Solid State Drives for HTPC - Page 2

The Gigabyte i-ram looks interesting with it's 4G and price point. The backup battery is only good for 4-16 hours but I wonder if a 120v trickle charger could be used with it?

a couple of things:

1. That Tom's hardware piece has been MERCILESSLY castrated by several other review sites as WILDY inaccurate and flawed.

in fact iirc, there was another post on TH that basically admitted it was flawed.

2. Wear leveling on SSD's basically means that they will last approximately 20 years with average reads/writes on a windows system. So the whole "not reliable in the long run" things is a fallacy.

3. Heat. Yes it's significantly less and this is at all operational levels, unlike a hard drive which gets hotter when its being heavily accessed.

I currently use a Dell XPS m1330 with a 64GB SSD on it, and the last i looked, this was about a $200 upgrade. So the cost isn't that high.

My laptop sees about 13 hours of use a day, and i can say that it boots significantly faster than the same laptop with only a regular 5400 rpm HDD. It also seems to access most apps faster and have significantly longer battery life (nearly 40 minutes in a side by side), mostly due to the fact it doesn't run the fans nearly as often.

this is my unscientific testing of course, but i trust it more than TH's guide which, as i said, has been crucified.

I don't have the link either, but I agree that the TH article was flawed and they published new numbers. I forget how they screwed it up. Also, Vista supposedly "doesn't use SDD correctly", so the power savings and performance "should" significantly improve.

Regarding the wear leveling, I've seen clear specifications of how many writes can be done to an SDD, but I've never seen a good study of how Vista writes worst case. I'd really question that 20 years until I saw good data. So what if the number is 2 years? That may still be good enough.

Regarding the $200 price for 64GB, that isn't a huge amount of money. But consider that you can get 1TB for less money (and still not have a problem with heat, performance, power, noise, etc.) then I'd wonder why the heck I would WANT to buy 64GB for $200.

any idea of the manufacturer of your SSD in your dell ?

Here's the website with the TH correction:

http://www.tomshardware.com/reviews/...rive,1968.html

Here's the comments regarding potential Vista speedups:

http://www.infoworld.com/article/08/...n_Vista_1.html

Lastly, here's recent Dell pricing for SDD:

http://www.engadget.com/2008/08/12/d...330-and-m1530/

It looks like they're charging $450 to "upgrade" from a 250GB HD to a 128GB SSD.

If you can find a 64 GB SSD for $200 post the link. I would also like to see some speed numbers for said SSD. A single 7200 RPM drive can do 100/90 MB/S read/write.

So where does the perception come from that SSD have low reliability? Is this based on any tests or is that just speculation from consumers trying to counter the over-the-top reliability figures that the ssd marketing guys are throwing around?

Also, the tuning of Vista to get better performance out of an SSD applies to virtually any OS... XP, Linux, OS X, they all are currently tuned to get the best performance out of a typical HDD. That's one of the downfalls of an older OS like XP... MS probably won't bother updating it since it's a legacy product for them now.

Flash memory will survive a limited number of write cycles, often in the 100K range. Wear leveling is employed to prevent writing to the same blocks repeatedly. It is a well-understood semiconductor phenomenon.

Sure... and HDD have a spinning platter with bearings that wear out and are susceptible to physical shock. That's also a well understood phenomenon. But I have yet to see anyone that has pointed to any real world test to counter the manufacturer's exaggerated claims or consumer's pessimistic fears.

...I know a lot of people have poked holes in the Toms hardware piece (for valid reasons) on ssd power consumption but at least Tom's tested that aspect of SSD's. I'm hoping someone at least tries to do the same for the reliability claims... even if their first attempt is less than perfect it's better than what we have now.

I bought the newest, lastest , OCZ 128gb core 2 SSD in the last week, its like 500$ bucks, give or take with a rebate.

I do like the product, and may keep it or argue with OCZ about taking it back, I am a fair guy, and have bought some high quality OCZ memory and power supplies over the years and been VERY satsified, so no issue with OCZ themselves, great company, again, good outfit.

