-plug PCIe cards.
Check out http://www.iocrest.com/en/product_details321_a.html

That is both hot-swap & hot-plug.

PCI EXPRESS BASE SPECIFICATION, REV. 3.0

6.7.
PCI Express Hot-Plug Support
The PCI Express architecture is designed to
natively support both hot-add and hot-removal (“hot-
plug”) of adapters and provides a “toolbox” of
mechanisms that allow different user/operator
models to be supported using a self-consistent infrastructure. This section defines the set of hot-plug mechanisms and specifies how the elements of hot-plug, such as indicators and push buttons,must behave if implemented in a system.

 

[VirtualLarry]

I don't believe that M.2 PCI-E slots are hot-swappable in consumer motherboards,

Note that I used the word "consumer". I was aware that the bus specification supports hot-add/remove, but generally, most home-user motherboards don't support that, and I don't believe that the M.2 slot does either. I mean, I could try it, but I would be concerned about burning something up.

Edit: There are a specific few higher-end mobos, that have DIP switches to enable/disable PCI-E slots on the mobo. In theory, you could switch a slot off, remove/hot-plug it, and switch it back on. I have no idea what would be required on the software side of things, or in Windows.

Edit: Are you talking about SATA hot-plug, or PCI-E hot-plug? I was talking about PCI-E hot-plug, although I mentioned using SATA hot-plug to fix my SATA SSD that had gone south.

 

[mikeymikec]

I'm sticking with SATA SSDs for the time being because IMO M.2 is a mess of standards with advantages and disadvantages. Maybe if a customer came along who needed all the storage I/O they could get then maybe I'll consider M.2 (or more likely, pure PCIE NVMe), but I'm still not that enthusiastic about adopting a storage medium that I can't easily externally mount, and why on earth would I bother with the SATA M.2 drives (apart from say in a NUC-type unit).

Maybe M.2 will undergo a fundamental revision to make it far more elegant than it currently is, though I think they will have to throw out backwards compatibility in the process just for that revision.

 

[whm1974]

Larry, to begin with you brought TLC SSDs. Which I do not trust, especially with NVMe as the higher I/O will cause the the drive to die sooner.

 

[Red Squirrel]

There are limited PCI-E ports, and they arn't really hot swap. You can get a much larger number of SATA ports and they support hot swap.

Though with the introduction of GPU crypto mining, there are now motherboards that have like 13 PCI-E slots. So I suppose this could be used for PCI-E SSDs too. But you still lose out on hot swap which is kinda important for any kind of mass storage.

 

[razel]

Yes, if your system has other limitations. I played with the Nvme off my office work rig. It's Core2 2010 era with 500GB 7200 RPM HDD, but with plenty of RAM. It's PCIe Gen 2 with x1 (500mb) and x16 (8gb) slots only. Samsung 960 Evo m2 Nvme put into a m2 to PCIe x4 adapter. I can only put it the x16 slot since the x1 slots don't have an open 'back'. Doing so I lose one of my digital video outputs. Performance was as expected since the adapter is limited to x4 and the motherboard is gen 2 so that's 2gb. Benchmarks are 1.2 to 1.5GB range.

I ended up removing it since two monitors are more important than NVMe and as you mentioned, I can still put SATA SSD in there. Nvme was awesome for the week I had it and I do notice subtle improvements over SATA SSD, but again given the choice, SATA SSD with 2 monitors beats out NVMe with 1 monitor.

Well technically, if I Dremel out the back of the PCIe x1 slot I could still use it at the max 500mb/s speed, but it is an office work computer that I didn't buy.

 

[Glaring_Mistake]

I had a TLC 2.5" SATA SSD, plugged into a test rig, that hadn't been booted in nearly a year. Well, you know TLC, it degrades from being non-powered, and when I finally fired it up, it was having severe trouble. It couldn't really boot Windows, would hang at 7% at the disk check stage, etc. A real mess.away.

I'm a bit curious what drive it is, could you mention which it is?
Because from what I've seen even TLC can keep from degrading fairly well.

 

[XavierMace]

I mentioned this in a post in another thead, but I got to thinking, it's an interesting example of how SATA is inherently superior to consumer PCI-E SSDs, really. It's all due to the inherent hot-swappability of SATA.

I had a TLC 2.5" SATA SSD, plugged into a test rig, that hadn't been booted in nearly a year. Well, you know TLC, it degrades from being non-powered, and when I finally fired it up, it was having severe trouble. It couldn't really boot Windows, would hang at 7% at the disk check stage, etc. A real mess.

