[LEFT]
Quote Originally Posted by GNU View Post
X25-E is better than the Vertex no? Please some Benchmarks between both SSDs.
I'd like to see benchmarks between an X25-E and an array of OCZ drives, preferably Vertex. However I don't think anyone is going to argue that a single X-25 is better than a single Vertex. The Vertex with the new IOPS friendly firmware seems to close the gap a little - it turns out there's a clear advantage with the new firmware.

The X-25 controller would seem to be the best around at the moment. I suspect a higher proportion of the cost of the drive is on the controller than anyone elses.

The Vertex's advantage is price, and when you look at the 30GB drive it just seems to be asking for RAID0.

X-25 owners have said that RAID0 doesn't seem to give any noticable speed increase. Either that means RAID0 doesn't do much, or the drive is already so fast any speed increase isn't going to deliver any real world performance increases. Something factual and absolute is that any file sizes around the stripe size or smaller gain no benefit from RAID. If these file sizes are the ones taking up your drives time then RAID0 isn't going to do anything.

Most reviews of the drives don't look at RAID0, so we'd be relying on an end user who had both an X-25 and a couple of Vertex in RAID0 doing proper testing, and not having an agenda. I just can't see someone who owns an X-25 buying Vertex drives, or vice versa.

At least the good news is that more and more information is coming out about SSD related things, which can only lead to better products, better configurations and pitfalls being avoided.

Ultimately though if an application was going to open in 5 seconds on a mechanical drive, and opens in 1 second on an SSD drive, then RAID0'ing it to shave that down to 0.75 of a second isn't really much of a benefit.

I was quite dubious about the IOPS performance of drives, until the Vertex drive was shown with both a firmware which allows high sequential speeds, and a firmware which facilitates as many IOPS as possible (I assume within reason). The result was lower "simple" benchmarks (which don't mean anything) and better results in simulated "real world" benchmarks.