Will installing a PCIe SATA III card improve SSD performance?

[dg27]

Unless you're using some old hardware for something like a server (as do I), you'd really want the PCI_E ports to be version 3.0.

My motherboard is circa 2010:

Manufacturer Dell Inc.
Model 05DN3X (CPU 1)
Version A00
Chipset Vendor Intel
Chipset Model X58
Chipset Revision 13
Southbridge Vendor Intel
Southbridge Model 82801JR (ICH10R)
Southbridge Revision 00

I read this on wikipedia:

PCI Express 3.0 Base specification revision 3.0 was made available in November 2010.

Should I assume that mine is version 2.0?

 

[WhoBeDaPlaya]

PCI-E 3.0 was only supported with IvyBridge and later.
X58 is definitely too old for that.

 

[dg27]

Thanks, WhoBeDaPlayer.

 

[BonzaiDuck]

PCI-E 3.0 was only supported with IvyBridge and later.
X58 is definitely too old for that.

That's what I thought. Depending on my experiment with a 960EVO M.2 in a PCI-E slot, I may replace it with a larger "Pro" drive. If I do that, I'll want to move the 960EVO to another system. Almost lucky -- we have one Ivy Bridge Z77 i5-3570K in the house.

 

[dg27]

So going back to the original question, if I have PCI-E 2 and I install a PCI-E SATA III card (which won't perform like a III), what will be faster: a drive connected to the PCI-E card or to the motherboard's SATA II port? I believe the latter, right?

 

[PingSpike]

I've been hemming and hawing about this myself for my own project.

dg27, here is a good article on this:
http://www.tomshardware.com/reviews/storage-controller-performance-ssd,3540.html

Unfortunately, its kind of old. Fortunately, it doesn't seem like any different cheap controllers have come out since so the asmedia 1061 and marvel ones are still around. I think they're missing jmicron controllers which seem somewhat common. The results for me seem to indicate it was a bit of mixed bag but that the pci-e controller won out. It seems like access time and random transfers is a bit better with the onboard SATA2 but sequenciel is clearly better on the discrete card.

 

[dg27]

Thanks for the article.

It seems like access time and random transfers is a bit better with the onboard SATA2 but sequenciel is clearly better on the discrete card.

So in practical terms, what would this mean?

 

[PingSpike]

I don't know, LOL. If you're copying big files a lot, the card should stomp the onboard SATA-II controller. But the random access is a muddy picture where the card doesn't do so well losing to the SATA-II controller in at least some of those types of tests. And wisdom seems to indicate that random access is what really makes SSDs feel fast.

I'm thinking in the end, you probably won't really notice any significant difference unless you're copying a large file.

Also remember, if you're booting off a card it usually takes longer since it has to spin up the controllers own option ROM first.

 

[dg27]

Thanks for all this. The original reason I posted was to explore ways to improve performance specifically running Pro Tools. Pro Tools and the associated libraries are installed on my Samsung 840 Pro boot drive, which is connected to the onboard SATA II controller. The project files (which one opens in Pro Tools) all live on an internal WD Caviar Black 2 TB HDD, also connected to the onboard SATA II controller. I currently have a SATA III card: two external HDDs are connected to that. I could conceivably move the PT project files to a small, dedicated SSD and connect that to the SATA III card if it would make any difference. That's what I'm currently trying to decide whether to do.

 

[BonzaiDuck]

If I had an SSD, I wouldn't connect it to an SATA-II port. With my PCIE-x1 Syba 4-port SATA-III, I believe the expectations were 500MB/s tops, but that's well above SATA-II performance.

By the way -- and I'll ask it here -- If the spec sequential throughput specs for a 960 EVO are 3,200 and ~1,500 respectively, what would that translate to for putting the M.2 NVMe in a PCI-E 2.0 slot? PCI-E 3.0 gives me the full spec performance.

 

[dg27]

This is where it keeps going around and around. Some people maintain that an SSD will perform better on the onboard SATA II than on a PCI-E card.

 

[dave_the_nerd]

This is where it keeps going around and around. Some people maintain that an SSD will perform better on the onboard SATA II than on a PCI-E card.

Because of the differences between how the two controllers are integrated. And because some people don't realize that there's more than one way to measure storage performance, so they'll argue to the death for "their" benchmark of choice.

It will likely have lower latency on the internal SATA2 port. It will have higher maximum sequential bandwidth on the SATA3 controller.

