> The hard drive shows better write throughput and latency as compared to most of the tested SSDs. Yes, except for the initial few minutes where the cheap SSDs tend to be faster (except for the Kingston & Crucial SSDs) but how much does that matter?
I love the idea of this article, but it lost me here.
“Except for the initial few minutes” is a weird thing to dismiss, since the majority of desktop operations will be done in less than a few minutes. Most are in the span of seconds.
The only time someone will typically go past a few minutes of sustained writes is during very large file copy operations. It’s weird to put an emphasis on this relative edge case while downplaying the importance of burst performance.
Anyone who has switched from even a fast hard drive to a cheap SSD can see the difference an SSD makes. It’s true that my NAS can sustain higher throughput for longer than a cheap SSD, but it’s much slower at doing directory listings and scanning random files than even the cheap SSDs.
Although cheap SSDs do perform fine regarding reads, sustained write performance can be really atrocious.
Notice the sustained write performance.
So you consider copying large files a fringe / or edge case, but personally, I'm not so sure. Especially for people who buy cheap 1TB or 2TB+ SSDs, they may be unpleasantly surprised when they copy some media files or start downloading a game from steam.
Don't forget that it will take quite a while - due to the slow flash - to empy the SLC cache before you can benefit from it again.
> Anyone who has switched from even a fast hard drive to a cheap SSD can see the difference an SSD makes.
This article is not about the merrits of SSDs, that's a done deal, a good SSD beats an HDD by every conceivable metric.
My point is that people should watch out for cheap SSDs, as their sustained write performance is so slow.
> So you consider copying large files a fringe / or edge case,
No, I consider sustained writing at maximum speed for more than several minutes to be an edge case.
Consider the 240GB ADATA in the article. It can write over 400MB/sec for over 150 seconds before throttling kicks in. That's 1/4 of the entire drive.
The Crucial drive has no problem doing at least 100MB/sec for the entire capacity of the drive, if that's your thing.
> but personally, I'm not so sure. Especially for people who buy cheap 1TB or 2TB+ SSDs, they may be unpleasantly surprised when they copy some media files
You can write 40GB to 3 of these drives and never even throttle. That's an entire Blu Ray, and it would occupy 1/3 of the 120GB drives tested.
If someone has a use case that involves writing 50% of the drive at full speed over and over again, a cheap SSD is not the ideal tool. But that's really an edge case for a budget 120GB drive.
> or start downloading a game from steam.
The slowest SSDs in the article can consume the entire bandwidth of a Gigabit internet connection until the drive is full.
The ADATA has dramatic throttling, but it only kicks in after 60GB written at full speed.
You're not going to encounter these throttling scenarios under normal operations. If you're only getting a drive for 100GB sequential transfers over and over at the highest possible speed, get something else. But then again, you're probably not looking at $20 120GB SSDs anyway.
> The Crucial drive has no problem doing at least 100MB/sec for the entire capacity of the drive, if that's your thing.
That's quite slow to be frank, as stated slower than an HDD and that's quite disappointing for an unsuspecting consumer who does want to transfer larger files.
P.S. see the footnote where after 1 hour or about 300 GB the transfer speed start to collapse entirely.
> Consider the 240GB ADATA in the article. It can write over 400MB/sec for over 150 seconds before throttling kicks in. That's 1/4 of the entire drive.
> No, I consider sustained writing at maximum speed for more than several minutes to be an edge case.
It's not specifically about the small 120/240 GB SSDs, the 1TB Crucial shows that larger drives exhibit the same problem. If your SSD is 1TB or 2TB, that 50GB transfer doesn't feel so enormous anymore as compared to drive size.
I'm going to agree that most people won't have a use case for transferring large files. I've actually repurposed the Kingston as an OS drive for my lab server.
But the key objective of my blog post is to show that cheap SSDs exhibit this behaviour in the first place. Many people are not aware / don't know. They can still decide that it's no problem for them. But for some, it will be an issue.
This is not about the smal <$20 SSDs, but about the concept of cheap SSDS often having terrible sustained write speeds, regardless of capacity. And many review sites don't highlight this issue or actually show when the throttling kicks in.
> Especially for people who buy cheap 1TB or 2TB+ SSDs, they may be unpleasantly surprised when they copy some media files or start downloading a game from steam.