These SSDs are bleeding edge, though this new price point from OCZ is making major headlines on all the tech sites.

It is actually suggested to either use NO swap file on the SSD, or one on other hard drive. In the end, I found some times to be screamingly fast, and others horrifically slow, and I mean thatttt slow. Lots of interactive reads and writes bring this thing to its knees. So can copying a large file.

Now this was under XP, Vista has another whole set of issues until itself, and some settings recommended are different from XP.

I did also use it with Fedora Linux 9, and it screamed for the two days I tried it with that OS.

Microsoft is working with Samsung on software tweaks for Vista, and those will hopefully benefit all SSDs and other Microsoft OSes when they are done.

The OCZ forum advice, leave write caching enabled,and try disabling/using NO page swap file at all for WIN XP. Some of us are using a swap file on a USB hd as an experiment.

I am seeing/hearing of all kinds of different issues with Vista for SSD users, I would not go that route. Vista apparently accesses the hard drive much, much more than XP, by orders of magnitude, so that is probably part of the problem.

This fall Intel, Samsung, and a whole slew of mfgs are all jumping into the SSD fray.

I will say this the OCZ price of 500$ for 128gb is a huge jump forward in value. The problem is, it has no sram caching and is not SLC, so at times its actually 4-8 times slower or more than my Hitachi 200gb 7200rpm 16gb 2.5". Installing my ATI 600 graphics card drivers on my HP 8510p , 4 gig dual core t7300 notebook took forever (over 10 minutes).

Usual install is about 1 minute give or take on the above Hitachi Drive. Yes, I timed it, that is an honest time. The huge number of reads and writes to install the ATI drivers and sound package brought this drive completely to its knees, the green hd light was solid for the entire time.

Now the positive bottom line, Win XP starts in 10 seconds, and many operations are instantaneous which is a fun kick to experience. Surfing, downloading, and many simple operations are *INDEED* instantaneous. Good times

But for compiling large numbers of small files, i.e. software development and large numbers of interactive read/writes, my 16gb Hitachi 7200rpm 200gb 2.5" was much, MUCH faster, 14 minutes faster in one case with a large Java compile with lots of related XML and web files in a project on the Eclipse Java development tool. Sorry, thats just awful, I was hoping for the opposite, my primary reason for buying the drive

Here is my link to my Newegg post, it says pretty much the same thing.

My review is under the name Mike34 is here

Now, this SSD may be useful for some, so also in additional step to be fair and promote the product for that niche group here is a link to the OCZ support forum.

I HIGHLY SUGGEST you read, about your O.S. and the various issues people are having, and THEN determine what you want to use the drive for.

I would NEVER use this to record video to though, I pass that out to my USB drive.

OCZ FORUM LINK

That page lists advice & experiences.

I think OCZ got to markt first with a VERY reasonably priced product, but as a daily practical drive replacement, your intended use and O.S. should be very carefully considered first, AFTER reading that page.

The forum manager does make, timely, helpful post, and does admit limitations, so I give the forum credibiilty

A safe compromise, buy the 64gb version, and install your O.S. on it, its half the price of the 128gb basically. That way you can find out for yourself, without risking as much investment.

And for the record, I DO think these things will get MUCH better, they say the next generation of Intel Mobile Centrino 2 laptops and SSDs are the bomb, and will render all current laptops junk in the next 2 years.

Huge battery lives, less heat since no HD, and new display technologies.

Desktops will similar, if less dramatic improvements.

I post all this detail, for the benefit of those considering SSDs, as it took me some time, experience, a bit of surfing and research, and 500$ to learn what I have !

IMO, SSDs are a better option for Laptops, where they need the extra power and heat efficiency. Not to mention the speed of SSDs vs typical 5400 rpm drives. Laptop drives can get noisy as they get old too.

