Your real world testing was not in fact real world. Although you acknowledge that “Systems don't idle for long in the Benchmarking Sweatshop, making it difficult to test the garbage collection schemes on the Indilinx and Samsung drives”, you went on to do the test without allowing for this cleaning. In the real world these drives idle and are cleaned through-out the week, thus these drives are rarely being operated under the false completely full “real world” conditions of your test.

You also did not test it with the easy to use wipe.exe which does an excellent TRIM to the disk in just 2 minutes on a 256 gig drive (many of us use the Microsoft driver). Additionally, although you tested using Indilinx firmware version 1370, the much improved 1531 was already being shipped at the time of your test. Since Indilinx firmware is easily user upgradable, not using it was another non-real world handicap that materially distorted the outcome.

If TR wants to create an average of 5 runs, can you include horizontal error bars showing error of the mean?

As expected, TR's results were correct the first time. This kind of in-depth followup and analysis is why I don't bother reading hardware reviews at any other site. I wonder if all of the OCZ fanboys will be back to apologize.

Great follow up.

I can't say I agree with the conclusions entirely, though. I think the file copy test is a bit over emphasized. These are small drives and I don't think it is likely they will be used that way. I find it more likely that your OS and Program Files (and maybe games) will be loaded on for the best responsiveness, while storing or shifting around large collections of video files will be left to larger mechanical drives. Next to the Summit, even the Intel drive looks so-so.

The IO results for the Summit are interesting. After a quick glance at the 7200.12 review it looks like it even falls behind mechanical drives at the highest IO levels. I'm unsure how bad IO gets when used as an OS and Program Files drive, but I suppose the rest of the realworld tests show the Summit still does well.

Still not pulling the trigger. I think I'll wait for W7 gone gold and see what the landscape looks like. Still looks like the Intel drive is the safest bet for a premium on $/GB.

I'm still flabbergasted people are shelling out over $350 for devices that offer, by comparison, minute storage capability, and virtually unnoticeable performance increases.

Is a couple seconds really worth paying 3x as much money for 10x less storage space?

The pricegrabber links to
"Intel X24-M 80GB Solid State Drive "

And leaves out newegg (possibly because of that?)

Wow. Congrats to Geoff for an enlightening and informative article.

Your thoroughness and the ability to take criticism and suggestions constructively have ensured that you are my most trusted (and favourite) writer on TR.

Awesome work, dude.

A Velociraptor and then the fastest 3.5" drive should be thrown in for comparison. The Velociraptor does give a nice boost IMO. It feels snappier then my older 7200.10 drivers.

But then, i go with the model that the OS and apps are on the Velociraptor, then all data is on my 4 disk raid on another server and loads over the gig ethernet. Copying files between the velociraptor and the raid just about gives me 98% utilization of the network. So right now the Velociraptor and the network are rather balanced in performance with up to120MB/s read/write. The raid on the other hand is overpowered, but thats beside the point for this exercise really.

Thank you, TR, for this review. I really appreciate the work that your new testbed must have required, and the extra data is welcome.

But no review's perfect, and I still have a few questions.

1) Why were no mechanical drives included in this test? Because we can't compare these results with the previous review's, we don't really know how the technologies rank. I consider this a serious omission when Geoff concludes that he wouldn't purchase any of the SSDs now. Are any of these 3 SSDs ever slower than a standard HD in the new testbed?

2) Is there a performance difference between Vista's native and Intel's AHCI drivers? If wiper doesn't like Intel's drivers, perhaps some of the SSD benchmarks were also affected.

3) While your concerns about wiper & TRIM make sense, I'd like to know how much performance wiper can restore. Would you please include just a couple Vertex benches post-wiper?

Your boot/load tests are off and you changed a lot of stuff compared to the last round. Far cry level loads got cut in more than half (almost in three) compared to the last round of SSD testing, but the OS boot up load actually gained 10 seconds. According to you guys, between both rounds of tests, a WD black actually beats the intel X-25M in OS loading by nearly 20 seconds which is in direct contradiction with what every other reviewer is showing...

And why would you guys do COD4 level loads when the last review you used DOOM3? If you're trying to validate results, it only makes sense if the tests are consistent.

And the vertex fell on its face again. Who would have thought.....

Hope you got to have a drink on Canada Day while writing this
XD

And there, the Vertex has officially fallen.
No review could have more authority than this one.

