About SSDs in DAWs

[Sir Les 2]

Thought this might help in some way.

http://www.recordingreview.com/blog/...mputer-part-1/

Follow up:

http://www.recordingreview.com/blog/...ding-computer/

Kenneth H. Williams January 15, 2014 at 2:25 pm
"Hi Brandon, I recently changed out my mechinical hard drive for a SSD drive on a Macbook Pro running OS 10.6.8, UA Apollo Quad, Logic 9, Waves plugins, All the Spectrasonic VI’s, Battery, AD drums & Keys and a cast of others. Yeah I know old OS BUT everything was working great and I didn’t want to bear the expense that goes with the upgrade of OS. It was a beautiful thing for about one day, then Apollo started dropping audio and computer started the gray screen of death. I found out the Apollo does not play nice with SSDs at this point or fusion drives. I had to reinstall a mechanical hard drive. Has anyone else had this problem or know of a workaround? The speed of bootup, apps opening and response of VI’s was amazing to me. I”m hoping UA gets it sorted out soon. I will say there was no BS about it when I contacted them they were straight up about the issue."

Just thought about my new build with the New MAC PRO...and the solid state PCIe Drives....I wonder if any problems will occur for those buying in like me?

[mesaone]

Second link is dead. I think you accidentally pasted a shortened URL with an ellipsis.

[TOM@METRO]

We are running an SSD system drive in every computer we use here. No regrets.

[Darryl Ramm]

I use SSD drives for everything, except backups and archives and those SSD drives perform stunningly well.

If you are really pushing performance (or reliability) then all SSDs are far from created equal and there are many ways to hang yourself by doing things wrong with a HDD or SSD. SSDs do take a little attention to set up to get maximum performance. e.g its a good idea to -

0. Always have a backup plan, with bootable image and incremental backups. And test they work.

1. Get a decent high quality SSD, Samsung 840 Pro or Evo are my current SATA based SSD recommendation. I expect SATA III/6 Gbps SSDs will be rapidly replaced by PCIe based SSDs.

2. Run SATA III/6 Gbps SSD drives on a SATA III controller. A single fast SSD can max out a SATA II/3 Gbps interface. Obviously if you have an SSD with a PCIe interface then you are even further ahead. A SSD on a USB2 or Firewire interface is leaving being a lot of performance, and if you have a choice use Thunderbolt or eSATA III/6 Gbps for external SSDs (because they are fast and supports using TRIM). USB3 is the next choice.

3. Check the drive firmware is up to date when you install it, and update it if needed.

4. Make sure you have TRIM support enabled (especially on write intensive drives) for those drive models that really benefit for TRIM and are stable with TRIM enabled.

5. Make sure SSD drives are sufficiently over-provisioned. Having that unformatted/unused space on the drive provides for better write performance and for better drive wear life. Most SSD drives default to around 7% or so over-provisioning. For write intensive drives this should be up more around 20%. Just leave that space unformatted, or see utilities like the (Windows only) Samsung Magician software to set this up on Samsung SSD drives.

6. Did I mention have a proper backup plan and test it?

But even without worrying about this a good SSD will typically outperform a good HDD and be more robust and reliable.

[mesaone]

Originally Posted by Darryl Ramm

2. Run SATA III/6 Gbps SSD drives on a SATA III controller. A single fast SSD can max out a SATA II/3 Gbps interface.

Can it really? I have a question, because I'm always wary of quoted speeds - and I only have a tenuous grasp of how this stuff works.

On paper, SATA II bandwidth is 3 Gbps. Also on paper, those Samsung Pro 840 drives are 540 MB/s read and 520 MB/s write... 4320 and 4160 Mbps, respectively. So on paper the SSD has read and write speeds that do exceed SATA II bandwidth. In a real-world scenario, does SATA II ever actually bottleneck SSD?

[Darryl Ramm]

Originally Posted by mesaone

Can it really? I have a question, because I'm always wary of quoted speeds - and I only have a tenuous grasp of how this stuff works.

