Installing a SSD in the optical bay?

All SATA ports are SATA2 / 3G but the good thing is that optical port is on a different controller. The 4 drives are all sharing a (limited) 800MB/s and not even the "300MB/s for each, independently operated". There should be 6 ports on 1.2GB chip.

Some go for PCIe SSD or controller with mixed as not everyone likes or had good results from OWC product.

CCC or even a clean install.

You should be fine. There is an SATA power and data cable to use.

Startup Manager: How to select a startup volume

Hello, sorry to jump in.

I too have just bought a 512 SSD for my 2012 mac pro. I've seen varying accounts of the pros/cons of using the optical bay.

Does it still carry the same 3gb/s speed as the hard drive bus? So there is no reason why not to put my SSD there and just leave my HD where it is in bay 1?

thanks

mark

The hatter wrote:

All SATA ports are SATA2 / 3G but the good thing is that optical port is on a different controller. The 4 drives are all sharing a (limited) 800MB/s and not even the "300MB/s for each, independently operated".

Is that actually correct? When I look at the current "Tech Specs" for a Mac Pro (late 2012 version), it states:

Storage

Four 3.5-inch cable-free, direct-attach drive bays with built-in independent 3Gb/s Serial ATA channels; four internal drive carriers included

That suggests to me that each drive has its own 3Gb/s SATA connection to the CPU. Or am I missing something?

Well there sure ain't 300MB/sec x 4, there is a single controller and more like 800MB of bandwidth. In fact 2.5 SSDs of the older variety would max it out - see Barefeats from say 2010 - SATA3 controllers / SSD 6G were more like mid-2011 or later by the time bugs worked out.

the optical SATA cables get around some of that.

Don't believe me or the reports. And DO take most of the "specs" with the proverbial 'grain of salt.'

www.macperformanceguide.com has done a bit of testing too, along with www.barefeats.com

Well I'm now able to speak from experience on this topic. I've got my new MacPro (the 12 core late 2012 model) with maximum RAM and an 8 TB striped OSX software RAID 0 (four 2TB, 7200rpm, 6Gb/s drives). Booting Mountain Lion from an OWC Mercury Electra 3G SSD installed in the unused optical bay. I noted while installing the internal drives that they connected directly to fittings soldered on the motherboard. Whether they were all in "independent 3Gb/s Serial ATA channels", as promised by Apple's specifications, is something I could not visually determine.

From the authoritative Wikipedia:

SATA revision 2.0 (SATA 3 Gbit/s)

Second generation SATA interfaces run with a native transfer rate of 3.0 Gbit/s, and taking 8b/10b encoding into account, the maximum uncoded transfer rate is 2.4 Gbit/s (300 MB/s).

==end quote==

Using a 27.76 GB data file, I performed the following: (a) a simple duplication (using command-D) of the file on the SSD; (b) a simple duplication of the file on the RAID; and (c) copying the file from the RAID to the SSD and back (by dragging the file). Here are my findings.

(a) Dupe SSD to SSD: 225 seconds. Since the file is both read and written, this is 0.247 GB/s (247 MB/s) =1.97 Gb/s. This is 82% of the anticipated SATA maximum, and 88% of the "peak read data rate" given by OWC, of 280 MB/s.

(b) Dupe RAID to RAID: 119 secxonds. Since the file is both read and written, this is 0.467 GB/s (467 MB/s) = 3.73 Gb/s. This is 156% of the SATA maximum, which suggests that the drives are in fact able to transfer at a rate modestly over the single-drive maximum. However, it is nowhere near the 9.6 GB/s that is seemingly promised by Apple.

(c) Copy RAID-to-SATA (and back): 215 seconds, which is 0.258 GB/s (258 MB/s) =2.07 Gb/s. This is 86% of the SATA maximum, and 92% of the "peak read data rate" given by OWC. It would appear that the overall transfer is being limited by the SSD, although the overall thruput is marginally higher than the SSD-to-SSD dupe.

I have no idea where you (The hatter) come up with the facts to say "The 4 drives are all sharing a (limited) 800MB/s", but it certainly does seem to be supported by my experiment. The RAID-RAID dupe I performed was well within that limit.

It may be that during formatting of the array I selected non-optimal parameters for this sort of a test, but I am a beginner in this, and I don't have other diagnostic software yet. Since most of my work will involve reading and writing files of this size (10-40 GB), it would behoove me to seek advice from you (or others) as to how to improve the performance of this RAID array. Is there anything, short of buying a PCIE RAID card and external drives, that i can do to speed things up to approach that 800MB/s?

Thanks in advance for any suggestions here. I realize I've drifted somewhat off-topic for this thread, so if a moderator moves it I won't be surprised; however, if the OP (Dickie7) happens to see it, s/he may find the results of my SSD in the optical bay to be useful.