But it would be great having a no noise htpc, fanless everything and an SSD as main and only drive. Then read all the movies from a server down in the basement.

Regarding the reliability and wear-leveling, does anyone know what happens when a region of the SDD fails? Do the SSDs always perform a read-after-write? It seems like the performance hit would be minimal. It would be nice to notify the OS of a write failure so that it can remap it without data loss. Or does the SDD store ECC, like a hard drive, to correct the data when it's later read? Any errors on read, even if correctable, could cause the OS to remap the block. Given that I haven't read about either feature I'd assume that they're not currently implemented.

If you have an SSD with, say, 5 times the space you actually need. Then, with the wear leveling ability, it would seem that you could get a lot of life out of it, since you will have lots of free area for it to move into as it goes.

As long as it provided a way for the OS to report the amount of good storage left so that you could migrate it before it fails, if you had five times the nominal lifetime to work with, that might not be a bad deal.

Or just use write filtering and don't worry about it as nothing gets written to it anyway.

Unfortunately I don't think it works that way. SSDs just emulate ATA/SATA drives, and the OS doesn't expect the capacity of a drive to spontaneously change. - it's not constantly checking for this and the drive doesn't have a way (like SCSI Unit Attention) to notify that the capacity has changed. And anyway, the bad space is located at random locations within the logical range, not at the end. The OS could certainly map around bad regions, but it would need to know before the data was purged from the OS's memory. If the error doesn't get returned until the data is next read then that data is lost. The error needs to come back on the write command.

If I'm wrong about this, let me know. But I think it's that intimate knowledge of SSDs that needs to be added to OSes to do the things that you describe.

technically, the bad bits will be in strewn all about in the physical range, as far as the logical range goes, that will be all fine and dandy... the drive should handle all that itself... although i am guessing here, that is how the old M-systems DoC worked, it handled everything itself, but they generally came with a little more then their claimed capacity to make up for lost bits over time... and this was from back in 2000, new SSD should be at least as good/self fixing?? this is 2008 (isn't it?) they wouldn't have lost functionality?

actually i heard somewhere that they were getting it up into the 1M write range without wear leveling... maybe it was from the Intel dudes.... oh well, anyway, its only bound to get better...

the big thing now is writes... anything that needs to write takes forever... so just don't do it and then you don't have to worry about it wearing out either...

I'd agree that the DoC handled the wear leveling, but I don't think the OS was aware of what it was doing and I don't think the DoC was aware of bad sectors. I suppose it could be writing ECC, just like a disk drive, but I've never seen that in the specs. Unfortunately I think it's as simple as undetected data corruption when a region has exceeded it's write limit. I mentioned it before, but if the drive would do a read-after-write then the error could be detected. (Of course if the bit error occured later, without ECC the user would be screwed. Data corruption!) I wonder what OSes would do if it got a write error. Since drives don't really use IDs any longer (to increase capacity) it may be impossible to get a medium error on writes, so OSes may simply ignore those errors, or hang, or worse, so a read-after-write may not even fix the problem. But maybe you know more about DoC than I do. We used them to boot Linux on an iSCSI box. I thought they just looked like a drive, so there was nothing fancy going on to account for wear leveling other than the fact that the DoC moved write data around and kept track of a redirect table to map logical to physical. I could be wrong.

Ah, right. I think I remember that. Intel had announced something that could basically replace DRAM on a disk controller (of course with other uses as well) where memory is constantly being re-written. In fact I don't think it had a write limit. And it didn't require a battery, which solves a lot of problems. Dang, what the heck was that called? I think they weren't going to ship until 2009. It was some sort of high speed flash. Maybe I'm hallucinating...

to be honest, I haven't looked at the datasheet for DoC since the evil sandisk corp. bought m-systems and EOL'ed it without telling me about it... bought I thought it did something about errors on its own, as far as the OS was concerned, it never knew anything other then it was a basic IDE HDD...
but anyway, we have never had a bad DoC,and we write to them all the time.. but we have had HDD's fail in the field...

your not hallucinating about the intel part, but I really don't remember the specifics, I was talking to an Intel sales drone at one of those meet and greet things that distributors put on for their suppliers and customers... Arrowfest I think? not to be confused with Ozzfest, which is completely different...
free beer, good free food... actually, come to think about it, the free beer is probably why I don't remember the specifics...
wonder if it will ever be a real product someday?

