Hi, I'm considering purchasing a 2 TB WD SN570, but I was able to find reviews only for the 1 TB version: if anyone has the larger one, could they confirm whether the sustained speeds and temperatures are the same as for the 1 TB one? Sean's reports a sustained speed of 600 MB/s post-cache and a peak temperature of 61 degrees (temperature at idle is not mentioned). Are these values the same for WD SN570 of 2 TB?
Question Feedback on WD SN570 of 2 TB: sustained speed and temperatures
- Thread starter GVM2014
- Start date
Don't know of anybody that's tested it. Maybe someone on my discord has it and would be willing to test it, or on Reddit (or even on here). The 2TB SKU came later and has higher sequential writes (+500 MB/s) and IOPS (+150K) so it's probably using 16x1Tb dies and possibly in the four-plane orientation of BiCS5. If so, it could be as fast as 1GB/s or so in TLC mode. It's bound to be at least as fast as the 1TB. The peak power listed is the same (4.5W) so it's probably the same number of dies (different density) and the same controller (1200 MT/s bus) although it's possible there's been optimization on it instead of faster flash.
If peak power is the same, does this mean peak temperature is likely to be the same?
I'm surprised though that nobody reviewed the 2 TB version of SN570, most other reviews received updates when newer versions came out.
Edited: speaking of which, SN570 (both 1 and 2 TB versions) seems to draw almost the same power than A2000 in all scenario. SN 570 draws 5 mW at sleep, 30 mW on low, 90 mW on average and 4.5 W maximum. A2000 draws 32 mW at idle, 80 mW on average, and 4.5 W maximum. Does this mean their thermals are going to be similar?
For comparison, SN750 (2 TB version) is listed in its datasheet at 5 mW at sleep, 100 mW on low. Average is not listed, and peak power is listed in amps, at 2.8A. Having in mind that NVMe voltage is around 3, this means SN750 draws 8 W at peak, which is almost double than A2000 (Tomshardware review indicates 1.2 W at idle and 5.6 W in a 50 GB transfer). No wonder it gets so hot (someone correct these figures if I got something wrong).
Last edited:
People have maybe reviewed the 2TB but none that show sustained write performance, AFAIK. It's possible there's some out there in Chinese or Russian but I didn't look too hard. I asked Sean and he would have to buy one to test this. I am extrapolating potential performance from the spec sheet.
Power draw does correlate with heat generation. Measuring average power draw is complicated as workloads vary, although you can measure with a trace that includes lots of idle time and bursty I/O (which is how it is in normal use). NVMe drives being used in a laptop will usually be idle taking little power but at higher performance states the difference can be wide, some drives peak at 11W or higher in fact. Most 4-chan DRAM-less will be in the 4-5W range although the SN740 (OEM SN770) goes up to 6-7W IIRC. 1.5A for 3.3V is <=5W which is quite reasonable.
Power draw does correlate with heat generation. Measuring average power draw is complicated as workloads vary, although you can measure with a trace that includes lots of idle time and bursty I/O (which is how it is in normal use). NVMe drives being used in a laptop will usually be idle taking little power but at higher performance states the difference can be wide, some drives peak at 11W or higher in fact. Most 4-chan DRAM-less will be in the 4-5W range although the SN740 (OEM SN770) goes up to 6-7W IIRC. 1.5A for 3.3V is <=5W which is quite reasonable.
SN750 went to 5.6 W in a 50 GB transfer in Sean's test, but its peak power is listed as around 8 W in its spec sheet (2.8A, more exactly).
Ok, since no one tested a 2 TB SN570, I think my results could present some interest: I bought this model and installed it in my laptop. Until now, thermals seem to be good: it stays at around 50 degrees at idle, which in a tight gaming laptop is fine. I performed 3 types of tests, in order to find out its max temperature, its cache and average sustained speeds:
- first test: I hammered it with aprox 400 GB of data from the laptop's other drive: it never throttled and wrote the whole package at speeds between 1000 and 1700 MB/s; the temperature got close to 75 degrees;
- second test: I hammered with the same amount of data from an USB 3.0, wrote the whole package at 400 MB/s, the max temperature was around 73-74 degrees (that is probably because it started from a higher idle temp, because I performed this test immediately after the first one - if I were to perform the same test starting with a cold SSD, I think it would be unlikely to surpass 70 degrees).
