SanDisk CloudSpeed Eco 1.92TB 2.5" SATA $120

[Sean Ho]

In any case, use fio for testing SSD write speed

 

[UhClem]

Pny cs3030, which I have, does something similar

PNY XLR8 CS3030 NVMe SSD review: Cheap and exceptionally fast for everyday chores

PNY's CS3030 NVMe SSD is a great everyday performer. For the average user, it's a bargain, but it slows more than usual for a bargain drive during long writes.

www.pcworld.com

It can write zeros all day long at 1.7GBs but random is a lot slower.

Thank you. That is interesting.

I suspect that you could demonstrate that same behavior using, e.g., dd :

dd if=/dev/zero of=/dev/nvme0n1_of_cs3030 bs=4M count=12000

vs

dd if=/mnt/Xtmpfs/random_48GiB of=/dev/nvme0n1_of_cs3030 bs=4M count=12000

After a quick perusal of the NVMe spec, there IS a command (optional feature) for Write_Zeroes, which would offer a mechanism for this optimization--and, hence, require a workaround to prevent its possible availability, and use, from (deceptively) skewing NVMe test results.

Two points to note:
1. No such mechanism exists in the ATA spec (applicable to SATA SSDs) [and very unlikely in SCSI spec (for SAS SSDs)]
2. (assuming that one can replicate the linked test result) For that controller, said optimization only appears to be applied when the pseudo-fast-cache is in full play; else its write-zero rate would (probably/likely) always be >> 1.7G/s.

 

[nexox]

Well I have looked through my old benchmark data and don't have anything specifically about writing zeros to the early Sandforce drives (though I do have a chart that shows writing the same block of random data repeatedly is quite a lot faster than fully random data, there's not enough documentation of methods to go with that one to be very convincing.) I did a quick survey and none of my personal drives seem to differentiate between zero and urandom writes, but they're all either Intel or these Sandisks, so don't really cover a lot of the market.

[Specifications (ATA & NVMe) dictate that a WRITE (device cmd) results in a write to the media.]

Technically an update to the metadata is a write to the media, and in my experience "technically" is how SSD makers like to follow the specs, that's how some of those Sandforce drives could lose 20+MB of data which had successfully been fsync'd on power loss. The spec said they could return the fsync command when the data had been written durably, but, because at that time it was written with just spinning disks in mind, it did not explicitly require that the metadata required to read the data back must be written durably, so after reboot you would read back zeros or older version of the pages. Even Intel played that game, to a lesser extent, until they added PLP in the third gen drives.

 

[A33]

Hi,

I have tried the drives for a few days and all seem to work fine. I'd like to update de firmware as the are running ancient ones (from version 0 to version 3 vs. currently ZR11). I have search online but the firmware upgrade process does not seem to be that straightforward.

Firmware available on: CloudSpeed Gen. II Eco & Ultra SSD Drive Firmware (ZR11) (wd.com)

1. Connect the drives to a Windows machine
2. Download de firmware from wd.com. There seem to be several firmwares on the wd zip file, how should I choose the correct one?
2. Run the SanDisk Guard Tool for Windows.
3. Update the firmware. Here is where I'd like input from someone who has actually updated it. Some people online speak about using a 256 bytes offset.

Thanks

 

[sohailoo]

should I use these for torrenting(random read/write)? I'm downloading and uploading 24/7 with 10-20 torrents active at the same time
my speed is 500/200, or should I stick with HDDs?

 

[EasyRhino]

should I use these for torrenting(random read/write)? I'm downloading and uploading 24/7 with 10-20 torrents active at the same time
my speed is 500/200, or should I stick with HDDs?

A generic piece of advice for torrenting drives is they are subject to a ton of random write activity.

If your current hard drives are not keeping up, then SSD upgrade would probably be great. If they ARE keeping up, then there may not be any urgency.

Also, a generic piece of advice for using a SSD for torrenting is to be careful about using up the write endurance. However, these guys are rated at 0.6 DWPD for 5 years.... and I don't think you're going to be writing 600GB every day.

 

[nexox]

Strictly speaking, torrents don't have to mean random writes - if the torrent client waits until an entire piece (usually 1MB, 4MB, or larger) is downloaded to memory and passes checksum before writing it to disk (a reasonable thing to do if you want to get decent download performance on spinning disks,) then that's pretty much a sequential write for an SSD and won't take as much life out of the NAND as a random write workload with 4k or 16k block size will.

But yeah even if the client is particularly awful and writes every 4k downloaded directly to disk with an fsync you would still need 6-12 months downloading at a constant 500Mbps to wear out one of these drives.

 

[UhClem]

There seem to be several firmwares on the wd zip file, how should I choose the correct one?

ZR11RPA1 for this ( ECO Gen II 1.9TB)

3. Update the firmware. Here is where I'd like input from someone who has actually updated it.

I used their CLI on Linux; straightforward, no problemo.

 

[A33]

I've tried the windows app as described, but failed miserably, again. Then succeeded with the linux version with a little trick:

./scli update /dev/sdb -p ZR09RPA1_w_boot_BLR-0080_QR.dob -m dnld-wo -l 1
./scli update /dev/sdb -m activate
./scli update /dev/sdb -f -p ZR11RPA1_w_boot_BLR-0080_QR.dob -m dnld-wo
./scli update /dev/sdb -m activate

I found that doing separate download + activate worked well if I've used the older firmware first. Using ZR11PA1 straight from ZR03PA didn't worked.

Updated 4 drives using ubuntu without problem with these commands. One drive was on ZR01, two on ZR04 and the last one on ZR07. All completed without errors using a sata to usb sabrent adapter. From Windows (using the same adapter) was impossible to get the drives detected.

 

[Fritz]

Updated 4 drives using ubuntu without problem with these commands. One drive was on ZR01, two on ZR04 and the last one on ZR07. All completed without errors using a sata to usb sabrent adapter. From Windows (using the same adapter) was impossible to get the drives detected.

Hooray for Linux.

 

[ribroc]

My lack of common sense aside, I want to mention some behavior here in case anyone else comes across this.
I was losing hair over this issue, I had 24 of these burning in, and I was getting frequent issues with the drives removing themselves at random at 1-2 drives an hour, requiring a physical reseat to operate normally again. This had no predictable cause, did not follow the bay, or drive, or even a predicable drive over time. I tried a different SAS backplane and connecting via SATA to no luck. It seemed like "bad batch" behavior.

I decided to finally spend some time and update the firmware on all of these up to ZR11 using the steps above, and I noticed that all the drives which had been troublesome were on ZR06 firmware. Since updating all to ZR11 the drives have worked predictably, and the data which had been on the drives was not affected by the firmware update.

sample dmesg

[ 8575.361441] program smartctl is using a deprecated SCSI ioctl, please convert it to SG_IO
[ 8575.362095] blk_update_request: I/O error, dev sdu, sector 0 op 0x1:(WRITE) flags 0x800 phys_seg 0 prio class 0
[ 8575.370608] sd 0:0:32:0: [sdu] Synchronize Cache(10) failed: Result: hostbyte=DID_NO_CONNECT driverbyte=DRIVER_OK

 

[Fritz]

Updated 4 drives using ubuntu without problem with these commands. One drive was on ZR01, two on ZR04 and the last one on ZR07. All completed without errors using a sata to usb sabrent adapter. From Windows (using the same adapter) was impossible to get the drives detected.

Could you go into a little more detail on how to update these in Linux? What program did you use? Those commands look a little wonky. Can I take them literally? Thanks.

 

[A33]

Yes, sure. I installed Ubuntu on an old laptop and connected the drives using a usb-c to sata adapter.

Then I downloaded the linux tool: http://downloads.sandisk.com/downloads/ess/utilities/scli-1-8-0-12-linux.zip and unziped it.

Open the terminal, cd into the folder of the unziped tool and then use the commands I wrote above. You just need to be careful that you are applying the firmware to the correct disk. If you have any doubt please ask and I'll try to help.

 

[Fritz]

Yes, sure. I installed Ubuntu on an old laptop and connected the drives using a usb-c to sata adapter.

Then I downloaded the linux tool: http://downloads.sandisk.com/downloads/ess/utilities/scli-1-8-0-12-linux.zip and unziped it.

Open the terminal, cd into the folder of the unziped tool and then use the commands I wrote above. You just need to be careful that you are applying the firmware to the correct disk. If you have any doubt please ask and I'll try to help.

Thank you sir.

 

[Jorge Perez]

This popped up making me think it was a good deal.

Anyway the thing with Sandforce controllers is that they wrote compressed data to the drive, which improved performance and extended the life of the NAND.

It wasnt an optimization for writing zeroes.

 

[BackupProphet]

We have tested these with uncompressible data. You still get 200-300MB/s and excellent latency. I am very happy with these.

 

[e97]

FYI - these seem to have increased in price - the lowest I see now on eBay is $60

 

[zack$]

Has anyone experienced an issue where if the temperature exceeds the threshold the drive reports as failing smart, even when rebooted and temperature is back to normal?

 

[T_Minus]

FYI - these seem to have increased in price - the lowest I see now on eBay is $60

Now 85-95 is lowest.. crazy times!

 

[T_Minus]

Has anyone experienced an issue where if the temperature exceeds the threshold the drive reports as failing smart, even when rebooted and temperature is back to normal?

No, curious if anyone else has.