"In related news, Intel announced a new Z-P230 PATA (Parallel ATA) SSD drive that comes in 4 gigabyte (GB) and 8GB capacities, with a 16GB version following in September. Pricing is $25 for the 4GB version for 1,000 unit quantities and $45 for 1,000 unit quantities for the 8GB version. "

http://guru3d.com/news/intel-turbo-m...peed-up-vista/

Thoughts?

yea, just one before I go off and kill any remaining brain cells with a heathly dosing of alcohol.... WHY PATA?????

Because PATA still requires less power than SATA, and those particular drives are designed for ultraportable devices, not computers.

As for wear leveling, something that most people are missing is that all current SSD's have approx 25% more storage capacity than listed. This is to allow new Flash memory to be used as the old memory is worn out. No, there is no way to access that 25% extra from the beginning and make a "bigger drive". It is there to help offset the fact that over time, even with the best wear-leveling algorithms, flash memory can/will/does wear out.

A final note is that this is true for the higher end systems that I have been reading about and researching, such as MTron, Memoright (both SLC), and Samsung drives. (which are also rebranded as G.Skill and OCZ.) I have no idea how "budget" SSD's design their products, but I would hope that they follow suit.

Until recently, all SATA SSD's were using SATA-PATA bridge chips to convert the native PATA drive to the SATA signalling for it's outside connection. This converter chip would consume anywhere from .5-2w on it's own. This was the cause for a lot of confusion regarding power consumption numbers vs. the claims of the marketing department.

Currently SLC still maintains several advantages over MLC, namely durability and speed. The current rating on most SLC flash cells is 100,000 writes. On MLC it is much lower, around 10,000 writes. SLC is what has been keeping SSD prices so high. In order to either match or beat traditional winchester hdd technologies speeds, they had to use the faster of the two technologies.

The new OCZ Core SSD's that another poster mentioned, and that all of the review sites have been drooling over is an MLC based drive. This is what has allowed them to produce the drive at roughly half the cost of previous drives, as well as boost capacity.

One final thing: The thing that differentiates one drive from another ultimately has nothing to do with the memory used. It's the controller. The controller chip, combined with how many banks of memory it can access simultaneously. A good example is that the MTron drives handle roughly 64 banks of memory It has good speeds at around 120MBps read, and 80MBps write. However it still tanks during random writes.

The FusionIO drive which is a PCIe card with onboard memory offers sustained sequential & random throughput in the range of 800MBps read/600MBps writes, accesses roughly 160 banks of memory simultaneously. It's all a matter of the controller chip.

IDF is supposed to be the release of the NDA for the Intel SSD's, and we'll see what their prowess with processor and memory design brings to the SSD market.

Yeah, I use a SSD in my HTPC setup. Its only a 32gb though so I ended up not using it for the system as I first intended but its sort of acting as a recording cache. All temp files are on it, pagefile and as soon as I record tv it dumps into that one. The result is that its 100% quiet, i.e. no hd seek sounds, while youre viewing live tv or recording it. It works pretty good, no sound coming from it, it seem more 'relaxed' compared to the hd's when recording as well.

m

James5mith,

Thanks for the info. I read your desciption of SAS vas SATA connections to RAID cards also.

Always good to have someone around who has taken the time to thoroughly understand the vagaries of these computer subsystems.

This is cool:

http://www.youtube.com/watch?v=96dWOEa4Djs

I put an OCZ Vertex II 30GB SSD in my HTPC. Works great!

This is a must read:
http://www.anandtech.com/storage/showdoc.aspx?i=3531