SSDs - State of the Product?

Looks like an overall decent m.2 nvme drive with 5100TBW over 5 years is pretty generous (2.8TB/day writes). Why would the firmware be whacked and no good for every day use? When I read it's labeled for NAS-use, I think it implies higher durability or larger amounts of writes per day or even may have write loss protection (I didn't confirm this). The performance metrics look pretty decent to me, what is your expected use case?

My 1TB SSD is built in so I can't expand but I'd have no issue using this nvme if I could.
 
I would be using it multiple times/places over 2-3 year lifetime, e.g., in a laptop part time, otherwise externally enclosed, and maybe even in a desktop for data. I'm wondering how that SN700 can be so cheap, yet still be TLC and have the DRAM.
Mostly I want to know what the sustained write speeds are for example if I want to copy 1+ TB to SN700 in one shot. Most of the benchmark sites are always full of it, using test that fit into the SLC part of the drive and not what happens down the line when the drive is dirty from heavy use.

The new SR site has some data on the SN770, but it makes no sense. The writes are plotted as µs vs. MB/sec. and most others as µs vs. IOPS. Some of the other graphs are just nuts with more than one Y value for the same X value. Is this some "new math" that changed in the 21st century? The VDL looks like some kid drew them by hand. I don't know what to make of all this.
 
Great. :( I don't like to order products with vague delivery dates as packages occasionally don't arrive or disappear when I'm gone. By the time they are in stock who knows.
 
They are using a bunch of Seagate SATA SSDs and a sprinkling of others? I don't think that data has much value.
 
What are people using as USB-C enclosures for 2.5" SATA SSDs? My older USB 3.0 enclosures are dicsonnecting with the 4TB drives. I think it is a power issue because there are no errors when writing/reading the whole 4TB on a natural SATA port. :(
 
I have a bunch of Sabrent 2.5" USB enclosures for general purpose backup drives and I've used SSK nVMe enclosures for faster stuff. I can't say I've seen many disconnects.
 
It's starting to seem like 2.5" SATA SSDs are becoming a thing of the past. I have a bunch of server systems out in the world that don't have u.2 backplanes and still have a couple years of life in them. Intel D3-S4510s, which were released in 2018, are kind of my product of last resort, somehow offering ~2 DWPD endurance for a rated 5 year lifespan on TLC NAND.
 
Pretty much everything I want to do is becoming a thing of the past. I'd be fine with U.3, but don't expect to ever see it in a normal computer. M.2 is stupidly small for power/heat production. All the TLC 2280 SSDs >2TB have chipotle on the underside, which is even worse.
 
Pretty much everything I want to do is becoming a thing of the past. I'd be fine with U.3, but don't expect to ever see it in a normal computer. M.2 is stupidly small for power/heat production. All the TLC 2280 SSDs >2TB have chipotle on the underside, which is even worse.
You can add these easily to a normal computer. I've setup a few machines to use U.2 SSDs in 2.5" form factor with this PCIe card to SFF-8639 with a U.2, SFF-8643 to SFF-8639 Cable. Not terribly expensive components and you can leverage the full speed of an enterprise U.2 NVMe SSD assuming you have the available PCIe lanes on your motherboard.
 
I don't think an adapter will fit in my current system due to the clearance. I'd like to see six of those ports like we have SATA.
 
The pcie card is tiny for what it's worth.

There are other more expensive adapters with 4 ports but they need lots of pcie lanes (16) if you want nvme.

An intermediate would be a SAS 12gb adapter to connect more SSDs with 4-8 ports.
 
Looks like an overall decent m.2 nvme drive with 5100TBW over 5 years is pretty generous (2.8TB/day writes). Why would the firmware be whacked and no good for every day use? When I read it's labeled for NAS-use, I think it implies higher durability or larger amounts of writes per day or even may have write loss protection (I didn't confirm this). The performance metrics look pretty decent to me, what is your expected use case?

My 1TB SSD is built in so I can't expand but I'd have no issue using this nvme if I could.
The RED SSDs are a bit strange. For no good reason the write performance is reduced in the x570 setup, but normal in the laptop.
 
The pcie card is tiny for what it's worth.

There are other more expensive adapters with 4 ports but they need lots of pcie lanes (16) if you want nvme.

An intermediate would be a SAS 12gb adapter to connect more SSDs with 4-8 ports.
That's a big part of the problem with the mainstream boards that only have 16-24 lanes. Unfortunately the higher boards use the stupidly expensive CPUs with excessive cores that don't help most mainstream programs that need decent single-thread performance.
 
I hear ya. An alternative is to look for a used server class system that is a generation or so older to save some money but gets you more pcie lanes. That or go with much larger nvme drives to use fewer lanes but they get really expensive too.
 
I'm leaving the 2x 4TB in the laptop and hope to use that in the summertime. I was hoping to also use it in the desktop when in the states, but the slower write speeds compared to the original 2x 2TB 970EVO+ is not worthwile.
Computing its becoming very unpleasant and upsetting with the Win 11. I built or rebuilt 5-6 Win 10 systems in 2021 so I'm not really planning to buy any more computing hardware this year, but the Big MAC may be in my future. I was planning to upgrade storage this year.
 
