The new SSD Thread (Benchmarks, Brands, News and Advice)

As I am running my Samsung SLC encrypted with Truecrypt, I am unable to test SteadyState on it (because SteadyState is not compatible with Truecrypt).

I did test SteadyState on my friends Core SSD and saw some pretty impressive results. As I have posted, the Core SSD + SteadyState basically approached the performance of my (encrypted) Samsung SLC in regards to random writes (I timed file copy operations as a reference).

With that in mind, I would easily go for a $50 Core V2 30GB if I needed one (which I don't).

 

Yeah I was thinking of that for $50 for my desktop's boot drive but I really don't need one.

 

someone needs to create a steadystate guide

 

Meh. If you think about it, steady state will wear out your SSD twice as fast if you do commit changes at every power down. To me, EWF is still the way to go. Running out of ram when you have 3gig of it is hard to do in a casual usage session, if you made an unprotected data partition on the side. I wonder if anyone tried benching EWF, wonder if the write speed would be crazy, since it writes directly to ram, unless the overhead implies something like reading from the disk, i wouldn't know why though.

 

I'm planning on getting an Intel X-25 SSD for my new Asus G71G if I decide to get it.

Reports of an hour of extra battery life and a 3000 point boost in PCMark05 is way worth it for the $600.

 

i'll have a go at it when i get home.... i've had ewf disabled for a while now though.

 

Did I miss something in this thread? EWF? is EWF? Does it cure the problems of cheap SSD's? Getting harder and harder to keep track of all the "cures"!

I have followed this thread from the absolute beginning, so it is possible that I missed something. Thanks, Dave

 

ewf is a write filter that writes to ram instead of disk, used in embedded systems, that often run on flash, too. so that helps.

steadystate is similar to ewf, in that it writes to a cachefile instead of the ordinary disk place, to make system restore possible. it's used for kiosk-installations where the system has to be identical at every boot. as it writes to the cache file in a sequencial matter, that helps, too.

disk encryption like truecrypt or bitlocker may help, as they collect the data they need to crypt first, and then write down the crypted data in a sequencial matter.

ewf and steadystate need to be user aware. they're both not ment for this, but for "steady state", means allways the same (100%) system on reboot. to prevent this (to be able to save _anything_) you have to circumvent those systems to write their cache into the system partition again.

disk encryptions take up cpu power to encrypt. they thus remove all your battery gains you may get with such an ssd. they may even bottleneck your read/write performance (in my case, with the ultralowvoltage cpu for example) as they max out the cpu.

the only real way is a direct software layer that only fixes the random writes by making them sequencial. MTF or how it's called promises to do it. but I never got the demo working so far, not for my ssd's, not for the flash in the asus eee. so I can't comment on them. but that's the most fitting solution (except a real hw fix for the problem).

besides that, on desktop environments, one could try a raid controller with a big cache, to solve the issue. they may make writes more sequencial, too.

in the end, having correct hw is the most fitting solution. that's why i love my mtrons. they just work

 

The Intel X25-M also facing stuttering issues? Does it mean all the MLC SSD also having the same problem?

 

well, according to one user, yes. i've no experience with it, but in general, no it shouldn't. what it does, though, is it shares the same mlc default problems that may result in stuttering (rather low random writes on a chip). but it's much better in working around those problems.. so in general, no. then again.. (i like my slc chips )

 

Well, I'm not sure whether I should go with the Intel XD25-M or downgrade to a second hand Samsung SSD SATA 64 GB SLC (not SATAII). Any recommendation?

 

There's one available on the OCZ forums: http://www.ocztechnologyforum.com/forum/showthread.php?t=47183

 

From what i 'v heard .........the SLCs experience lesser issues than the SSDs........but the Intel X 25......must have no problems

 

MTRON 3500? $400... SLC... fast... but delayed again, now 12/14. Should be faster than the Samsung.

I see the Intel X-25M dropped below $600 at Newegg today. That's the only MLC drive I would consider.