So you're saying SATA is better because it makes it easier to recover your drive when you do something you shouldn't be doing with it either way? That's quite the view point there.

 

[VirtualLarry]

So you're saying SATA is better because it makes it easier to recover your drive when you do something you shouldn't be doing with it either way? That's quite the view point there.

Pretty quick to accuse, don't you think? Or is leaving an SSD disconnected, outside of a system, without power, "something you shouldn't be doing with it"?

I didn't do anything "strange" with the drive. It was pulled from a working system, and set aside for nearly a year. Plugged it back in, errors, non-detection by BIOS, generally not feeling well. Explanation? TLC degradation, it's real.

 

[whm1974]

Pretty quick to accuse, don't you think? Or is leaving an SSD disconnected, outside of a system, without power, "something you shouldn't be doing with it"?

I didn't do anything "strange" with the drive. It was pulled from a working system, and set aside for nearly a year. Plugged it back in, errors, non-detection by BIOS, generally not feeling well. Explanation? TLC degradation, it's real.

In that case, do not buy TLC SSDs. It is very simple.

 

[XavierMace]

Or is leaving an SSD disconnected, outside of a system, without power, "something you shouldn't be doing with it"?

That's correct. The risk of leaving SSD's disconnected from power is long established (especially cheap drives) and an issue that normal people won't run into. If this is the drive you're using for "regular" backups, it wouldn't be sitting unused for a year. Nobody in their right mind would use an SSD for cold storage as, even ignoring the data loss risk, it's literally the most cost inefficient device to use for long term archival. However none of those scenarios apply here, your situation is more unique and entirely self made because you have a mountain of computers that sit around unused for extended periods of time.

With all that said, I've re-read your original post 4 times now and still can't figure out how hot swapping the drive was a factor in getting it fixed. But like @whm1974 said, don't buy TLC drives. Or at least don't leave them sitting around... unplugged.... for over a year.... with data on it. Plus, given you seem to have gone out of your way to avoid mentioning the specific drive, I'm assuming this is probably one of the bargain bin SSD's you bought.

 

[Red Squirrel]

Wait, it's bad for SSDs to be left unused for a long time? Like for data integrity or the drive itself? First time I hear of this.

 

[VirtualLarry]

because you have a mountain of computers that sit around unused for extended periods of time.

Exactly the scenario, this was an SSD pulled (working) from another rig, plugged into my Skylake Pentium "test rig", for testing / benchmarking / erasing SSDs, and not booted for quite some time. (Might not have been a full year, maybe 8 months.)

The reason that I had to hot-swap the drive, is because if I booted with the drive attached, sometimes the BIOS wouldn't detect it properly, and Linux Mint wouldn't detect it properly either, some sort of ATA SENSE error. Mint wouldn't even detect it was an SSD. I had to hot-swap with an Adata SSD, and swap the defective / non-detecting SSD in at the last moment, before hitting it with the HDPARM commands. I've had to use that method before, to Secure Erase a finicky SSD, that wasn't always detected by the BIOS or Linux properly.

It seemed to work, though.

The SSD in question was a "Radeon R3 SSD", which you can't even source the actual mfg of it.

 

[XavierMace]

Wait, it's bad for SSDs to be left unused for a long time? Like for data integrity or the drive itself? First time I hear of this.

Sometimes I wonder if you vets realize there's tech information on the internet from other sources than AT. But in this case, AT had an article on it as well, like 2-3 years ago.

 

[mxnerd]

This SSD will die prematurely seems to be a myth.

According to this article, it only happens if it's reaching its end of life then unplugged and store at abnormal temperature.

https://www.pcworld.com/article/292...nt-lose-data-if-left-unplugged-after-all.html

 

[VirtualLarry]

Sometimes I wonder if you vets realize there's tech information on the internet from other sources than AT. But in this case, AT had an article on it as well, like 2-3 years ago.

The thing is, and I believe that this is mentioned in the article, JEDEC specs for P/E cycles, basically state that the SSD should retain its information at room temp for a minimum of a YEAR, ONCE rated P/E cycles were EXHAUSTED.

Yes, TLC sucks, and I guess I didn't realize how badly, but even still, this was not (IIRC) a heavily-used SSD, and shouldn't have been at the limit of its P/E cycles, and thus should have retained its information for over a year.

