Skip to content

Benchmarking Storage for The 2013 Mac Pro

2017/06/18

In early 2017, Apple admitted what everyone already knew, that the “innovative” late-2013 Mac Pro platform has some pretty severe limitations.  As some reviewers noted when it was first released, the internal bus design of the system had technical throughput limitations that caused Apple to design for only a single internal solid-state drive.  From Apple.com, your options were 256GB, 512GB and 1TB of space, and the upgraded capacities weren’t very affordable.  Being different can be expensive and being inside the Mac Pro was the most expensive place to live.

After studying my storage options for awhile, I opted for the entry-level 256GB SSD.  Going with the base level of storage allowed me to justify upgrades to the CPU and GPU, opting for the 8-core CPU and the twin-D-700 GPUs.  But knowing I had the 256GB limitation was good because it will force me to keep the clutter out of the drive reserved for “system” purposes.

I’ve always been a big fan of RAID as a storage mechanism whether it is the speed of RAID-0 Striping or the durability of RAID-5, it’s rare I feel comfortable using a system for anything I care about without RAID.  You can see my impressions of the OWC ThunderBay 4 unit in a previous post.  It’s a nice expansion chassis for Thunderbolt 2 systems.

So, back to the Mac Pro. When you realize how limiting the internal storage is on the Mac Pro you become painfully familiar with issues around connectivity.

The Mac Pro has a single USB-3 Controller with four (4) USB3 ports.  These are USB 3.0 ports, while the world has moved on to USB 3.1 Gen-2, a significant upgrade.  But, using USB 3 ports on a single controller means one port can impact the other, so the sustained throughput on the Mac Pro ports is combined across all ports.  You can see this by connecting two drives, each with their own cable, and watching throughput on each be cut in half. Oh, and you need to be sure the USB3 device was the first thing connected to that port, or it is further degraded to USB2 speeds, which cuts the speed to 480Mbps.  In summary, the USB3 story on the Mac Pro is not a good one.

The Mac Pro has three (3) Thunderbolt 2 Controllers.  This sounds better, right? They are configured with 2, 2 and 3 ports, the third including an HDMI port.  Again, with ports sharing a Controller you want to avoid having more than one big bandwidth consumer on a controller, so you might have a large storage array on one, a monitor on another and so on. As with the USB situation, this is Thunderbolt 2, which is now legacy with the release of Thunderbolt 3 to the marketplace. But, for most consumers, Thunderbolt 2 bandwidth should be fine. For those with larger storage arrays and multiple high-resolution displays, maybe not so great.

And, finally, knowing there are consumers who use Thunderbolt 2 laptops, there are Thunderbolt 2 “docks” that promise to allow you to detach a single cable and grab your laptop.  These can also be useful for those who want a second USB3 Controller to be available on their Mac Pro.  I grabbed a used Elgato unit for the purposes of this test to see if a second USB Controller was worth the clutter and expense.

So, this was about Benchmarking, right?  I decided to test some commonly-available consumer devices.  I paid for these myself and have never had an offer of equipment to write some positive reviews.  So, I don’t need flames and seldom, if ever, post comments.  I do read the comments, however, and do respond.

First, I wanted to establish some kind of USB3 performance baseline. What throughput would you expect to be ideal under the various USB3 options the platform provides, in this case the Mac Pro USB3 Controller versus the Elgato Thunderbolt 2 Dock USB3 Controller.  For this test, I selected the StarTech USB312SAT3CB, a USB 3.1 (10Gbps) SATA 3 adapter cable.  This is a simple, unpowered cable and is perfect for temporary use of your favorite 2.5″ SATA-SSD.  I decided to use the Samsung 850 Pro 256GB SSD because it sits at a performance-price sweet spot, even though it isn’t the fastest nor the cheapest.  It was good to see this affordable combination read 367MB/s WRITe and 361MB/s READ sustained throughput when connected directly to the Mac Pro ports.  This became the gold standard for USB3 testing on the Mac Pro.

I wanted to include a range of products you might find at reasonable prices.  I also conducted multi-day 3-pass certifications on the storage media to make certain it was “perfect” before I conducted any benchmarking tests on the media itself.  For this I used the SoftRAID.com Certify component of their SoftRAID product.  I used the free Blackmagic Disk Speed Test software to conduct the benchmarks.  This product produces a nice table of what kinds of video editing work the drive under test might effectively support, so even if you aren’t a video editor you can get an idea of relative performance against an admittedly-challenging real-world use.

I tested the following storage devices are part of these tests:

  1. Samsung 850 Pro 256GB SSD
  2. Seagate Constellation ES.3 7200RPM 3TB HDD
  3. Seagate Green 5900 RPM 2TB HDD
  4. Seagate Green 5400 RPM 2TB HDD

For my purposes, I really wanted to find an affordable two-drive enclosure that gave me the option of hardware-based RAID, or no RAID at all.  Because all the Thunderbolt 2 devices were priced at some multiple of the same cabinet with USB3.1, I decided to wait for Thunderbolt 2 prices to fall and leverage USB3.1 devices if possible.  Most analysts agree that USB3 is sufficient for 1-2 drive purposes.

