pretty dismal thing!!!

aren't flash drives supposed to have a limited number of read/write cycles... like some CFs abt a yr ago.. like 100K read/write cycle... what happens after that?!? will hybrid's flash mem have maximum read/write cycles too?!?

http://www.dvnation.com/nand-flash-ssd.html is on top of the SSD market with new SSDs up to 64GB, sata and ata from PQI and the Samsung line of SSDs any Hybrids coming out soon. I got a PQI 16GB drive from them. It really does boot a PC faster than a mechanical disk, even with a lower throuput rating. My laptop boots 21% faster on an SSD thats only rated for 15MB/s read, 13MB/s write... Probably due to the 1ms seek time. 13X faster than a mechanical drive. They are touting 5,000,000 write cycles, 10 year life expectancy, etc....SuperTalent drives are not very good. Do NOT use them as system drives. Windows installation will last about a day before a blue screen of death and you'll have to completely reload windows. For a data drive, they may be o.k. Stick with PQI and Samsung. They are robust.

Edit: ment as a reply to Dirge.

Put 2 Gig of system memory in, and turn the friggin pagefile off. All of my machines with 1Gig or better have the pagefile off. There are a few programs that complain without one tho (none that I use).

Sorta half arsed design imho, looks like they just slapped some flash on a board and called it a hard drive... might as well make a 20000rpm ATA drive with zero cache while yer at it :P

... someone do the job right. The potential is there, just needs better engineering.

Looking at the first picture on Page 1, I'd like to know who the pin-bender is. ;-)

USB 2.0 flash sticks are not what I would call fast either, especially writing. I was sorta wondering about why everyone has been drooling over flash hard drives supposedly being faster. They will save power though. I'm sure that's their primary reason for existence.

I wouldn't use this solution on a desktop, but definitely for something low-powered and silent. (sitting quietly in the corner).

I use those IDE "Disk-On-Modules" (DOMs) when I build BSD and Linux based firewalls. (I set it such that its Read-Only for everything except when you need to save the settings such as IP filter settings. Logs go to a NAS or an admin PC via Syslog)...They're wonderfully silent and ridiculously (overkill) powerful for the purpose.

I hope one day, we'll reach to the point where we can completely replace these hard disks, with solid state solutions that are very reliable and acceptable in performance, but reasonably affordable. I guess we're still at the very first baby steps to that path.

Maybe they should have a write buffer of say, 256MB for constant read/write operations? (I suppose it would be a bit difficult in finetuning an algorthim such that it knows when to write data from buffer to storage only when necessary to prolong life of the drive...Of course, there's also a problem when there's a sudden blackout or batteries running out, etc or some power-out scenario).

I know a potential problem arises with WinXP (and possibly Vista). Even when you turn off the swap file in WinXP, it will override you as a "safety feature" by creating a small one. (In Linux or BSD, you don't create a swap partition and that's it...But make sure you have plenty of RAM, because if it ever runs out, it will crash!).

Flash drives seem to be in some kind of dilemma, the easy design like this one work like the classical drive spanning with raid-systems and are quite slow on sequential transfer patterns.
Interleaving needs a more complex logic (it must have very small blocks) and thus draws more power. Unfortunately because of the simplified clear logic of flash rom even greater data must be accessed on a write access. Read acces would be mostly faster, but write slower.
I suppose the good old cache on the harddisk would work miracles combined with a very efficient drive logic.
I'm only afraid the power advantage would disapperar.
MRAM or any non-volatile memory with significantly faster read/write access is more suitable, but I suppose the old rule of computer industry wins - if the new technology needs several years and millions of dollars tobe better than current high-end parts of the old technology, better improve the old one (less risk involved).

yeah, your gonna need that extra power for dx10 running 24/7.

oh well, i guess i can chuck my super slow 15k scsi drive and get a speedy 7200rpm monster.

So:
Flash disks like this have awful transfer speeds and incredible access times, as well as minimal power consumption.

Traditional disks have excellent transfer speeds but relatively poor access time and power consumption.