 

Thank you daveperman for the straight scoop. Crystal clear. Tells me that I should still hold out for a version 3 SLC drive Dave

 

Yup.the Intel X-25 Prices have dropped.........

 

I recall the Intel MLC having some issues, maybe from somewhere in this thread (like 10 pages ago?)

 

more around 20 to 30 pages ago, but yes, it got discussed on this thread

 

X25-M price is now down to $550:
http://www.buy.com/prod/x25m80gb-80gb-2-5-form-factor/q/loc/101/209180500.html

X25-E now available for $750:
http://www.buy.com/prod/intel-x25-e-32gb-extreme-sata-solid-state-drive/q/loc/101/209884000.html

If you benefit from these links, don't forget to rep+ me

 

If it's not SATAII, don't waste your money, the SATAI version is dog slow.

 

The SATA 1 version of the 64 GB Samsung SLC is still fast, I have it running in my HP NC8430 that I gave to my brother.

 

Surely SATAI is 150MB/s which is plenty?

 

Well, yes SATA-I as an interface is plenty for both Samsung SLCs.

However, the SATA-I version of the Samsung SLC has only approx 50% of the performance of the SATA-II version (neither version is capable of operations in excess of 100 MB/sec).

Unless you are able to get the Samsung SATA-I version at substantial discount, I would not go for it.

 

Greetings all. After MANY days of reading multiple threads and gaining some wonderful insight on what laptop to buy I finally decided to sign-up and speak up.

SSD is the technology that may eventually dethrone the hard drive, and give us a quantum leap ahead in computer speed. But, in order for that to happen the SSD makers need to do a couple of things:

1. Get the OS to play nicer with the technology. IIRC Windows 7 will do this, so let's hope M$ does a good job of implementing it. If they don't it could be one of those setbacks that takes a long time to get over. Maybe Intel will push for this to happen since they're pretty much pot committed at this point, and would be tickled pink if it goes off well with the new OS.

2. Price needs to come WAY down! Once SSD manufacturers start selling for around $1-2/GB we'll see a much, much bigger audience buying them. Sadly, what's making this happen is early adopters who will benefit from the current technology, yet pay a hefty premium to do so. A big thanks to all of you with the deep pockets and need for speed.

3. Going out on a limb here - but why design a brand new technology around an insufficient architecture? Intel, MTron, Samsung, and the rest need to come up with their own data connection to interface with the mobo. Something that has an enormous bandwidth that, while not fully taken advantage of now would leave several years worth of headway to work with. Why keep bottlenecking the technology by trying to interface it with the SATA II line, thus making sure your technology hits a brick wall you can't get around. A new port that either runs to the PCI-e line, think of the Fusion ioDrive, or something altogether new makes more sense to me.

4. Design a new RAID-esque architecture that will better utilize the SSD technology base. So far all RAID systems are built around standard hard drives, making up for their deficiencies and building on their strengths. If a new RAID level was created specifically for SSD you could get really creative, and make these drives even faster. Hey, the person who thought of the SSD was thinking outside the box, so let's get the rest of the way out and make this technology work!

The last two are probably the least important in the short run, but something that the manufacturers should be looking at now. If the first two fall into place around the release of Windows 7 we'll see a massive paradigm shift to SSD. Standard hard drives will still hold a place for years to come for people who do DV editing, store massive amounts of digital photos using cameras like the D3x, Canon 1Ds Mk III and 5D Mk II, multi-terabyte storage options will be around until 1TB and larger SSDs come down to an affordable level.

Anyway, these are my thought on the whole SSD future. I bought a new laptop with dual hard drives in it to house a couple of really fast 256GB drives I'll be buying in about two years for around $200 each. That is unless the SSD manufacturers build a new data line and my old SATA II won't take them.

 

Do you guys think a cheap MLC SSD would be alright if jsut used for games? that's all i really want the speed for, or would my games constantly hang and stutter?