I tested the following storage configurations as part of these tests:

  1. StarTech USB312SAT3CB, USB 3.1 (10Gbps) SATA 3 adapter
  2. StarTech SDOCKU313E USB 3.1 (10Gbps) & eSATA dock for SATA drives
  3. IcyDOCK IcyRAID MB662U32SR1 USB3.1 Enclosure.

Knowing the Mac Pro has a single USB3 Controller makes you wonder how limiting a factor that will be.  On a Mac Pro equipped with a USB2 keyboard, mouse and (Bluetooth) Magic Trackpad 2, I wondered if the USB 2 work or potential temporary connections would cause instability in performance.  I wondered if isolating my storage to an external dock (or my USB2 devices to that dock) would be a more predictable hardware architecture.

I tested the following connectivity as part of these test:

  1. Mac Pro USB3 Port
  2. Elgato Thunderbolt 2 Dock USB3 port

The purpose of these tests was to determine where the sweet-spot is in terms of sustained throughput and storage capacity.  After achieving disappointing results with other forms of RAID, this test was limited to RAID-0 Striping of two (2) storage units with the hope of achieving both high capacity and high sustained throughput at an affordable price.

I tested the following RAID combinations as part of these tests:

  1. No RAID, just a disk
  2. RAID-0, Hardware-based
  3. RAID-0, Software-based

The following are the results of the baseline SSD tests, withe an observation that for SSDs they are as fast as USB can go, so a RAID-0 SSD is a waste, and the Elgato dock performs consistently at 98-99% of the Mac Pro USB port on READ and 93% of the Mac Pro USB port on WRITE.  This seemed to hold true throughout the tests.:

TEST WRITE READ USB DOCK Bay RAID DRIVE

  1. 367 361 Mac Pro StarTech None None Samsung 850 Pro 256GB (1)
  2. 366 361 Mac Pro StarTech Dock None Samsung 850 Pro 256GB (1)
  3. 346 341 Mac Pro IcyRAID JBOD None Samsung 850 Pro 256GB (1)
  4. 346 341 Mac Pro IcyRAID RAID-0 HW Samsung 850 Pro 256GB (2)
  5. 342 325 Mac Pro IcyRAID RAID-0 SoftRAID Samsung 850 Pro 256GB (2)
  6. 335 354 Elgato StarTech None None Samsung 850 Pro 256GB (1)
  7. 336 354 Elgato StarTech Dock None Samsung 850 Pro 256GB (1)
  8. 320 336 Elgato IcyRAID JBOD None Samsung 850 Pro 256GB (1)
  9. 320 337 Elgato IcyRAID RAID-0 HW Samsung 850 Pro 256GB (2)
  10. 315 318 Elgato IcyRAID RAID-0 SoftRAID Samsung 850 Pro 256GB (2)

Because SSD technology is still relatively expensive for consumers when purchased in higher capacities (a 4TB OWC SSD is $2,000+ today), the hope was to find an affordable hard drive that could come close under USB3 platforms. I selected the Seagate ES.3 drive because I had very good previous experience with it, and it is now on the market for $80-85 for a 3TB unit, a fraction of its price just a couple of years ago.  These are “New, Surplus” stock, so they are probably soon to be unavailable.  But for me, it seemed like the ideal drive to test.  What we see here is that hard drive speeds, under RAID-0 Streaming, are 98-99% of an SSD on WRITE, with Software RAID READ speeds falling behind at 85% READ.  For a US$70 Enclosure, this feels pretty good.  A 6TD RAID-0 Enclosure for under US$250 seems like a pretty good deal yielding these speeds.  Here are the same tests, using the Seagate ES.3 7200RPM drive:

TEST WRITE READ USB DOCK Bay RAID DRIVE

  1. N/A N.A Mac Pro StarTech None None HDD not supported
  2. 175 180 Mac Pro StarTech Dock None Seagate HDD ES.3 7200 3TB (1)
  3. 175 180 Mac Pro IcyRAID JBOD None Seagate HDD ES.3 7200 3TB (1)
  4. 342 319 Mac Pro IcyRAID RAID-0 HW Seagate HDD ES.3 7200 3TB (2)
  5. 340 275 Mac Pro IcyRAID RAID-0 SoftRAID Seagate HDD ES.3 7200 3TB (2)
  6. N/A N/A Elgato StarTech None None HDD not supported
  7. 180 181 Elgato StarTech Dock None Seagate HDD ES.3 7200 3TB (1)
  8. 179 180 Elgato IcyRAID JBOD None Seagate HDD ES.3 7200 3TB (1)
  9. 315 315 Elgato IcyRAID RAID-0 HW Seagate HDD ES.3 7200 3TB (2)
  10. 311 265 Elgato IcyRAID RAID-0 SoftRAID Seagate HDD ES.3 7200 3TB (2)

For completeness, I will include the 5900RPM and 5400RPM drive test results, because these are cheap today and run cooler and quieter than non-Green counterparts.  I would consider these if the performance levels meet your needs, and you can put together a 4TB RAID-0 Enclosure for under US$130.

