HDDs vs SSDs: It's all about the random speeds

-October 09, 2017

With my Mac mini's Thunderbolt port resurrected back to solid reliability, it was now time to similarly resurrect my search for faster mass storage. For those new to this particular post series, the computer (a mid-2011-era model) came standard with 500 GByte Toshiba 5400 RPM 2.5" HDD (a MK5065GSXF, to be precise). As long-time Apple aficionado and tech expert Glenn Fleishman also found with his system, mine was getting slower with the passage of time and O/S-and-app updates. And like Glenn, as I've mentioned before, I've long considered the seemingly complicated surgery needed to swap the HDD for a SSD.

My hope, as mentioned last time, was that I'd instead be able to clone the HDD image to the spare 1 TByte of space in an otherwise RAID 1-mirrored two-external-drive array connected to the Mac mini over Thunderbolt. The thinking here was that even though I'd then be running the mass storage subsystem over an external versus internal bus, that external bus would be 10 Gbps full duplex first-generation Thunderbolt instead of 3 Gbps (half duplex) SATA (the system supports third-generation SATA, but the HDD is only second-generation SATA). Plus, that spare 1 TByte of external storage space happened to be located in the 7200 RPM (and 64 MByte buffer-equipped) member of the two-drive Seagate GoFlex Desk for Mac array (a ST3000DM001 3 TByte 3.5" HDD, to be exact ... the other was a ST2000DL003 5900 RPM 2 TByte 3.5" HDD), hopefully leading to a further performance boost versus the seemingly pokey 5400 RPM (and 8 MByte buffer-equipped) 2.5" HDD inside the Mac mini.

So that's what I did, using Carbon Copy Cloner to duplicate the internal HDD image to the external HDD. I could have used SuperDuper! (which I've mentioned and leveraged in the past) to accomplish the clone, but as it turns out, CCC has an additional nice feature: the ability to clone not only the primary partition but also the hidden recovery partition on the source drive. Then I selected the external clone as my boot device, crossed my fingers, and started counting the seconds as the boot sequence progressed. I ended up both baffled and disappointed ... the system still took several minutes to fully boot, and application launches were still molasses-slow. A "bit" faster than before, perhaps ... but definitely not the dramatic increase I was hoping for.

My system is no slouch, based on an Intel "Sandy Bridge" i5-2415M, with dual physical cores plus HyperThreading support for two additional virtual cores, and running at a baseline clock frequency of up to 2.3 GHz (2.9 GHz in Turbo mode). And I've already upgraded the system memory from its default 2 GBytes to 8 GBytes; Activity Monitor therefore gives no indication of "memory pressure." But on a hunch, I bought a 1 TByte ADATA SU800 SSD on sale for $229.99 at Newegg, mounted it in a used FireWire 800 enclosure I'd picked up on Ebay, cloned the HDD image to the SSD, rebooted to the external drive, and about fell out of my chair when the system completed boot in a handful of seconds, versus the previous attempts' multi-minute boot times.

Keep in mind that FireWire 800's peak full-duplex transfer rate is, as the name implies, "only" 786.432 Mbit/s. Compare this to the half- and full-duplex rates for previously mentioned SATA 2 and Thunderbolt 1 interface alternatives, used to connect the Mac mini to HDDs. Clearly, the storage media is the fundamental performance determinant in this particular setup, not the interface. But my initial attempts to more precisely quantify the storage-and-interface candidates' comparative specs only further muddied the waters. Using Blackmagic Design's Disk Speed Test, here are the results I got when testing the internal HDD:

the Thunderbolt-tethered Seagate external HDD (specifically the non-mirrored partition running solely on the 7200 RPM hard drive):

and the FireWire 800-tethered external SSD:

For good measure, I also ran Disk Speed Test on my Thunderbolt-tethered Western Digital Thunderbolt Duo, containing dual 3 TByte WD "Green" HDDs configured in a RAID 1 mirror:

Based on these results, you'd expect either of the Thunderbolt-connected and HDD-based external storage devices to blow away the FireWire 800-connected SSD. But, at least from the standpoint of initial system boot and subsequent application launches, the converse is the case ... dramatically so, in fact. Why? After pondering a bit, I realized that Disk Speed Test, intended for video editing i.e. data files, measures sequential-access read and write performance, whereas system boot and application launch are heavily biased toward random access (mostly read, in fact) patterns. And, as I've pointed out in plenty of past SSD-vs-HDD writeups, SSDs dominate in storage scenarios when random access ... particularly read ... speed is the priority.

In retrospect, given the numerous past case studies I've published on systems that have gained new life thanks to a solid-state storage upgrade, I'm not sure why I didn't expect the same result this time. I guess, though, that the improvement is always so dramatic that I'm pleasantly surprised each time I experience it. And it's nice to know that cracking open the Mac mini and temporarily removing half its guts in order to replace the HDD with a SSD (hopefully reinstalling those same guts intact and fully functional afterwards) isn't my only option for gaining the improvement on this particular system. Stay tuned for a future post where I'll explore my options and explain which SSD-upgrade path I ended up selecting.

Also see:

Loading comments...

Write a Comment

To comment please Log In