Difference between older and newer hard drives

braigduck

Prominent
Jul 20, 2018
5
0
510
Hey!

So this has bothered me for awhile and seems to be pretty much swept under the rug in PC discussions.

When comparing hard drives of the same capacity and rotation speed (i.e. 500GB 7200RPM), what OTHER factors play a considerable role in the performance of a drive?

For an example: I have a Seagate Barracuda 7200RPM (16Mb cache, 3Gb/s) hard drive from 2007. It has just under 3,000 hours on it and reports good health. It has a large and clunky PCB and overall design.

I also have a Seagate Barracuda 7200RPM (16Mb Cache, 3GB/s) hard drive from 2014, and it has just under 4,000 hours and also reports good health. It is considerably more streamlined and has a flat and neat PCB.

Now, in comparing the two, they are both Seagate Barracudas, both at 7200RPM and both with 16Mb in cache. They have a comparable amount of usage, so wear is likely not a large factor between them. The only obvious difference is the 3GB/s vs 6Gb/s sata interface.

Despite all of this, the NEWER Seagate Barracuda is considerably faster than its older predecessor, EVEN WHEN PLUGGED INTO A 3GB/S SATA PORT, thereby bottle-necking it to the same speed cap as the older drive.

----------------------
My question is, what are the underlying factors that make the newer drive so much faster? Are these factors readily available on product pages when purchasing hard drives, or is it simply a guessing game in the end?

I.e. if I had the choice between a brand new 2018 WD Blue hard drive or a brand new 2018 Seagate Barracuda with what appear to be the same specs, how would I reasonably compare the two in terms of performance, since the RPM and cache size arent necessarily all there is to know?

----------------------

I realize this is long winded, but I deal with hard drives a lot, so this would be very good to know.
Thanks!
 
Solution
The main difference is areal density. With fewer platters holding the same data, more data flies by under the heads in the same amount of time, which makes an enormous difference in sequential transfer rate. This STR is the most important performance number on the data sheet that can be directly compared, and of course only tells you about performance with large files.

Note that unlike SSDs, no HDD yet reaches the limit of SATA 2.0 which is 3Gbit/s or about 300MB/s after overhead, so that cannot present a bottleneck. The very fastest 14TB drives are just now approaching 250MB/s.

The other major determinant of performance is the cacheing algorithm that determines what exactly is kept in the cache. Because HDDs are horrifically slow...
The main difference is areal density. With fewer platters holding the same data, more data flies by under the heads in the same amount of time, which makes an enormous difference in sequential transfer rate. This STR is the most important performance number on the data sheet that can be directly compared, and of course only tells you about performance with large files.

Note that unlike SSDs, no HDD yet reaches the limit of SATA 2.0 which is 3Gbit/s or about 300MB/s after overhead, so that cannot present a bottleneck. The very fastest 14TB drives are just now approaching 250MB/s.

The other major determinant of performance is the cacheing algorithm that determines what exactly is kept in the cache. Because HDDs are horrifically slow when accessing many small files due to both seek time and rotational latency (have to wait for the data to rotate around to where the heads are), anything in cache RAM can be fetched literally 10,000x faster. The onboard RAM is really small though (16MB) and optimizing for one intended use will hurt performance for other uses. The only way to determine which drive would be faster in your applications is as bmockeg suggested, to try to find reviews with benchmarks.

Note that Flash is only 1,000x faster but easily 1,000x more can be used--the WD Black SSHD comes with up to 24GB of cache Flash (24,000MB). And of course a SSD is entirely Flash so the whole drive is 1,000x faster.

The other numbers on the data sheet such as average seek time, and rpm (which tells you rotational latency) , only really tell you about uncached performance on small files and not as important as how well the cache works.
 
Solution