When will we see a disk that combines both technologies, but specifically - with logic that determines which files go on which media? Horrifically oversimplified, but files over 16KB in size go on the mechanical platter and the smaller files on the flash disk.....

In other words, when will we see the "performance" hybrid drives rather than the "power-saving" hybrid drives?

I think these kind of drives have alot of potential in the future. Right now i don't need a ton of read/write speed on my laptop that I mostly use for creating documents, email and webbrowsing. While I don't need a massive ammount of computing power I'm always running out of battery during the day. to bad IBM decided to use a freaking 1.8 in drive over whatever the next size small is in my laptop.

What I found interesting was how many apps seemed incredibly latency limited rather then Bandwidth limited. While the drive had something like a tenth of the bandwidth it still scored within 20% of the leading score in world bench.

I'd Imagine that in the future the speed of flash chips will go up drasticly. The way I see it, it doesn't seem like most flash chips are designed for high speed. while it might not be feasible to drasticly increase the the rate at which a bit is read and writen to, it seems relativly easy to increase the number of bits being read/writen to at the cost of more over wiring and suport overhead.

IANAEE

Edit: I love the article. This is the cool off the wall kind of technology that Tech Report should be covering all the time.

I have one request though. Record some boot times for windows XP starting up. I wonder what kind of results you would get.

added random thoughts.

Did you folks check to see how the limited write cycles affected the flash memory? I imagine repeated erase/write cycles would severely limit the longevity of the flash-based drive compared to a magnetic-platter design.

imagine 8 of these in raid-0, might turn around the transfer rate problem

Doesn't look like it's a good design. these guys trying to just put something on the market in a hurry or something? :P

I love how the benchmarks it won were completely unsuitable for a laptop. Who runs a webserver on a mobile computer?

Besides that, what are the advantages of this over 8GB USB flash drives which go for about half the price?

that would complement the panasonic r5 nicely - 11hours already.
http://dynamism.com/r5/specs.shtml

when you're pulling that little, even small changes make a difference.

I would think this would be a good candidate as a UMPC drive.

Hybrid drives are not a requirement for Vista Premium. TGDaily was confused:
http://arstechnica.com/journals/microsoft.ars/2006/6/14/4328

Power consumption: thumbs up

Everything else: thumbs WAY down. Diss had the application areas right too - for those who would likely never notice the loss in speed. Most IT guys would rather chew nails than suffer that level of performance.

This is the letdown of the year.

Flash drives' limited read/write capacities have me worried, even more so if it's in a hard drive. People may say it's not a big deal, but once you've got a system installed and running, the leftover space is going to be written to and read from a lot.

I wonder if this would find use in a server. TR has reviewed a server hard drive array as I recall that only used 2.5" drives.

Provided you don't need huge amounts of storage, a few of these in some form of redundant array (write cycle limitations and all that), and you've got a wicked fast array for small transfers.
-Mole

Edit: Not a drive array, I was thinking of this:
http://techreport.com/reviews/2004q4/seagate-savvio/index.x?pg=1

I wonder how it would have done in comparison with things like IOMeter...

Lets not forget these guys...//http://www.bitmicro.com/products_storage_devices.php

Test the laptop's uptime with this thing!

Goooood Lawdddd it's early.

What's the purpose of Vista requiring a hybride drive when the flast portion of the drive may suck worse than the mechanical part.

I expected at least 30% faster than mechanical drives.... which is what many have been preaching.

FP

If the Read (<17MB/s) and write (<5MB/s) are this slow ... I don't want it. Need Read>35MB/s and write>15MB/s and about 32GB bare minimum. Make the internal interface wider and faster (more power consumption but still much less than HD). It might be OK for the really small devices but doubling the time required to complete a task means I get half as much done per charge even if the battery runtime is longer.

Nice review, Geoff. Thanks.

Wait a couple more years for MRAM based drives....=)

Why the frack are these running on fricken IDE?

Sorry if I seem ungrateful, but whatever happened to the notion that if you're going to do it, do it right. SATA doesn't seem like very much to ask, at all. Good lord. It's like friggin' SATA optical drives.

Very nice power Consumption numbers.. though it's kinda slow in some things...