ssd for my mac pro (early 2009)

http://eshop.macsales.com/shop/SSD/OWC/

You will also need a 2.5 to 3.5 inch adopter. like

http://eshop.macsales.com/item/Sonnet%20Technologies/TP25ST35TA/

People have been installing SSD since they came out especially in summer 2008. No reason you need to shop OWC, you can find a lot of options - any almost will do - on Newegg, Amazon etc.

I've opted for a number of Samsung EVO 500GB, larger than needed for system. $250 500GB, 250GB are only $139 which is usually ideal size. Even the 1TB is now around $400-440.

The SATA II drive bay is fine for system, not much benefit using PCIe SSD to boot from, but is if using SSDs for scratch and graphic libraries.

you also want to not 'burden' and take up space on any boot drive with user data and media, all that should be on other drives.

At this writing the Samsung EVO that The hatter has suggested are faster than just about anything else available.

Almost all are 2.5" form factor, so an adapter is needed to attach it to a drive sled.

TRIM is not really optional if you want system stability and continued good speed. A third-party utility such as TRIM Enabler (base version is FREE) is needed.

This type of Boot Drive plus Data Drive(s) organization yields the fastest overall results:

User Tip: Creating a lean, fast Boot Drive

As mentioned, you'll need some form of adapter and the best I've seen (and use in my Mac Pro) is this one because rather than being an adapter, it's a replacement sled made for the 2009, 2010 and 2012 Mac Pro's.

Also note that while some SSD's are faster than others, nearly all will be limited by the SATA II bus on the Mac Pro.

Finally, while, like everyone else here, I'm a big fan of TRIM, at some point, Apple should be releasing the public beta of Yosemite which they've already announced, and I've read that enabling TRIM on that may cause real problems.

The way to think about drive speed is Bottleneck Analysis. As long as the connection methods are faster than the Drives, there will be no appreciable slowdown.

All single rotating and SSD drives available today and most Arrays are MUCH slower than SATA-2, PCIe, or ThunderBolt, so there will be no real-world slowdown.

"A chain is only as strong as its WEAKEST Link", and a drive is only as fast as its SLOWEST connection. In this case, that is the speed at which the data spins under the read heads, or the access time of an SSD. At this writing, these are almost all quite a bit slower than the Busses available. So there will be no real-world difference in where you attach those drives.

RE: SATA Bus speed:

SATA 3 is rated at 6G bits/sec, which theoretically is about 750 Mega Bytes/sec

SATA 2 is rated at 3G bits/sec, which is theoretically about 375 Mega Bytes/sec

SATA 1 is rated at 1.5G bits/sec, which is theoretically about 187.5 Meg Bytes/sec

Most Rotating drives available today, whatever their SATA spec, can source data off the spinning platters no faster than about 125MBytes/sec.

None of the SATA Busses is a bottleneck for consumer Rotating drives you can buy today. Trying to speed up the SATA Bus will not provide any real-world performance increases for Rotating Drives.

Even MOST common SSD drives are not bottlenecked by SATA 2.

NB>> Samsung EVO will have its peak performance clipped back by the SATA-2 bus, but it will still be really fast.

Grant Bennet-Alder wrote:

The way to think about drive speed is Bottleneck Analysis. As long as the connection methods are faster than the Drives, there will be no appreciable slowdown.

All single rotating and SSD drives available today and most Arrays are MUCH slower than SATA-2, PCIe, or ThunderBolt, so there will be no real-world slowdown...

Indeed, with rotating drives, even where on the disk the reading/writing is being done will affect throughput so there's an additional bottleneck.

However, to quote SanDisk (which ought to know about these things), "SanDisk Extreme SSD, which supports SATA 6Gb/s interface and when connected to SATA 6Gb/s port, can reach up to 550/520MB/s sequential read and sequential write speed rates respectively. However, when the drive is connected to SATA 3 Gb/s port, it can reach up to 285/275MB/s sequential read and sequential write speed rates respectively." So, in this case, the Mac Pro bus would seem to be the bottleneck.

At the same time, replacing a couple of hard disks with SSD's has given me a remarkable boost in performance so I'm not complaining.

"seem to be" is operative world. There is, in benchmarks and such, and under EXTREME stress when used for scratch (which the boot drive may be forced to do).

But in everyday use, and here even MacRumors forum where people benched and ran tests AND everyday usage, SATA III 6G is not that important, certainly not a deal breaker.

I do notice on my PC with native SATA III (6 ports not just the more common two, so the controller has more bandwidth as well) t here is "some" and it feels a bit more "responsive."

But held back? No.

It also matters the size and NAND and find that larger SSDs have more pipes and channels to push I/O concurrently.

Samsung 128GB were slow, 130MB/sec for writes and not as good as their 250GB and took steps to bring it closer in terms of IO and writes. Larger is better.

I agree with all of that, including that size matters, which is one reason why both of my SSD's are 960GB. And the fact that an SATA II bus could be a limiting factor is certainly not a deal-breaker given the remarkable boost in performance I experienced with SSD's both when new and after a good bit of use. I think far too much attention is paid to benchmarks (because of their artificiality - though Anandtech seems to do a good job of testing real-world use too), like daily commuters proud of zero to sixty performance, given the use most users put their computers to; more concern should be directed at longevity and efforts to ensure that, which is one area I believe TRIM shines in.