Audio editing is primarily a sequential task. So it should work slightly better (overall) on the PCI-E card. But if you were building a database server (random performance matters a lot more) it would be better to use the SATA2 port. Likewise if it was your OS drive.

 

[BonzaiDuck]

That actually might be true. I have a Crucial MX100 500GB in an old laptop with an SATA-II controller. It would only come close to 300 MB/s. I replaced the RAM with an 8GB 2x4 kit and cached the disk. The benchies now show variously around 1.2 GB/s depending on the amount of RAM used.

So it would depend on the controller, I would think. Is it x1? x2? or x4? The Syba 4-port SATA-III I purchased uses an x1 slot. And it would make sense comparing the x1 slot configuration against the onboard SATA-II ports.

In my case, I only need the PCIE controller for eSATA and hot-swap backup bays and caddies. And I'd only plan to use it for spinners, unless I need to transfer or clone something to an SSD.

 

[PingSpike]

Yeah, there's no clear answer which is why it will go round and round. The SATA-II onboard proponents are correct in some respects and the the PCI-E SATA-III card proponents are right in others. The various benchmarks in that Tom's hardware article bear out what dave_the_nerd suggested though.

 

[dg27]

It will likely have lower latency on the internal SATA2 port. It will have higher maximum sequential bandwidth on the SATA3 controller.

Audio editing is primarily a sequential task. So it should work slightly better (overall) on the PCI-E card. But if you were building a database server (random performance matters a lot more) it would be better to use the SATA2 port. Likewise if it was your OS drive.

Thanks--this is useful. So my OS drive stays on the SATA2 port.

And since latency will be lower on the SATA2, then that's preferable as well. I have no problems with latency now (knock on wood).

But in terms of the Pro Tools project files that live on an HDD connected to the SATA, would those be better off on a small SSD? Would that make much of a difference? I use Pro Tools every day, so it is a priority.

 

[dave_the_nerd]

Thanks--this is useful. So my OS drive stays on the SATA2 port.

And since latency will be lower on the SATA2, then that's preferable as well. I have no problems with latency now (knock on wood).

Latency in terms of drive access (the minimum amount of time a drive takes to respond to a request) is not the same as audio recording/playback latency, which is dependent on CPU and RAM. If you haven't had trouble with audio latency, good. If you have had trouble with audio latency, faster storage probably won't help much. (Assuming you're not using some antique.)

But in terms of the Pro Tools project files that live on an HDD connected to the SATA, would those be better off on a small SSD? Would that make much of a difference? I use Pro Tools every day, so it is a priority.

I am finding conflicting information: some people say it changed their life, other people are lukewarm. So it probably depends on your individual case, workload, etc.

I think your best option (before you spend money!) is to use Resource Monitor (on Windows anyway) to watch your HDD access while you go through your audio workflow. If you ever hit a point where you're seeing >80 I/O per second, or you're seeing a burst of read/write activity at the same time you're waiting for something to happen, that indicates you're (probably) bottlenecked by your HDD, and that's where an SSD will help you.

Maybe it's fine and you won't benefit from one. That would save you some bucks.

 

[dg27]

Latency in terms of drive access (the minimum amount of time a drive takes to respond to a request) is not the same as audio recording/playback latency, which is dependent on CPU and RAM.

Thanks for clarifying. I was strictly talking about audio recording/playback latency: On my last box (i7 930 @ 2.8 GHz w/ 16 GB RAM as opposed to i7 960 @ 3.20 GHz w/ 24 GB RAM) I did occasionally have problems with audio recording/playback latency. Now I never do.

If you haven't had trouble with audio latency, good. If you have had trouble with audio latency, faster storage probably won't help much. (Assuming you're not using some antique.)

I am finding conflicting information: some people say it changed their life, other people are lukewarm. So it probably depends on your individual case, workload, etc.

On another (audio) forum people insist faster storage makes a difference. I'm not quite convinced.

I think your best option (before you spend money!) is to use Resource Monitor (on Windows anyway) to watch your HDD access while you go through your audio workflow. If you ever hit a point where you're seeing >80 I/O per second, or you're seeing a burst of read/write activity at the same time you're waiting for something to happen, that indicates you're (probably) bottlenecked by your HDD, and that's where an SSD will help you.

I'm looking at resmon right now; not sure where to look for this. Will have to research how to use this.

Thanks for all your help!