 

[XavierMace]

Exactly the scenario, this was an SSD pulled (working) from another rig, plugged into my Skylake Pentium "test rig", for testing / benchmarking / erasing SSDs, and not booted for quite some time. (Might not have been a full year, maybe 8 months.)

The reason that I had to hot-swap the drive, is because if I booted with the drive attached, sometimes the BIOS wouldn't detect it properly, and Linux Mint wouldn't detect it properly either, some sort of ATA SENSE error. Mint wouldn't even detect it was an SSD. I had to hot-swap with an Adata SSD, and swap the defective / non-detecting SSD in at the last moment, before hitting it with the HDPARM commands. I've had to use that method before, to Secure Erase a finicky SSD, that wasn't always detected by the BIOS or Linux properly.

It seemed to work, though.

The SSD in question was a "Radeon R3 SSD", which you can't even source the actual mfg of it.

Hot plugging doesn't get around ATA commands. Keep rebooting and it would have been detected eventually. Basically your argument is amounting to "I like cars without hoods on them because it's easier to jump start my dead battery". Rather than declaring SATA SSD's superior due to this ridiculous scenario, perhaps you should reevaluate your tech habits. I've CURRENTLY got 30 SSD's in the house. If I count past SSD's, add at least another dozen to that. If I count PCIe Flash Accelerators, that technically adds another two dozen as well. I've never had to jump through any of the hoops you've described in this thread.

 

[VirtualLarry]

Hot plugging doesn't get around ATA commands.

Actually, it does. Let me explain. The BIOS, when it boots, does some ATA SENSE and DRIVE IDENTIFY, and also attempts to perform a READ SECTOR 0 (boot sector), among other things. If the drive is in a (bad) state, these initial commands can cause the drive to internally hang, until power is remove, and re-applied.

Hot swapping the drive, allows that to happen, such that the FIRST command it sees, is the SET PASSWORD, followed IMMEDIATELY by the SECURE ERASE command. Without the other BS commands the BIOS spews at the beginning.

I had to discover this the hard way, as I had another drive that was so badly corrupted, that the BIOS drive init stuff would cause it to internally hang, but I discovered, using a working mule drive, and then hot-swapping in the broken drive, would allow it to get Secure Erased, and thus another power cycle, a TRIM pass, and it was generally good to go again. (OK, the TRIM pass may be redundant.)

Edit: Not this Radeon R3 drive, but the original OCZ drive that I revived this way (Hot-Swap SE technique), would basically hang on any "Read" of the drive. And since the BIOS tries to read from the drive at boot, well, that caused it to hang. My theory is, the "Read" command, causes the SSD to try to parse the mapping table internally, and if that's royally screwed up, then it panic hangs, to protect the data, but if you hot-swap it, it can process the ATA PASSWORD and SECURE ERASE commands without processing the mapping table (although, the SECURE ERASE should clear / reset it).

Edit: This hot-swap, is done with both drives with power applied, but no SATA cable on the bad drive, instead, it is plugged into the donor drive, until just before hitting ENTER on the command-line containing the HDPARM commands. (Don't Secure Erase the wrong drive!)

Also, "Hot Swap" was DISABLED on that SATA port in the BIOS (Intel). If it were Enabled, it might generate some signal, causing the drive to attempt to re-identify the drive, and interfere with our "raw" hot-swap procedure.

 

[cbn]

The SSD in question was a "Radeon R3 SSD", which you can't even source the actual mfg of it.

Interesting, according to Johnny Lucky's website and this one that SSD has Toshiba TLC. (This is the first time I have heard of that NAND doing this)

 

[VirtualLarry]

I've never had to jump through any of the hoops you've described in this thread.

One word: "OCZ".

 

[XavierMace]

One word: "OCZ".

My OCZ history:

OCZ Apex
OCZ Onyx
OCZ Agility
OCZ Agility 3
OCZ Vertex
OCZ Vertex 2
OCZ Vertex 3
OCZ Vertex 4

At least two of each. I've still got two of the Vertex 4's and one of the Vertex 3's on a shelf in the closet. But outside of those last two, I've never left any of them sitting around without power.

 

[whm1974]

With all that said, I've re-read your original post 4 times now and still can't figure out how hot swapping the drive was a factor in getting it fixed. But like @whm1974 said, don't buy TLC drives. Or at least don't leave them sitting around... unplugged.... for over a year.... with data on it. Plus, given you seem to have gone out of your way to avoid mentioning the specific drive, I'm assuming this is probably one of the bargain bin SSD's you bought.

Another rule of thumb to add, don't buy bargain bin SSDs either.