OCZ Vertex solid-state drive OCZ has been rattling out new SSD drives with gay abandon. The 120GB Apex scored 70 per cent in our recent review and now it’s time for the 120GB Vertex to take its place on the test bench. OCZ Vertex SSD OCZ's Vertex: standard 2.5in form-factor Cast your eye down the specifications of the Apex and the Vertex and you’ll …
Please be consistent in the graph colours - it's much harder to make meaningful use of them when the colours representing the drives change from graph to graph (eg pg 4 - the OCZ Vertex starts out as yellow, the Apex green, the later graphs on that page have the Intel using yellow and the Vertex green).
The test with 4MB of MP3's isn't interesting as interesting as a test of random 4KB writes.
The reason is that the 4KB random writes simulates stuff like writing to swap, updating file meta-data and similar stuff, that will cause the OS to stutter (which, if you had tried to use one of the JMicron-based drives as the main drive in your work machine, you _would_ have noticed).
Really, you should try to use a JMicron based drive as the main drive of your machine, before testing any more drives (try for example to install a service pack on an XP installed on one of the drives, while browsing the net). Then you'll figure out why you really ought to read the Anandtech article, and include the result of random 4KB writes.
Smart people tell me the only benchmark for any SSD that matters at all is random write speed; all SSDs are ridiculously fast at sequential and burst read and writes, so there's no point measuring those, but where SSDs really struggle in a way that translates precisely into real-world use is in random read and writes. Any drive that's good at random writes will be good at random reads, so all you need to measure is random writes.
So, um, could we see that graph please?
I've gotta agree with holding off on a more final verdict until we can see the final specs of the Summit. However this thing right now looks like a very nice bit of kit and coming out around the same time I'm beginning to seriously consider an SSD option in my next windows box. So I'm looking at all the SSD reviews I can. My only complaint is the addition of the Samsung drive in this review, for the simple reason that it's OEM only. For me when I'm reading a review, unless I can outright buy a piece of hardware that is included in the review or specify it from a computer manufacturer then comparing a piece of consumer equipment against it is useless in a review.
I'm sorry, but 4MB is *NOT* a small file, and does literally nothing to simulate a world-world test. The reason that most people do a 4KB small file test is two-fold: 1) to simulate real-world usage such as browser caches, saving/updating documents, access your OS swap file, etc; and 2) flash memory uses a 32KB page/cell size, so it gives a world-world performance measure of those real-world writes. A 4MB file will be written as many sequential writes which is hardly a real-world scenario. Yes, you most definitely will have time when you're copying 4+MB files between drives, but it won't be nearly as often as writing small (<=4KB) files, especially considering your OS swap file will be using those small writes. That's also the reason the last test suite ran multiple tests from 0.5KB through 8192KB. That's really the only test you need to show.
I bought one of these for my Vostro laptop which was grindingly slow with the standard 5400rpm HD. I used to switch it on, log in, go for coffee, have a chat at the coffee machine with my colleagues, come back and it would still be booting, and yes, I did defrag. The Vertex absolutely transformed the machine, and now the time from login to a useable prompt is just a few seconds. The system is snappy and resposive under any realistic load. I can't say that I never get pauses - it is running Windows after all - but I don't think the SSDis the problem here.
I use the SSDs and HDDs that I review both as Windows drives and also as data drives and the widely reported JMicron small file stuttering problem hasn't caused me any problems.
I can, however, see strange results in testing. Atto and CrystalDiskMark show horribly reduced transfer speeds with the 4KB data size so I am prepared to believe that certain SSDs may cause problems in certain applications.
My question is this - do you ever transfer a large block of tiny files around your HDD/SSD?
I was unimpressed with the OCZ Apex but the Vertex is a properly good piece of hardware, as is the Samsung 256GB and the Intel X25-M.
"My question is this - do you ever transfer a large block of tiny files around your HDD/SSD?"
If you use your SSD as your system drive, then yes. You can use SysInternals' File Monitor to verify this. The concern isn't really about small files; it's about small transfers (whether those transfers are individual files or parts of one file). The reason is that an SSD controller can only write data in 32KB blocks (as opposed to 512-byte blocks for HDDs). So for an SSD to write less than 32KB, it needs to read that 32KB from the flash memory into RAM, save the new data to the appropriate spot in RAM, then write that 32KB block from RAM to the flash memory. This is the block-level, so whether the data being written is a file or not is irrelevant.
Here's a small sampling of activities that use small writes (most are 4KB or less, but all less than 32KB): Windows writing to the registry, Firefox 3 writing to its sqlite databases, IE and Firefox writing cache files, reading your email, compacting your folders in Thunderbird (compacting my folders just used 18,000 writes of size 4KB or less), opening an OpenOffice document (OOo issues many 1KB and 8KB writes to temp files), installing programs, VMWare (almost all writes to the virtual disk are 4KB or less, and all that I saw while booting the guest machine were less than 32KB), and Apache (writing to log files).
So yes, you do make a LOT of small transfers if you're using an SSD to replace a HDD (which is, after all, the intended purpose of SSD). I'm not sure about Vista (I don't touch it), but when formatting a disk in XP using NTFS, it still defaults to 4KB cluster sizes (and the HDD standard of skipping the first 63 sectors), which will result in an SSD with even poorer performance. It's certainly possible to get great performance with SSD, but software needs to rewritten to use 32KB transfers.
As my comment already said, small file random write isn't only useful in itself, it's a perfect proxy for small file random read. Which happens all the damn time. Like, whenever you launch any program - the executable (usually a fairly small file) is read into memory, and so are a whole bunch of ancillary files (again often quite small ones) - think data files, icons, shared libraries, all that garbage.
And, say, system bootup is just a series of executions of such fairly small processes, each of which gets read off the disk into memory.
You the user may not interactively shift around large amounts of small blocks of data on your disk very often, but the system's doing it all the damn time.
Yesterday, I had to write 70,000 tiny tiny files over to a flash stick (and no, couldn't zip 'em). Only a few gig, but took hours before I finally gave up and chucked them on to a whacking great external HDD instead in the space of 20 minutes.
For the hordes of tiny tiny files, flash really can't cut it.
I find the comparisons between drives really useful in Reg reviews, many thanks.
I took the plunge and bought a 60Gb Vertex SSD as I wanted a quiet drive for my P4 HTPC in the lounge. The HTPC runs Vista Ultimate and took 1m25s to boot on a standard 7,200rpm Samsung HDD. The Vertex SSD brought it down to 1m15s with no disc vibration or clicking noises! Very pleased.
Also took advice of some sites and turned off the paging file to prolong the life of the drive. No ill effects so far as the 4Gb of RAM seem enough for HTPC duty.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
