Reviewer Tests $3 SATA SSD, Gets Exactly What They Paid For

An anonymous reader shares a report: StorageReview went through the remarkable journey of testing a $3 SSD from AliExpress. The Goldenfir-brand SSD was reportedly given to the storage site by one of its Discord users for testing. The good news is that Goldenfir is actually using an SSD controller for its NAND drive. The controller is a Yeestor YS9083XT, which the Chinese company announced as a SATA3.2 controller in 2019. [...] StorageReview tested the drive by putting it into a Lenovo SR635 1U server with an AMD Epyc 7742 processor and 512GB of DDR4-3200 RAM. StorageReview also decided to, admittedly "unfairly," put it up against Kingston's DC600M entry-level enterprise SATA drive. You can guess what happens next. With a 64GB file and the CrystalDiskMark benchmark, StorageReview reported that the "Kingston drive finished the entire test before this piece of turd [the $3 drive] could even build its test." With the VDBench workload benchmark filling up the entire drive, the $3 drive hit a wall at around 15,500 IOPS when running the 4K random read test, compared to the Kingston drive's approximately 80,000. The cheap SSD ultimately finished the test at 13,000 IOPS and 10,225 ms, compared to the Kingston's 78,000 IOPS and 1,630 ms.

wait is that bad?

[Chalex]

15,500 IOPS per $3 sounds pretty good to me!

Re:wait is that bad?

[Luckyo]

It's fucking amazing. I was expecting one of those scams where it comes with a 16 MB chip or something.

Re: wait is that bad?

[guruevi]

It likely is a scam like that, most reviewers donâ(TM)t have the technical chops to test that though. These synthetic benchmarks write all zeros linearly, at that point youâ(TM)re just testing the SSD controller, because all it does is ack the packets youâ(TM)re sending it.

In real torture testing, even a Samsung Pro NVMe doesnâ(TM)t get much above 20K IOPS (latencies average about 1mS once you get beyond the cache)

Re:

[Can'tNot]

Storagereview knows what they're doing, assuming they're the same people they once were. (I thought they had shut down.)

Re:

[guruevi]

Depends on the staff, this doesn't look like a serious article where tons of research went into it, more like ad-filler. Most of the articles after their shutdown have been just some CrystalBenchmark, I remember the days a serious review was 7+ pages of data. Now it's barely 5 tests and a conclusion.

Re:

[Midnight_Falcon]

Yeah I think they shouldn't have went straight for the IOPS to illustrate how bad the drive is... it has 50-120ms latency on read writes for example. On a hard drive (non-SSD), 10-20ms is acceptable. Also at some point the drive could only tolerate about 5.2mb/s write speeds, because it can't handle its own garbage collection and TRIM commands. So they rightfully conclude a crappy SD card, USB stick or basically anything else is better. A drive that reduces to 5.2mb/s after being full is about the sa

Re:

[gnasher719]

So you read something that's not in the article. Great. So where do you get this from? From what I read, this drive is f***ing amazing for $3. Only 6 or 7 times slower than a drive costing real money.

Re:

[Ksevio]

The bigger issue is it's probably a 16gb chip with a controller that's just re-writing the same memory over and over again

Re: wait is that bad?

[dwater]

"Probably"? Either it is, or it isn't. Anything else is you making stuff up.

Is it really so difficult for them to find out? They're only $3 ffs. Surely they've ripped one apart...

Re:

[Ksevio]

Given that it's a common scam for flash drives, it's quite likely, hence why I said "probably". I'm not affiliated with the reviewers so I don't know for sure.

It's pretty easy to tell if you just fill it with files and then read the ones written first.

Re:

[Midnight_Falcon]

It's in the article you just have to watch the video -- the article is just a summary of someone else's video.

Re:

[i.r.id10t]

Not sure if amazing or not... the price certainly is.

I'd be interested in how it does with longer term light use - ie, as storage for a remote camera taking pics on motion activation, maybe for mp3 storage after gutting an old radio and rebuilding internally with a Pi or similar small minimal machine to act as your player, maybe as a base-os storage for a thin client, etc.

But then... it is also kinda hard to beat a plain old USB flash drive type thing for a lot of that as well....

Re:

[NewtonsLaw]

Forget the quality... feel the width! :-)

Re:

[lsllll]

It's called "girth", and that's what SHE said.

Yeestor?

[jddj]

Time to yeet that thing.

You gotta be crazy.

[Major_Disorder]

I would never buy one of the cheap no name drives. Just not worth it. My data and my time is worth SO much more than I could ever save. Recognized brand name drives only for me. Kingston, Samsung, Crucial, WD. That is just about it.

Re:You gotta be crazy.

[serviscope_minor]

Same, but it's $3 for a real drive with the advertised capacity. That's pretty amazing

Re:You gotta be crazy.

[KiloByte]

And if we only could build something like a Redundant Array of Inexpensive Drives...

Re:

[Major_Disorder]

And if we only could build something like a Redundant Array of Inexpensive Drives...

Inexpensive, sure. But not this kind of cheap crap.

Re:

[Opportunist]

Why exactly? For that price, I can stuff 12 drives into a Raid6 with 24 hotspares running circles around it and still be cheaper than any competing solution without hotspares.

Re:

[Opportunist]

Recalculating that... this setup would be cheaper than the average Raid1 setup. Sans hotspares.

Re:

[Powercntrl]

Same, but it's $3 for a real drive with the advertised capacity. That's pretty amazing

Maybe you've missed it, but the drive in TFS is 120GB. So that's $0.025 per GB. Just checking Amazon, you can get 1TB NVME drives made by a brand name manufacturer for $0.04 per GB.

Considering how little extra you have to pay to get a lot more storage from something that isn't a steaming pile of crap, no, the $3 drive is not really an amazing deal.

Re:

[demonlapin]

It’s ridiculously cheap and it’s slow but usable. ATA/33 was a standard, you know. For some retrocomputing project, this could be great.

Re:

[slazzy]

Nor would I normally. But it's great when the lower priced drives come out because it will force the big names to put on some good sales, and start working hard to develop larger capacity drives and newer faster technology to compete.

Re:

[SuperDre]

Well my experience with brand name drives is that they also fail. I got an Intel, Samsung, Crucial, WD and Seagate fail on me. But ofcourse I expect a $3 nobrand chinese SSD to fail much MUCH faster, but it still can have its uses.

I realize they are just talking about performance

[93 Escort Wagon]

But, given fairly recent and repeated history, did they also carefully check to make sure the drive (or its cable - I imagine it included a cable) isn't doing something untoward with their system and/or their data?

Re:

[nsbfikwjuunkifjqhm]

So you're suggesting that this product may actually be an $80 rubber ducky, or a $180 O.MG cable sold for $3, and comes with a legit bonus 120GB SSD? Where the fuck do I sign up? Security researchers and hackers the world over would put these guys out of business so fast they wouldn't be able to pay for the CnC servers you're proposing they're using to steal Minecraft logins from little Johnny's PC.

The real news here is ...

[davidwr]

... that they didn't lie about its size.

Re:

[youn]

you know what they say... it;s not just about size, in the end it's really how it's used that really matters :p

the big difference if you finish quickly, it is really good :)

Reads like a hit piece

[WaffleMonster]

This is the strangest review of an SSD I've ever seen. They focus entirely on IO latency and not read and write bandwidth, endurance or real world benchmarking. They show a picture of the PCB cut in half for no reason and their descriptions are full of loaded language and unsubstantiated opinion.

Re:

[Opportunist]

Sounds like Kingston has a new USB-attached backup drive they need to sell.

Re:Reads like a hit piece

[thegarbz]

They show a picture of the PCB cut in half for no reason

The PCB isn't cut in half. That's all there is to it. Most 2.5" SSDs are mostly empty inside with small PCBs and usually air filling the rest of the form factor. If you're talking about the video thumbnail, it's in pieces because they snapped the SSD at the end of the video after saying don't buy this, we're not even going to give it away for free.

They focus entirely on IO latency and not read and write bandwidth

The review contained read write bandwidth. They focus on IO and latency because they identified massive issues occurred when IOPS reached a very low number. In actual sequential read/write it performed the same as the reference SSD maxing out the SATA bus, but only at insanely low IOPS. Read just the bandwidth and you'd think you have something that performs well but in the real world it would perform significantly worse than even spinning rust HDDs.

are full of loaded language and unsubstantiated opinion.

If you watched the video you'd realise nothing about the language is unsubstantiated. It's a piece of shit that no one should ever spend their money on for any reason what so ever.

The problem is what exactly?

[Opportunist]

That it's slow as molasses?

The question is, is it actually the size it's supposed to be, will it retain its data for longer than the memory span of a goldfish and will it survive a few write cycles? Because if the answer to all of these is yes, then it's an amazing backup drive.

I mean, who attaches backup drives with anything but molasses-slow USB connectors anyway? Twice so if you're not a geek and don't even know which connectors are USB 2 and which are 3 on your laptop.

Re:

[evil_aaronm]

Is "memory span of a goldfish" an ISO standard, yet?

Re:

[hawk]

We're not sure, as the discussions are transmitted over an RFC 1149. connection.

I *told* them to use 2549, but, no-o-o, they wouldn't listen, so . . .

Re:

[Opportunist]

It's in the same suggestion as libraries of congress as a unit for data amount and Olympic swimming pools for liquid volume.

Re: The problem is what exactly?

[zmollusc]

Laptops don't have connectors, grandad.

Re:

[Opportunist]

Real ones do.

Fashion accessories disguised as laptops don't count.

Due to latency to China

[skaag]

What do you expect, it takes time to transfer all of that data to those secret servers in China! and it's probably done in a synchronous manner

Does it Store Data?

[Thelasko]

I've heard stories of fake flash drives [youtu.be] overwriting data. Are they sure this thing is actually writing data to the disk?

Re:

[Ed Tice]

It doesn't seem that they even checked.

Install ISOs

[John Allsup]

I don't have an SSD this cheap, but I have some supercheap brand ones. Something these supercheap SSDs are useful for is as an alternative to memory sticks for installing an OS. They are much quicker to write to (many times quicker), and quicker to read from, and can be rewritten more times. So this speeds the process. On the other hand, if it stops working, chuck it in the bin and grab another one.

I've not tried these, but at this price I may grab a few and see.

Special Offers

[John Allsup]

This is a Welcome Offer, so you can only apply it to one item, and if the order is less than $8, you have to pay shipping.
Given that you can get proper brand (e.g. WD Green) second hand 120GB SSDs off eBay for around $10-$15, I think I'd go for those instead.

sounds perfect for a small project.

[PinkyGigglebrain]

A small SATA III 120GB SSD for $3. Sounds perfect for several Raspberry Pi projects I can think off off the top.

If you want the best performance get the Enterprise level hardware. If you just want a big block of storage for some small project that and you won't care if it gets stolen/lost a $3 SSD might fit the bill perfectly.

Too bad the article doesn't have a link to where I can get a few of these, now that I think about it it might be useful to have a few handy.

Re:

[DarkOx]

I supposes there is exists a unique set of circumstances where there is a perfect application for just about anything.

However as others have pointed out:
1) on a $/unit of storage basis, this is actually as cheap as it seems. You can probably get 1TB of SSD (3d nand anyway ) from a brand you have heard of for $40. Do the math.

2) That $40 SSD will be much faster.

3) If you really are hell bent on optimizing for cost and your application isn't going to consume more than 120 gigs of storage, its probably also

What, it _worked_?

[gweihir]

That is a surprise. Ordinarily these things just pretend to be storage and simply throw away most or all of your data.

Getting a working 120GB SATA SSD for $3 is a pretty good deal. Well, question is how long it will work.

Nobody will steal all your data.

[stooo]

>> pretend to be storage and simply throw away most or all of your data.
Efficient computer security against intrusion, right there.
Good.
Nobody will steal all your data.

Re:

[gweihir]

>> pretend to be storage and simply throw away most or all of your data.
Efficient computer security against intrusion, right there.
Good.
Nobody will steal all your data.

Including yourself ;-)