I've read some rumblings that PCIe Gen 5 might run nVME on two lanes instead of four, simply because the bandwidth will be there to support it. That would definitely make desktop consumer platforms a lot more appealing again.
 
I suppose but somehow that won't help with current SSDs and might stunt the growth of future chipsets. Over 10 years ago I had 44 lanes IIRC.
 
I suppose but somehow that won't help with current SSDs and might stunt the growth of future chipsets. Over 10 years ago I had 44 lanes IIRC.

That was the driver for my choice to go to Threadripper instead of a consumer desktop platform.
I'm not sure if there are any plans for another generation of Not-Pro Threadrippers going forward, unfortunately.
 
If I wanted to take a motherboard (ASRock Rack C236 WS) I'm using to run TrueNAS and turn it into a desktop PC for light use, would I see much benefit to putting an M.2 SSD into a PCIe (3.0) adapter versus just using SATA? Is there a preferred reliable M.2 to PCIe adapter or are they all about the same?
 
When you say M.2 without specifying, I am assuming you're talking about B-keyed drives, in which case they are 100% just SATA drives, whether they're in an M.2 port or a SATA one. If you have M-keyed drives, those are the PCIe ones that have vastly higher read speeds. In practical terms, the only time there's REALLY an objective difference is on an OS drive during boot or large program loads, or someone who regularly deals with very large file loads. It's hard to see an objective difference otherwise.

The system I'm using right now has a bunch of 1 and 2 TB nVME drives on PCIe adapters and a number of SATA SSDs and other than assigning what drives I'm using for what purpose in the first place (e.g. making sure that an nVME drive is where I'm dumping raw data from my camera), I've never given any thought to what's going on. I don't think it matters for most people most of the time.
 
I would not waste a PCI slot in such an old system just for the boot drive. An 860 EVO or 870 EVO should be fine.
 
I have a bunch of Sabrent 2.5" USB enclosures for general purpose backup drives and I've used SSK nVMe enclosures for faster stuff. I can't say I've seen many disconnects.
All three USB-C enclosures disconnected sometimes and corrupted my data! At first I thought it was only the file being copied at the moment and deleted each before the next cycle, but back in the states it was observed there are a whole bunch of files with adreaded red X on them. (File size and date are normal, but they are internally blanks.) :mad:

What a cluster. My goal was 5 copies of <4TB of data (42,000 RAW) images from me and less than half od that from another person).
I ended up with only the internal drives and some of the 4x2TB drives in separate enclosures that I won't be able to access now until next year because they were returned to the states separately for security.
 
Also might want to consider something like TeraCopy, that allows verify-before-delete move operations. I've been using it for years for that purpose and it hasn't given me a problem yet. Some say it is quicker than Windows file manager for plain copies, but I really don't notice a difference. I use it for the verify option, not for speed.
 
Have you thought about syncing everything to a cloud service or are you places where bandwidth is a distinct issue?
Haha. >1TB/week in the middle of nowhere. We can use phones (mine won't work though), but no radios as the area is under a foreign control. Wi-Fi is very slow, maybe satellite or something like that. This is personal, so no company assets are available.

I'm still operating under the assumption that the problem is power related. I have two devices that have internal batteries and will charge via the USB also. Tests seem OK, but so did the others.
 
Also might want to consider something like TeraCopy, that allows verify-before-delete move operations. I've been using it for years for that purpose and it hasn't given me a problem yet. Some say it is quicker than Windows file manager for plain copies, but I really don't notice a difference. I use it for the verify option, not for speed.
Is there some way to have it make multiple copies in parallel?
 
What I mean by parallel is that there is one source folder and three targets. The program should read each file once, write it to three targets, read the next one, etc. The three targets should be able to use Windows to buffer the writes so the overall performance is similar to a single copy. I'm not sure what that type of copying is called.
 
If you really are looking to do this in software, something like using rsync in a loop would be a way to copy each file to multiple destinations but this feels very unpredictable especially if you have a large churn rate or long data sync times. You will have large periods of inconsistent backup sets assuming this is all for backup purposes.

I guess I don't fully understand this type of use case of why three separate copies with the potential for high inconsistency are needed like this versus replicating backups.
 
I've been using software to download up to three simultaneous copies since 2005 and was curious about the possibility of the data integrity check as part of the process as mentioned above. The main issue with the software is that if there is a copy error (such as a dismounted drive) every file generates an error dialog, but there is no easy way to stop the software. Errors were extremely rare until recently. I was just wondering if there was something better.

Time and risk are the main factors in making multiple copies now. Rest and sleep are valuable commodities on most of my trips, so two clicks to three copies is a no-brainer. It's critical to have multiple copies in multiple locations (at least different carry and checked bags) in case of loss, theft, flood or damage. In the states I look for east, west, and middle storage.
 
I exceeded the capacity of the 4TB drives and that created a data management nightmare in the field (ocean really). However, I'm not seeing 8TB SATA III drives (except the old, slow-writing QVO) or 8TB external USB drives either. It's 2022 so why not already? :(
 
Is there an USB enclosure for that? It must be powered from the natural laptop ports.
 
All the units I've looked at are just a straight attachment for a drive, but I don't see that as a problem for a 2.5" SSD; you can buy a little case to put it in when you aren't using it.
 
Back
Top