TEST WRITE READ USB DOCK Bay RAID DRIVE

  1. N/A N/A Mac Pro StarTech None None HDD not supported
  2. 135 140 Mac Pro StarTech Dock None Seagate HDD Green 5900 2TB (1)
  3. 140 140 Mac Pro IcyRAID JBOD None Seagate HDD Green 5900 2TB (1)
  4. 275 280 Mac Pro IcyRAID RAID-0 HW Seagate HDD Green 5900 2TB (2)
  5. 270 270 Mac Pro IcyRAID RAID-0 SoftRAID Seagate HDD Green 5900 2TB (2)
  6. N/A N/A Elgato StarTech None None HDD not supported
  7. 135 140 Elgato StarTech Dock None Seagate HDD Green 5900 2TB (1)
  8. 135 140 Elgato IcyRAID JBOD None Seagate HDD Green 5900 2TB (1)
  9. 265 280 Elgato IcyRAID RAID-0 HW Seagate HDD Green 5900 2TB (2)
  10. 268 268 Elgato IcyRAID RAID-0 SoftRAID Seagate HDD Green 5900 2TB (2)

TEST WRITE READ USB DOCK Bay RAID DRIVE

  1. N/A N/A Mac Pro StarTech None None HDD not supported
  2. 140 145 Mac Pro StarTech Dock None Seagate HDD Green 5400 2TB (1)
  3. 140 145 Mac Pro IcyRAID JBOD None Seagate HDD Green 5400 2TB (1)
  4. 255 260 Mac Pro IcyRAID RAID-0 HW Seagate HDD Green 5400 2TB (2)
  5. 198 216 Mac Pro IcyRAID RAID-0 SoftRAID Seagate HDD Green 5400 2TB (2)
  6. N/A N/A Elgato StarTech None None HDD not supported
  7. 140 145 Elgato StarTech Dock None Seagate HDD Green 5400 2TB (1)
  8. 140 145 Elgato IcyRAID JBOD None Seagate HDD Green 5400 2TB (1)
  9. 275 290 Elgato IcyRAID RAID-0 HW Seagate HDD Green 5400 2TB (2)
  10. 197 215 Elgato IcyRAID RAID-0 SoftRAID Seagate HDD Green 5400 2TB (2)

ADDITIONAL THOUGHTS

A couple of comments on product selected.  I didn’t use some other brands, which others might challenge.

I have never had good luck with Western Digital products and I could not get two drives with WD labels to pass the certification step.  This might be the luck of the draw, but my experience with their failure rate and long-disappointing support limited me to one attempt, buying a single drive from eBay, at which ppoint my other working WD drive failed. This entire project started becaue my WD MyBook Studio Edition II went belly-up, silently failing the Drive A bay on random occasions and reporting it as a drive failure.  In spite of what the box says about user-upgradability, you can only use the same (now obsolete) exact model drives that came with the unit, so these really do have an end-of-life.  As a result of my product and support experiences with WD, I don’t intend to go near their products in the future.

I considered the OWC Thunderbolt 2 dock because of excellent reviews, now a couplw of years old. The issue for me is I don’t really need the full dock, and the OWC unit is pricey considering the imminent demise of Thunderbolt 2 in favor of Thunderbolt 3.  If I was to invest in that unit, it would be at lower legacy technology pricing.

The IcyDOCK R1 was interesting because it was affordable and had an interesting data sheet. I found it on B&H Photo for US$104 before a US$35 rebate so I thought I’d give it a try.  After I finished the last benchmark and was going to reinsert the Seagate ES.3 drives and create my ultimate RAID-0 Enclosure, the door latch broke and the door wouldn’t latch.  The hinge in these units is so wobbly that it is easy to misalign the latch and it’s unclear exactly why it wouldn’t latch.  At this point, the fact that I filed for my rebate locked me in to getting a replacement unit, but I would caution anyone about two things here:  (1) don’t consider this a swappable dock unless you are really careful with the door hinges and lathes, and (2) never ever consider a rebate a done-deal.  I received an email saying my postmark was five days earlier than I mailed it and I haven’t heard from the rebate company since.

Finally, a word about SoftRAID, the OWC Holdings offering.  I used this to certify all my disks and the exercise really raised my confidence in the drives.  It is a tedious process, but for the user it’s the waiting that drives you crazy.  You could tell a lot about how a disk was going to perform by how long it took to certify.  If it’s going to take 72 hours to certify, you might not want to consider that as your RAID-0 drive.  And, while software-based RAID is competitive, if you really want that last MB/s of throughput, low-cost hardware-based RAID (without a software-license cost) still wins.

For me, I am now using the Seagate Constellation ES.3 drives in a 6TB HW-based RAID-0 Stripe that performs at close to SSD speeds for US$230, assuming I one day get my IcyDOCK rebate.

 

Advertisements

From → Uncategorized

Leave a Comment

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: