![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
I previously posted a set of tests like this on the Seagate ST2000DL003 four years ago. It was finally time to create a new RAID volume, and I decided on the Seagate ST4000VN000 for my next array.
In HD Tune Pro, there are six levels of test speed/accuracy: Partial test (five levels ranging from Fast to Accurate) and Full test. I've gone with Full test because it results in much cleaner data. It takes hours to do each test; with the eleven ST4000VN000 drives I tested, it took 7.5-7.9 hours to test each drive, reading every single sector on the drive. (Block size is also an option, ranging from 512 bytes to 8 MB; I've used a 1 MB block size.) This was a sacrifice, as it meant I had to delay my creation of a RAID array out of these drives by 5 days, but I feel it was worth it. (Testing them in parallel would have been great, but for some reason the Areca driver throttles simultaneous sequential transfers to the rate of the slowest transfer, even if they are on completely different drives/volumes — so I had to test them one by one.)
Here is an example of the difference between a Full test (with 1 MB block size), Partial test at Accurate (with 128 KB block size), Partial test midway between Accurate and Fast (with 128 KB block size), and Partial test at Fast (with 128 KB block size), in the form of an animated GIF:
(if I used a 1 MB block size in a Partial test, the transfer rate would come out significantly lower than that measured in the full test — in fact increasingly lower the less Accurate the test was.)
The data is so clean in a Full test that you can see the boundaries between zones. The ST4000VN000 appears to have about 30 zones. I was surprised to see that the outermost zone has lower density than the next zone in; this is the first drive I've seen whose zone template breaks the rule of zones always decreasing in density from outermost to innermost (albeit only on the first zone; from the second zone onward, it does decrease in density — although on one of the drives tested, there was a slight increase from the second zone to the third zone). Here is a 64 GB short stroke test on the same drive as the GIF above, demonstrating that the first zone really is a proper zone.
Here's an animated GIF showing the sequential transfer rate of each drive, sorted from slowest to fastest:
I've left out the access time data from the images above, but have included them here:
Full test data in screenshot form
Full test data in text format
(Access time ranged from 15.4 ms to 15.8 ms across the drives.)
All eleven drives are ST4000VN000-1H4168, and are connected through an Areca ARC-1880ix-24 dual-linked to a Chenbro CK23601 expander.
There are some little blips of lowered transfer rate in these tests. My apologies — this is because I was using my PC during some of the tests, and heavy access to other volumes in my system stole some priority from the benchmark. (This is also why the CPU usage is all over the map. It doesn't just measure HDTunePro.exe's CPU usage.)
In HD Tune Pro, there are six levels of test speed/accuracy: Partial test (five levels ranging from Fast to Accurate) and Full test. I've gone with Full test because it results in much cleaner data. It takes hours to do each test; with the eleven ST4000VN000 drives I tested, it took 7.5-7.9 hours to test each drive, reading every single sector on the drive. (Block size is also an option, ranging from 512 bytes to 8 MB; I've used a 1 MB block size.) This was a sacrifice, as it meant I had to delay my creation of a RAID array out of these drives by 5 days, but I feel it was worth it. (Testing them in parallel would have been great, but for some reason the Areca driver throttles simultaneous sequential transfers to the rate of the slowest transfer, even if they are on completely different drives/volumes — so I had to test them one by one.)
Here is an example of the difference between a Full test (with 1 MB block size), Partial test at Accurate (with 128 KB block size), Partial test midway between Accurate and Fast (with 128 KB block size), and Partial test at Fast (with 128 KB block size), in the form of an animated GIF:
(if I used a 1 MB block size in a Partial test, the transfer rate would come out significantly lower than that measured in the full test — in fact increasingly lower the less Accurate the test was.)
The data is so clean in a Full test that you can see the boundaries between zones. The ST4000VN000 appears to have about 30 zones. I was surprised to see that the outermost zone has lower density than the next zone in; this is the first drive I've seen whose zone template breaks the rule of zones always decreasing in density from outermost to innermost (albeit only on the first zone; from the second zone onward, it does decrease in density — although on one of the drives tested, there was a slight increase from the second zone to the third zone). Here is a 64 GB short stroke test on the same drive as the GIF above, demonstrating that the first zone really is a proper zone.
Here's an animated GIF showing the sequential transfer rate of each drive, sorted from slowest to fastest:
I've left out the access time data from the images above, but have included them here:
Full test data in screenshot form
Full test data in text format
(Access time ranged from 15.4 ms to 15.8 ms across the drives.)
All eleven drives are ST4000VN000-1H4168, and are connected through an Areca ARC-1880ix-24 dual-linked to a Chenbro CK23601 expander.
There are some little blips of lowered transfer rate in these tests. My apologies — this is because I was using my PC during some of the tests, and heavy access to other volumes in my system stole some priority from the benchmark. (This is also why the CPU usage is all over the map. It doesn't just measure HDTunePro.exe's CPU usage.)