Edit: this just looks pretty d@mn tough to pass up

http://www.tigerdirect.com/applications/SearchTools/item-details.asp?EdpNo=3978505&Sku=O261-3096

 

As long as the game (and other apps) aren't writing to the SSD much it should be fine. Personally, I will not buy any product that has a large rebate attached to it. It isn't worth the hassle for me.

 

So wahammer online... would Patches take rediculously long or would it be ok to use for that? (pretty much all i play anymore but i got spore crysis and WoW if i get board of WAR)

 

can't post links yet as I haven't done 15 posts but on newegg item=N82E16820227372

no rebate.

less capacity but higher read speed by a bit.

 

anyone taken a look at the exFAT file system for use on SSD?

 

My MOBI 3500 SSD's (16GB & 32GB) arrived this morning. Any test requests before I put them in production?

 

Where did you get it from? Rocketdisk?

 

would be grateful for any benchmark you can come up with, especially some real world usage benchmarks (in comparison to a 7,200rpm drive) e.g. boot time, shutdown, starting photoshop, opening a very large file, starting outlook, ...

 

If you are running a 32-bit version of Windows XP or Vista, you can always run Windows SteadyState which will take care of any random write problems you may encounter.

 

Another vote here for real world. Outlook running and browsers and other programs, and then do a AV scan or something

 

vista 64... don't really wanna go down to 32 it's a hassel and lose 1g RAM

 

With a stopwatch or is there a better way? I spend most of my time on Linux.

I'd love to compare to a 7200rpm (Scorpio Black 320GB) drive, but my wife would not be pleased if I took away her laptop. So if I do the comparisons in my Samsung netbook, it will be against a 5400rpm 2.5" drive on a slow machine. I could possibly do some testing comparing it to a Samsung HD300LJ on a Vista 64 machine, which is not the fastest thing in the world either ... It's just an Athlon 64 X2 4800+ or something along those lines.

Non-comparative tests are obviously a lot easier to do and take a lot less time.

Edit: Since what separates the SDD men from the boys is write speed, any particular suggestions there would be welcome. My first instinct would simply be to do a file copy speed comparison using the same data set.

 

Isn't patching more of a function of your download and the servers upload speed?

 

I agree on the write speed being the defining factor.
A simple file copy test does not really show real world performance impact IMO.
Outlook supposedly is very "write heavy". Editing very large files in Photoshop would be another good real world writing test.
I am trying to think of others.
When I tried an MLC disk, extracting zip files with many, many small files (e.g. 5 MB zip file with 300 files in it) was HORRIBLY slow (extracting took like 20-30 minutes).
Also installing something (driver, program, ...), running an AV scan and then extracting such zip file would be a good test.

 

You can use SuperSpeed RamDisk Plus to create a RAMDisk in the area of RAM which Vista 32-bit cannot access (3.12GB - 4GB) which is still recommended even with SteadyState installed. That way you are really losing nothing by going to 32-bit. You would place your internet cache and ReadyBoost files on the RAMDisk. This is what I'm planning on doing when I get my Core 128 next week.

 

Download patches are a function of the network bandwidth on both ends of the connection.

Installing the patches is about write speed to the drive.

 

Is it me or MFT costs 125$ a licence??? Why buy a cheap MLC drive and then pay that kind of money for a software to improve it's performance?

 

Yeah my thoughts exactly Spare:

Code:

MFT-SD16 	 MFT edition for Windows up to 16gb 	$40 	each
MFT-SD32 	 MFT edition for Windows up to 32gb 	$60 	each
MFT-SD64 	 MFT edition for Windows up to 64gb 	$80 	each
MFT-SD128	 MFT edition for Windows up to 128gb 	$110 	each
MFT-SD256	 MFT edition for Windows up to 256gb 	$180 	each

A bit steep for a glorified driver. Although the new version which is coming out in the next day or so is supposed to be very quick. Shame it doesn't support boot drives yet though.

 

The 3.5" 64GB version is in stock however...

 

Initial performance impressions of the Mobi 3500: It's fast.

I haven't had a chance to do any real world file write tests, but the Atto results are very encouraging.

I reduced my HDD down to 32GB and then copied the image over to the SSD...so dealing with two mostly-identical images, here are some real-world findings compared to the Samsung 300LJ:

Windows Defender, Full Scan (mm:ss)
SSD - 13:47
HDD - 33:51

Vista Boot (not counting time to enter password)
SSD - 54 sec
HDD - 1:21

(Note that between several boots there was as much as a 10 second difference for each one.)

For the Atto results: The first SSD test was booted from HDD with an empty, virgin SSD. The second Atto SSD test was booted from the SSD itself, and the SSD has been completely written with >32GB of data. The HDD test was done with the HDD booted, no SSD connected.

As far as the "seat of the pants-o-meter" ... programs pop up pretty much instantly. It would be difficult to time how long it takes Firefox to start up. I don't have Photoshop installed on here, and it wouldn't fit anyway - Vista is such a hog that I only have about 4GB free

Fortunately this toy is destined for an XP Pro netbook.

 

Yes but i don't know how bad the stuttering issue really is. Does it really slow things down or just a little anoying?

I'm just not to familiar with doing all that. I know there are guides for it but i would avoid it if i could... lazy...

 

Nice benchmark bicster.
Can you do the zip file test?
(Just create lots of empty text files and then zip and unzip them ... 5MB of empty text files should be enough).

and where did you get the MTRON 3500?

 

I wrote a small program to generate such a file. It turns out that a 5MB zip file of empty files contains around 36,000 empty files.

Extracting the zip file to an old Samsung SSD (MC4GE04G5APP, 4GB, ATA/66, Linux, P4 1.8 GHz) took 1 minute 29 seconds. On my work laptop it took 2 minutes 13 seconds (Windows XP, Core 2 T7300 @ 2 GHz, 7200rpm Fujitsu MHW2080BJ).

I'll do the test on the Mobi 3500 when I get home.

If anyone wants to do their own testing with the same file, you can download the zip file here: ssdtest.zip.bz2. I've compressed it again with bzip2, resulting in a more manageable 141K. (The free 7-zip utility can decompress it, as can WinRAR and others.)

I got the 3500's from Rocketdisk. I used Google checkout, ordered on a Sunday and they shipped on Monday. After I got the tracking number, I realized I gave them the wrong zip code. I e-mailed them, and they responded in 5 minutes and corrected it for me. It seems like they're on the ball.

 

Thank you

We try to stay on top of our game!!

 

Weird - I handily beat your Samsung SSD: Tested it on my Lenovo x60t Core2Duo 1.5GHz, Vista, 3GB RAM, 100GB Seagate Momentus 7200.1 ST9100821AS (7200rpm) HDD with several programs open (Outlook, OneNote, Powerpoint, IE, Word, Trillian):
1 minute 11 seconds


(this is only the measurement for extracting the text files from the second zip with WinRAR - I did not count unzipping the zip file)

 

I think just about any rotating media would clobber the Samsung MC4GE04G5APP SSD. It's a dog, probably not even worthy of the title "first generation" ... it has no buffer, and a basic ARM controller, but it's SLC.

The results for the Mobi are interesting. The program you use to extract the zip file matters a great deal! [Edit: This will be true for any drive.]

The Vista shell predicted it would take 7.5 minutes, so I hit cancel and deleted what few files it unpacked.

7-Zip took 2:11 (mm:ss).
Info-Zip took 2:25.
Winrar 3.8 took 1:23.

(Info-Zip on a Linux tmpfs filesystem takes less than 0:02 on a Pentium 4.)

So obviously, the Mobi 3500 is not always superior to a hard drive. But I suspect in the majority of circumstances, it is.

For grins I ran Atto while simultaneously running the WinRAR test. The Atto numbers were within 10MB-15MB/s of an unloaded system. WinRAR paused occasionally during that test. I'm not sure how Vista I/O scheduler works; I'd be curious to know if the pauses were due to the Mobi or due to Vista.