On paper, SATA II bandwidth is 3 Gbps. Also on paper, those Samsung Pro 840 drives are 540 MB/s read and 520 MB/s write... 4320 and 4160 Mbps, respectively. So on paper the SSD has read and write speeds that do exceed SATA II bandwidth. In a real-world scenario, does SATA II ever actually bottleneck SSD?

Yes it can, otherwise I would not have said it can

SATA II is 3.0 Gbps and uses 10/8 symbol encoding (SATA III uses lower overhead encoding as well as having twice the wire data rate). So you get an absolute maximum actual data transfer rate for SATA II of...

3.0 * ( 8 /10 ) / 8 (bit/byte) = 0.3 GB/sec = 300 MB/sec (depending on whether vendors spec a MB or MiB).

The Samsung 840 Evo and Pro can both easily exceed this performance for sequential IO, and yes you can absolutely see this in the real world.

[mesaone]

Originally Posted by Darryl Ramm

The Samsung 840 Evo and Pro can both easily exceed this performance for sequential IO, and yes you can absolutely see this in the real world.

Alrighty. I have two SATA III ports on my motherboard, and I've been shopping for something to replace my 120 GB OCZ Vertex. I'll take the plunge on a Samsung MZ-7PD256BW this weekend.

Thanks for the info, Darryl. You should be making commission. By the way, the "can it really" was an expletive, not an accusation

[Darryl Ramm]

And one more note I'll add is the performance of an SSD depends on many things, but one of them is the internal NAND-flash fan-out (i.e. how much data can the controller chip drive to/from the actual NAND flash in parallel). Because of this lower capacity drives may have significantly slower performance than their higher capacity drives of the same model. If looking at benchmark/review/performance data always make sure you are looking the exact capacity drive you are considering. With the current Samsung 840 Evo drives that means you likely want to get a 256GB or larger model.

[Darryl Ramm]

Originally Posted by propower

I have a Blackmagic Multidock which is all SATA II internal.
With a 1TB 840 EVO drive I get 340MB/s Write and 385MB/s Read.
Same exact result with 512G 840 Pro.

I understand that both drives are hitting the SATA II limit but they both seem almost 25% better on reads than your calculation would imply. Is the BM speedtest program the reason for the discrepancy? Not arguing - just interested.

The current MultiDock is speced as "SATA 3", which is damn annoying/sloppy marketing from BlackMagic, as it is unclear if this is SATA II / 3 Gbps or SATA III / 6 Gbps. They ought to be following the SATA/industry recommendations and just specing this as either 3 Gbps or 6 Gbps to avoid confusion. So are you really sure what your box supports internally?

And it always depend on what you are using to measure performance. And the easiest way of getting higher permanence measurements is to have performance measurements affected by OS disk caching or caching effects int he SSD drive. I have no idea what the speedtest program does, never seen it. IMNSHO Anadtech do some really good SSD performance testing and using the same approach they do is a great start if this stuff interests you. Unfortunately they don't make their test programs not available to use.

All my serious tests of SSD use either my own or other folks C code running on Linux. And they test stuff orders of magnitude faster than SATA III can ever get And to be honest I don't care much about SATA II or III performance, its so slow compared to PCIe to be uninteresting. But for affordable/desktop/consumer apps SATA III is mostly where it is at today. And no I don't get a commission (but if anybody is offering...).

[Darryl Ramm]

Originally Posted by propower

Hey Darryl,

I actually wrote Blackmagic before getting the Multidock and they confirmed it is SATA II internal. I totally agree that the marketing swag is totally ambiguous and unclear on this point!

Regardless- it does seem either a) the BM Speed test is not telling the truth or 2) Internally in the Multidock they are able to digest the data off the drive and get around the 10/8 rule before sending it on to the TB controller.

Whichever it is I know not but it is interesting.

There is no "getting around" this encoding (aka 8b/10b encoding see http://en.wikipedia.org/wiki/8b/10b_encoding) that is a fundamental part of the low-level electrical interface and needed for SATA II to work at all.

And "the truth" as always, depends on what exactly is being measured and how.