Dumb move of the day

[kitekrazy]

I have an external USB nvme 2TB drive and for got to format it and almost filled.

exfat is not a good idea unless you use Mac as well. I don't.

Now I have to start over. Hopefully it is faster. 

[Amberwolf]

note that many if not most ssd type storage is only fast for a tiny amount of data at a time.  after the "cache" (different terms, same idea) is full then everything slows way down.  nearly none are as fast as they claim to be for any kind of sustained write operation. 

[kitekrazy]

Transferring sample libraries with samples that have long files names goes quite slow.

External drives are no match for internal drives but a lot easier to install.

[Amberwolf]

In my experience, it's probqably not the file name length, but the many files involved (often tiny, but not always).   

 

If you have USB3, it's pretty danged fast as long as the port, drive/enclosure, and cables and any hubs in between support it.  

 

The biggest slowdown I've found so far is when the SSD / etc "write cache" (of whatever implementation) is filled by the transfer so the whole process is then slowed down as it has to first write all that to the flash locations it stays in, then clear that cache before it can accept any more data.  That's a slow process for some drives, and some drives have very fast but small caches, whcih work great for typical usages, but not so great for huge data transfers and sometimes not for our type of usages with mutliple big file writes. 

 

The Samsung Evo SSDs I have are terrible at these things, but the Crucial SSDs are much faster.   Both are in the same model and revision of the Sabrent USB-SATA cases, and have been tested on the same ports on the laptop, and the same hubs when that's needed. 

 

But they're all faster than the internal spinny drives I've got.  

Edited Sunday at 11:15 PM by Amberwolf

[kitekrazy]

My portable 2TB nvme drives transferring EW HO to another were writing 32mb a second on a power USB 3 hub.  That's HDD speed.  At least they don't load that slow.

 I don't care now, That project is done.

At last with spinning drives they let you know if they are approaching death, 

[mettelus]

On 12/14/2025 at 6:14 PM, Amberwolf said:

In my experience, it's probably not the file name length, but the many files involved (often tiny, but not always). 

This is typically true, since the file information, especially building/maintaining the Master File Table (MFT) is dependent on number of files and not their size. A million tiny files are going to take longer than one 80GB file to write. Cell complexity (QLC, TLC, etc.) will also affect write speeds fairly dramatically as well as expected longevity...

I didn't even realize that Penta Level Cells (PLC) were in the pipeline till I saw this article (supposed to be released this year, but none exist that I can find), and they are basically designed to be read-only (no proposed expected lifespan on them for writes/erases). The vast majority of consumer SSDs today are either quad (QLC) or triple (TLC) and are geared much more for read operations, with the writes being more intricate/slower per cell complexity. Only the SLC models are designed for "high intensity write operations" without burning out the cells, which is why they are highly preferred for O/S drive usage where they get temp files written to them often and in high quantity, and they are designed to last 30-50 longer than QLC/TLC variants for this specific task.

While QLC/TLCs are great for content that is "fairly static" like sample libraries, data archives, etc. They are not the best choice for something that will be written to often. Even with advances in technology I do still do not trust any SSDs for important archives (or images), so I keep one internal HDD and one external HDD for these purposes.

[Amberwolf]

52 minutes ago, mettelus said:

didn't even realize that Penta Level Cells (PLC) were in the pipeline till I saw this article (supposed to be released this year, but none exist that I can find), and they are basically designed to be read-only (no proposed expected lifespan on them for writes/erases).

Those used to be called WORM memories--write once, read many.  In the days of fuse-roms, they called them OTPs, one time programmables...

Good grief I'm old.