Barefeats came up or out with their ability to test 2, 3, and 4 SSDs and first time that we had drives that could push the SATA bus bandwidth - and found that 2.5 old SATA2 SSDs were it. SoftRAID suggested 135MB/sec which meant shaving off 25MB/sec which did seem like a lot. But that was life then too.

He also did some other tests. And on Apple Pro RAID fared worse.

MacGurus and others talked about how even in ideal our old SCSI Ultra160 was really only achieving 85% utilitzation and there is a lot of channel overhead. On SATA2 even a pure 300MB/sec if there was one (there isn't) comes out with 275MB/sec at best and sometimes a little less as well.

In a perfect world there would have been a 1.2GB/sec controller. Or 1GB/s and you can find motherboards but they really even with 6 x SATA ports aren't "independent" and share some Intel or marvel or other controller... and share the goodies!

Take pinch of salt over the right shoulder. There are schools that teach professional technical writing and seems that marketing, not technical, often overrides too many specifications. Don't want to school people or throw around EFI-32 or anything afterall.

RE: copying RAID-to-RAID

RAID is not inherently fast. It can be fast-ER when it can stay focused on large transfers of a single file. This is because RAID gets its increased speed by overlapping seek time for the second drive with transfer time for the data from the first drive.

Anything you do that interrupts this "best case" scenario reduces RAID to the speed of a single drive, such as:

• Reading files that are small (too small to sprawl across multiple tracks, for example).

• Reading/Writing from more than one file at a time (this moves the heads away from the target file, so no overlap, no speed gain).

• Allowing Mac OS X on the same drive (Mac OS X is always going after another part of an Application, writing a log entry, reading a preferences file, and your RAID performance improvement vanishes when it moves the drive heads).

There are schools that teach professional technical writing and seems that marketing, not technical, often overrides too many specifications. Don't want to school people or throw around EFI-32 or anything afterall.

Thanks for your reply, The hatter. Your comment about marketing is so right. At what point do we point to a marketing promise and cry "false advertising!"? Perhaps only when there actually is something better available that we might have bought instead? I don't understand the EFI-32 comment, but suspect it's some kind of inside joke. 🙂

From what I'm reading here, it sounds like you're saying I shouldn't expect to be able to read/write to a my internal drive array at more than I've already achieved with this RAID 0 arrangement, 3.73 Gb/s, because of the limited bandwidth on the SATA bus they are all connected to. I'm disappointed, but that's life.

Right now all four discs are part of the 8TB RAID 0 array. I understand that my risk of data loss is increases in a RAID 0 with every additional drive, so it seems that I would be wise to split this into two 4TB RAID 0 arrays, as my transfer speed would be roughly the same, with the same storage capacity, but half the risk of data loss. (I do have a backup 8TB array also, to mitigate the risk of loss).

Grant Bennet-Alder, thank you for your comments, they are clear and sensible. You may have been speaking generally, but in my instance: Yes, this was a single-file transfer; Yes, the single file was very large; Yes the RAID-to-RAID duplication was on a non-OS disk, however, even the SSD-to-SSD dupe on the boot volume approached what I'd expect the limit of the SATA 3Gb/s pipeline to be, at 82% of max (and the issue of drive heads moving is moot because SSDs don't have that problem). But it was good of you to remind me of these general principles.

Hello, sorry to jump in.

I too have just bought a 512 SSD for my 2012 mac pro. I've seen varying accounts of the pros/cons of using the optical bay.

Does it still carry the same 3gb/s speed as the hard drive bus? So there is no reason why not to put my SSD there and just leave my HD where it is in bay 1?

thanks

mark

Mark (and Dickie7),

I did just as you were mentioning: put the SSD in the lower optical bay. I didn't even bother to mount the SSD on a sled or bolt it in place, it's just dangling there in the bay. Since there are no moving parts, I don't expect vibration to be a problem. However, if I were to ship the MacPro somewhere, I'd want to secure the SSD so it didn't bounce around during transit and damage itself or the Mac.

After partitioning the SSD the way I wanted, I then used Carbon Copy Cloner to move everything from my original internal boot HD to the SSD. (And after that, I pulled the HD from bay 1--keeping it as a backup disk elsewhere--so that I could use all 4 bays to accommodate my RAID array.)

Regarding the SATA speed of the lower optical bay.... As I mentioned in another post in this thread, I also tested the speed of this OWC Mercury Electra SSD (and thus the SATA bus) by duplicating a large (28GB) file. The result was an estimated bus speed of 0.247 GB/s (247 MB/s) =1.97 Gb/s. This is 82% of the anticipated SATA maximum, and 88% of the "peak read data rate" given by OWC, of 280 MB/s. Note that the accuracy of my measurement is limited by having used "command-D" to duplicate a file, and my wristwatch to time the process. I don't have any fancy analysis software.