Anyone Have Experience with Cheap SSDs?

Banks on 27/06/2025 - 21:27
Last edited 15/07/2025 - 09:55

I was gifted a Orico Y20 for my birthday which my colleague got off Aliexpress.

I saw the reviews and the chip varies from batch to batch. I could run the popular Vadim Ochkin ssd tool, but I need to find time to do a backup and then run this tool before restoring my backup again. (I am not sure if I can trust it since it's not open source.)

I ran some preliminary tests on this SSD and it is relatively fast until the SLC cache is depleted at around 40% of the drive and then the speeds drop to around 70MB/sec, and I can say that this Orico seems to perform way better than a Crucial BX500 I bought after my latest hacking incident which seems to neutered deliberately in that the SLC cache is non-existent. I can copy a 40 GB windows VM without it dropping speeds on the Orico 256GB, but around halfway through the BX500 transfer it just drops to 40MB/sec. I did a bit of searching and found the SLC cache on the BX500 is around 12GB on the 240GB model.

The fact the speeds dropped near 40% suggests it might be a TLC nand, but I can't be sure without running the Vadim Ochkin tool. But no surprises it would be better than a BX500 which is QLC.

Nevertheless, I always wondered why the DRAM versions of Crucial ssds are faster, especially when you take a few seconds to think about the 1 GB DRAM cache is filled almost immediately which suggests that is not the real reason for the slowdown. In fact as far as I am aware the DRAM is only useful to minimise the impact of the OS writing in the background which I estimated is only really around 20 GB every 24 hours which is only a loss of 7 TBW every year which is trivial. The true reason for the slowdown is a deliberate neutering of the products which I have put down to deliberate programming of the controller to not use up to 70% free space for SLC caching.

I saw others recommending ACOS, Fanxiang, Kingspec, Gudga, Cusu, Billion Reservoir, Goldenfir, Fikwot. Anyone have any experience with those?

Apparently my colleague already set the partition to overprovision ~10% of the drive. I'm currently using it to run a virtual machine and open any incoming email attachments. Genius advice from my colleague I have to say. I now no longer have to worry too much about viruses as I can restore snapshots every few days.

Anyone have any other hints on what I should be testing to play around with this? I was thinking of running a AIDA64 benchmark but I already have estimates regarding where it starts to drop in speeds, i.e. SLC cache size which is around 30% of the drive size, although it could be a dynamic cache size. I did run a h2wtest which was successful but then my colleague told me afterwards it was already overprovisioned so that test might not have been very accurate.

I'm not too concerned about losing data and have an extra copy of the VM on my main SSD which I have been told to keep updated on a weekly basis before copying it over. I have done quite some writes and I'll probably keep using this SSD for a while. Hopefully it doesn't die, but I'll probably post some updates across my first year of using it.

Crystaldiskinfo I'm writing around 40GB per day to the drive, so I estimate it will last me around 5 years. fingers crossed

How has your experience been with these not-so-common SSD brands? Would you recommend them or am I going to be upset with this product in the first year? Feel free to comment or bet against yourselves on how long this SSD will last. haha.

Update as of 25/07/2025

Day 1? - Was in a PNG file on the drive

Day 21 - Crystaldiskinfo

I have a few concerns about how fast the number of valid blocks is dropping. From what I'm forecasting it will run to zero after a year. After this I am not sure if it goes into read only mode or if the overprovisioning means it will start taking some blocks from that overprovisioned area. Does anyone know?

Computing Computer Component SSD

Comments

  • Far Cough on 27/06/2025 - 21:45

    Sounds like you’ve set up that Orico Y20 pretty smartly, especially for a VM sandbox. Your findings on SLC cache behaviour line up with what I’ve seen in some TLC drives. Definitely agree: the BX500 feels throttled on purpose. As for those off-brand SSDs, they’re hit-or-miss; some decent ones are out there, but consistency isn’t their strong suit. Would love to hear how your setup holds up over time. I’m betting it gives you 2–3 solid years unless the controller throws a tantrum.

  • 8bitsperbyte on 28/06/2025 - 15:25
    +1

    Nevertheless, I always wondered why the DRAM versions of Crucial ssds are faster, especially when you take a few seconds to think about the 1 GB DRAM cache is filled almost immediately which suggests that is not the real reason for the slowdown.

    The DRAM on SSDs is not actually used as a cache for file data. Instead, it is used to hold a data map table that keeps track of where files are actually stored in the NAND memory.

    Basically all file operations on the SSD need to access this data map table to know where to read/write/move the data to. As a result, this data map table is accessed very frequently. Storing this table on faster and more responsive memory, like DRAM, would significantly boost the SSD's performance. 1 GB DRAM should be plentiful for this purpose.

    For DRAM'less SATA SSDs, there is no DRAM to store this data map table, so this table is stored and accessed in the much slower NAND memory, which makes the SSD feel sluggish/unresponsive.

    For DRAM'less NVMe SSDs, most use Host Memory Buffer (HMB), which avoids the issue of accessing the data map table from NAND. Instead, the table is copied to the computer's system RAM and accessed from there instead.

  • brokenglish on 28/06/2025 - 15:50

    Okay with cheap SSDs as long as it’s not my main rig. It’s a hit and miss.

  • Banks on 15/07/2025 - 09:57

    Replying to the comments seems to be the only way to bump this thread back up so people can read it again.

    I thought I would update on the crystaldiskinfo data, since it's been around 3 weeks. Let me know when is the next appropriate timeframe when I should update, otherwise I might make this a monthly task.

    Also, I have some concerns about the drive now…

  • Banks on 12/08/2025 - 10:42
    +1

    I'm not sure why the thread was locked for marketing speak, but yeah, it's dead.

    I am not surprised because ironbastion in a deal stated theirs died after a week.

    ironbastion's comments

    • 8bitsperbyte on 13/08/2025 - 19:18
      +1

      Thanks for the update. I hope you didn't lose any important data.

      Do you have any final statistics of the drive (day #, power on count, hours, total data written, etc.)? Another screenshot of the final CrystalDiskInfo state would be nice.

      If you're not planning on returning it, it might be interesting to open up the SSD to see what internals it was using, and share pictures of it.

      • Banks on 15/08/2025 - 11:24
        +2

        The SSD died catastrophically, but I sent it back to my colleague who might know how to deal with it. There wasn't anything important on it as the VM is encrypted but I think the way the these particular ssds do the caching to speed it up (p-slc cache?) and the snapshots I was using was putting too many daily writes onto the drive.

        Write amplification?

Login or Join to leave a comment
Login Join