That would be an entirely different experiment. This one has highlighted the most important pitfall of using critically undersized SSDs, so its conclusions say next to nothing about multi-TB drives.
The Crucial MX500 is a rather old piece of hardware now.
Tom's Hardware NVMe benchmarks include a "Sustained Write Performance and Cache Recovery" component. Whenever there's a good sale price on an NVMe, that is just about the only metric I hunt down now. Most of the worst drives they test will always beat a mechanical disk, but the worst drives Tom's Hardware ever tests are still decent drives.
I grabbed one of the cheapest SATA SSDs last week to replacing a failing lvmcache drive. It is a NETAC 1 TB that might still be on sale on eBay for $34. I expected the worst, and I did want to test its sustained write performance, but I wasn't as nearly scientific as you!
I just ran dd for a while and watched it stay between 420 and 470 megabytes per second for about 120 gigabytes straight before I stopped the test. The meanest I am to this cache is dropping 50 GB of video on two different days each month, so that was all the data I needed.
Had I known that I would be reading your blog four days later I would have let the dd finish so I could take better notes! Thank you for taking the time to do the science for us!
> The Crucial MX500 is a rather old piece of hardware now.
Depends on when you bought it. Crucial/Micron decided to stop introducing new branding when they updated their SATA SSDs, but the hardware inside has changed several times to incorporate new generations of NAND flash memory, and probably at least one update to the SSD controller by now. None of that matters to the top-line specifications they advertise, but such changes can be relevant for more stressful, more thorough or less realistic benchmarks.
> Especially for people who buy cheap 1TB or 2TB+ SSDs, they may be unpleasantly surprised when they copy some media files or start downloading a game from steam.
Especially seen that the one who has the means to pay for fiber to the home able to sustain 500 Mbit/s+ download (at less than that even cheap SSDs shall sustain the write speed anyway right?) and has the means to buy games from Steam probably can afford to spend 50 EUR on a fast SSD?
TFA mentions 137 EUR Samsung SSD from 2019 but prices have dropped since then. And nowadays all mobos ship with NVMe M.2 PCIe slots and you find stuff like that for 44 EUR: Sabrent M.2 NVMe SSD 256GB Interne Solid State 3400 MB/s read, PCIe 3.0 X4 2280 or for 80 EUR: Samsung 970 EVO Plus MZ-V7S1T0BW (that was a 200 EUR+ drive two years ago I think).
Would have been interesting to compare vs those beasts, which are also cheap.
My point is: if you've got a setup allowing you to max the write speed of a cheap SSD, you've got the 30 or 40 additional EUR to buy an ultra fast beast.
P.S: I don't mind paying a bit more so as adviced here, for my new build I bought a Western Digital SN850X Black.
For tiny SSDs, it's also important to consider how the volume of data written compares to the total capacity of the drive. Having the horizontal axis of the graph be time kinda implies the test could keep going indefinitely. But if the data were recast in terms of % of the drive, it would be easier to see how quickly it runs beyond any realistic use case: re-writing more than half the drive in one operation is simply not how people use their drives except on rare occasions (eg. restoring from backup).
Copying big file is not a fringe case. I often find myself copying files that take more than a minute because I work with videos that I took on my phone and camera and for video editing. I also have lots of data, which need to be transferred frequently and it is frustrating when the SSD cache is exhausted during copy and as a result, it becomes even slower than my old spinning hard drive (HDD).
Copying big files onto a 120GB SSD is a fringe use case, unless it's an external SSD you're using to transport those files. Editing and organizing large videos is usually done on drives large enough to hold more than two of them (plus the OS and applications).
I love the idea of this article, but it lost me here.
“Except for the initial few minutes” is a weird thing to dismiss, since the majority of desktop operations will be done in less than a few minutes. Most are in the span of seconds.
The only time someone will typically go past a few minutes of sustained writes is during very large file copy operations. It’s weird to put an emphasis on this relative edge case while downplaying the importance of burst performance.
Anyone who has switched from even a fast hard drive to a cheap SSD can see the difference an SSD makes. It’s true that my NAS can sustain higher throughput for longer than a cheap SSD, but it’s much slower at doing directory listings and scanning random files than even the cheap SSDs.