- since in my first test, I was not able to reach the cache limit with a 400 GB pack of files and I did not have a larger one on hand, I ran a benchmark with Aida64 linear write test: I started from an idle temp of around 48-50 degrees (having in mind what I was about to do, I let the system cool for several minutes before the last test) and I hammered it for almost 800 GB, then I stopped (I did not let it reach 100%, because, if I encountered no limitations after almost 800 GB written, there was no point stressing it further). I was still not able to run into any speed limitations, so I don't know how big the cache is (or whatever Western Digital has done to this drive). The maximum write speed during this test was 3046 MB/s, average speed 2902 MB/s, minimum speed 2234 MB/s. Obviously, the thermals during such a stress test were higher (when I stopped, CrystalDiskInfo indicated 86 degrees), but that was to be expected after such a demanding task - and, surprisingly, it never throttled. After I stopped the test, the temperature dropped at a steady rate over the next 5 minutes or so. Right now, at the moment I am writing this post, 20 minutes after the Aida64 test, SN570 has settled at 55 degrees.
Can someone tell me if Aida64 is a reliable benchmark tool, though? Because, I confess, when I saw the writing going past the 500 GB threshold and STILL not dropping, it looked awesome (700+ GB cache is KC3000 territory, at the very least) - granted, I also tested it the traditional way with 400 GB of data and still did not choke...
- first test: I hammered it with aprox 400 GB of data from the laptop's other drive: it never throttled and wrote the whole package at speeds between 1000 and 1700 MB/s; the temperature got close to 75 degrees;
- second test: I hammered with the same amount of data from an USB 3.0, wrote the whole package at 400 MB/s, the max temperature was around 73-74 degrees (that is probably because it started from a higher idle temp, because I performed this test immediately after the first one - if I were to perform the same test starting with a cold SSD, I think it would be unlikely to surpass 70 degrees).
- since in my first test, I was not able to reach the cache limit with a 400 GB pack of files and I did not have a larger one on hand, I ran a benchmark with Aida64 linear write test: I started from an idle temp of around 48-50 degrees (having in mind what I was about to do, I let the system cool for several minutes before the last test) and I hammered it for almost 800 GB, then I stopped (I did not let it reach 100%, because, if I encountered no limitations after almost 800 GB written, there was no point stressing it further). I was still not able to run into any speed limitations, so I don't know how big the cache is (or whatever Western Digital has done to this drive). The maximum write speed during this test was 3046 MB/s, average speed 2902 MB/s, minimum speed 2234 MB/s. Obviously, the thermals during such a stress test were higher (when I stopped, CrystalDiskInfo indicated 86 degrees), but that was to be expected after such a demanding task - and, surprisingly, it never throttled. After I stopped the test, the temperature dropped at a steady rate over the next 5 minutes or so. Right now, at the moment I am writing this post, 20 minutes after the Aida64 test, SN570 has settled at 55 degrees.
Can someone tell me if Aida64 is a reliable benchmark tool, though? Because, I confess, when I saw the writing going past the 500 GB threshold and STILL not dropping, it looked awesome (700+ GB cache is KC3000 territory, at the very least) - granted, I also tested it the traditional way with 400 GB of data and still did not choke...
The cache should be smallish on that drive, 24GB range IIRC. Many ways to check seq write performance, iometer, fio, hdtune (if set up correctly), dummy file creator, etc. Post-cache speed could be quite good on it though as mentioned above (based on assumptions).
Unless 2 TB SN570 has one of the highest post-cache speeds arounds, then the cache is much larger than that. Personally, I solved my issue and I am pleased with how SN570 performs, but, as a general observation, I think Tomshardware should seriously consider updating its previous SN570 with a section for the 2 TB version, because the 2 TB seems a completely different drive than the 1 TB version.
PS: I run a search on Google for disk benchmarking tools and the first which came out is Aida64 Extreme, which I used for testing SN570. Since I already have Aida on my computer, what is your opinion on its reliability? I've seen it used for linear writing tests (writing the whole storage space of the SSDs) by this guy: https://www.youtube.com/@prossd
Last edited:
The 2TB came out later. I have spoken with the person who reviewed that drive and he is curious about it. WD's older drives, all Gen3, have static SLC (which MUST be small), but if the 2TB SN570 is more like the SN770 it could have a different cache type. The SN770's is basically full-drive which means 1/3 the capacity, or ~670GB. This is only when the drive is completely empty and it will scale down with usage. Which is another area dummy file creator helps since you can fill the drive to say 75% then try your 400GB test again to exceed the cache (which at that point will be 150GB at most).
Sustained write testing actually has to be done carefully because it's easy to get the wrong results. IOmeter works and I have scripts for it, but NVM Express has their own ezIOmeter you can download to just run seq writes. FIO is also popular and can be run on windows (ioengine=windowsaio) without too much effort, I guess I could give a script for that as well. Although you will be able to test the way you have using the fuller-drive method I list above.
TRENDING THREADS
-
-
-
-
Discussion What's your favourite video game you've been playing?- Started by amdfangirl
- Replies: 4K
Facebook
Twitter
Instagram