Question Can one determine the write speed of an SSD after running the buffer out?

[pvwebb]

That is based on the published specs. Obviously one can measure it with the SSD in hand. Can one determine before purchasing?

I'm in the situation of copying 58 files over 330gb combined in a single transfer to a 1TB SSD as fast as reasonably possible. This is a routine procedure to update more than one computer (i7 32gb RAM) and not just a onetime process. The source is on a USB 3.2 Gen2 2x USB C port which is fast. When the buffer runs out, I've found the write speed changes drastically depending on the model of the internal SSD destination for the files. At least one model can keep a write speed of about 480 mb/s while other slow down to less than 50mb/s. I haven't found a way to check this speed in the published specs (write speed when the buffer runs out). I have found it is better to copy to a different physical drive than the drive with the OS. Formatting the destination drive had no effect, so it appears to be the physical parameter of the SSD. The models of the source and computers are all the same. The only difference is the destination SSD.

 

[Nine Layer Nige]

I guess it also depends on how fragmented the files to be copied are and how deep the queues. You dont mention the SSD's in question on either the read or write. Heat may also play a part in performance.

 

[Bob.B]

That is based on the published specs. Obviously one can measure it with the SSD in hand. Can one determine before purchasing?

I'm in the situation of copying 58 files over 330gb combined in a single transfer to a 1TB SSD as fast as reasonably possible. This is a routine procedure to update more than one computer (i7 32gb RAM) and not just a onetime process. The source is on a USB 3.2 Gen2 2x USB C port which is fast. When the buffer runs out, I've found the write speed changes drastically depending on the model of the internal SSD destination for the files. At least one model can keep a write speed of about 480 mb/s while other slow down to less than 50mb/s. I haven't found a way to check this speed in the published specs (write speed when the buffer runs out). I have found it is better to copy to a different physical drive than the drive with the OS. Formatting the destination drive had no effect, so it appears to be the physical parameter of the SSD. The models of the source and computers are all the same. The only difference is the destination SSD.

Tell the make and model of the fast ssd and the slow ssd.

 

[pvwebb]

The fast drives are model 1TB Samsung 870 EVO and the slow drives are Crucial BX500 - SSD - 1 TB.

 

[USAFRet]

The fast drives are model 1TB Samsung 870 EVO and the slow drives are Crucial BX500 - SSD - 1 TB.

And that is why.

The BX500 have no DRAM.
https://www.tomshardware.com/reviews/crucial-bx500-ssd,5377.html


Once that cache fills up...slooooow.

 

[Bob.B]

The fast drives are model 1TB Samsung 870 EVO and the slow drives are Crucial BX500 - SSD - 1 TB.

Yup...dram.

 

[pvwebb]

And that is why.

The BX500 have no DRAM.
https://www.tomshardware.com/reviews/crucial-bx500-ssd,5377.html


Once that cache fills up...slooooow.

But I don't think you are implying that the BX500 is the only slow SSD.

 

[USAFRet]

But I don't think you are implying that the BX500 is the only slow SSD.

No, not at all.

 

[Bob.B]

The fast drives are model 1TB Samsung 870 EVO and the slow drives are Crucial BX500 - SSD - 1 TB.

On a bx500 turn write caching on/off see what effect it has on the xfer.

 

[pvwebb]

On a bx500 turn write caching on/off see what effect it has on the xfer.

I saw no noticeable difference for the bx500, probably because the size of the files copied made the cash insignificant. With cache, one sees a burst of speed for a few seconds, then one sees the copying progress at the speed of the memory beyond the cache. So, so only a very small difference.

If I put 512GB of DRAM in the computer I might get caching software to release the external drive quickly, but it would still be risky if something interrupts the cache in the >1hr. to empty the cache. It costs less to get a fast SSD.

 

[pvwebb]

No, not at all.

Do OEM's publish any data that indicates this DRAMless issue, or is it up to finding out from reviews or customer testing?

 

[USAFRet]

Do OEM's publish any data that indicates this DRAMless issue, or is it up to finding out from reviews or customer testing?

No, they don't publish that.

"Hey look...our drives slow down more than others!"

 

[pvwebb]

No, they don't publish that.

"Hey look...our drives slow down more than others!"

But it seems OEM's would want to report "Hey look...our drives are fast and don't slow down like others. Thus, here's the write speed when cache is full. SSD's may not report this parameter when they are slow."

 

[USAFRet]

But it seems OEM's would want to report "Hey look...our drives are fast and don't slow down like others. Thus, here's the write speed when cache is full. SSD's may not report this parameter when they are slow."

A LOT of people won't notice the difference between a DRAMless BX500 SATA SSD, and a hotrod 980 Prp PCIe 4.0 SSD>
Seriously.
I know people like this.

That level of advertising would be lost on them.

It is faster response than an HDD....thats all they see and know. Sometimes, not even that.

 

[Misgar]

Obviously one can measure it with the SSD in hand. Can one determine before purchasing?

I look at reviews on Tom's and other sites showing IOMeter Sustained Sequential Writes when choosing a drive.

You observe initial high write speeds until the buffer is full, then a sudden drop to the underlying (unbuffered) native write speed. In the case of DRAM-less SSDs, a section of the drive is sometimes configured as a temporary SLC buffer, but when this fills up, writes drop back to TLC/QLC rates.

I observe this behaviour when copying 4K GoPro videos to my laptops when on vacation. A 15-minute video is roughly 11GB and backing up a 256GB microSD card in a fast card reader often saturates the buffer in my NVMe drives.

https://www.tomshardware.com/pc-components/ssds/samsung-990-evo-plus-ssd-review/2