That's a wicked review, Geoff. And I was all set to buy a Vertex today as they're in the latest NCIX.com sale. For a few bucks more, if I do buy a SSD, I'll just get a Summit. Anyone have a recommendation on a new case - full ATX at least, I've got a new Powercolor Radeon HD4890 1GB card to consider. It hardly fits in my Raidmax Smilodon ATX case.

I bought a couple g.skill falcons back on June 15, and they were both dead within 10 days. The more I read on forums, etc, these things aren't ready for prime time. Check newegg's comments on the 128gb "falcon".

Not sure if the issue extends to the vertex, which I understand has the same controller.

Very good analysis! The most important part was at the end. Windows 7 changes everything for SSD's. I am running 2 x 30GB Vertex drives RAID 0 firmware 1370 on a 780i mobo. Performance slowed very quickly on Vista. I have been running Windows 7 RC1 for a few months now and the drives are staying fast without wiper.

Here is a good read about what the windows engineering team have to say about SSD's http://blogs.msdn.com/e7/archive/2009/05/05/support-and-q-a-for-sol...

As people with enough cash and need for speed to buy an SSD will surely buy Windows 7. I think Vista and XP tests are a bit irrelevant. Can't wait to see you test the same drives on Windows 7!

Benchmarks aside, it will be interesting to see what your experience is with running the Summit in your desktop for a few weeks. If Anand is right about random write performance being critical for the "feel" of SSDs, then the Summit should turn out to be a stuttering mess despite the impressive benchmark scores.

Still feels like these things are at least 6-12 months away from being a safe purchase. Perhaps they'll even be affordable in more useful capacities by then as well.

The SSDs tested are really 1-1.5X generation SSDs (ihmo). These units are really the very first ones to hit the consumers and need the feedback in order to help test/upgrade for newer models.

Intel has the "real" 2.0 models coming (as I'm sure others are as well). Platter based storage has plenty of time left (it'll stay popular in RAID/NAS setups), but I'm itching to get SSD for my machines at home.

Thanks, Geoff for the great followup.

Now please, we really need you doing something else here, so let's try to move on... At least for a while until something new happens on the SSD front (perhaps Windows 7 and full TRIM support, and hopefully an updated Intel model).

I guess then that I would sit on the fence abit longer.

I've used the Vertex and the X25-M. I really want to use the Vertex more, but the wiper utility they have for TRIM is just not where it needs to be. I had it corrupt my ICH10 AHCI 120GB drive the other night and had to rebuild. Definitely not ready for prime time.

I criticised the benchmarks the first time because too many "problems", but this time the benchmark is very well done. Dissonance should have done this the tests like this from the beginning so nobody could have problems with them. The problems I had with the benchmark the first time were not in the results, but in the fact that nobody could know if the results were representative of other systems, now we know the new ones are and that they make sense with how things are supposed work.

Its not clear to me if in the test conditions you had AHCI enabled in the bios? That made a big difference in my system. I don't know if its running firefox in Wow32 in XP64 or what but every little while firefox freezes up even to text input then after about 30 seconds writes what I typed in the buffer. Enabling AHCI made it bearable. With AHCI enabled Atto benchmark showed a 3 to 5 fold increase in read/write throughput for the smallest file sizes; right where it helps.

Not a programmer since the days of FORTRAN but seems to me this is a driver issue. Somebody needs to man up and write a journaling file system for SDD's that skips around like the sugar plum fairy writing 4K blocks hither and yon and does not erase a 512K block unless it needs to free up storage space and then only in a wear leveling fashion.

I bought the Corsair M64 and its stuttering right now, especially when doing software installs on this system build. Huge improvement once I enabled AHCI in the bios and got Win XP64 to install the drivers. Off to read the article with great interest.

Thanks Geoff, great review. Your time and efforts - and pain - put into the SSD reviews is very much appreciated.

We do seem to be in a state of limbo right now, and given the price of the current crop of Intel, Samsung and Indilinx controller-based SSD's it may be a wise choice for enthusiasts sitting on the fence to choose to go over to the bench and wait this one out until the firmware issues mature and a fully-compatible (non-proprietary) garbage collection process arrives with Windows 7 and TRIM (we hope).

Didn't the Vista drivers work better on AMD storage controllers too? Interesting.