The Intel X25-M 80GB Solid State Drive Review

[DX]

Man that is fast. If only I had a job that paid 500K a year then maybe I could afford to build a system with those in it stripped RAID0 =)

[gwolfman]

hmm, very interesting, especially the first graph where it took about 7 or 8GB to spike

[lordmystic]

Can I use two X25-M in RAID 0? Amazing speed...

[Apoptosis]

Can I use two X25-M in RAID 0? Amazing speed...

of course... I've seen four in RAID 0 before and that is some serious performance!

[guiri]

Guys, I"m new to this forum and I wanted to ask you some specific questions. I just had a custom computer built with three Raptors in Raid 0 for performance but UPS trashed it when they shipped it so now it has to be rebuilt so my question is, should I go with the X25's instead.

Here is how I use my computer. I tend to multitask fairly often. I do webdesign so when i work I have often Photoshop, Corel Draw, many browser windows, ACDSee, Dreamweaver and maybe a few other windows open. This is so that I won't have to waste time opening and closing programs and also, because they retain the info such as where I got the last file from. This speeds up my work flow.

I tend to access pictures or graphics quite a lot BUT, my graphics are often fairly small in filesize. I would say that most are under 20 megs.

I also do photography and I tend to have thousands of pictures in a folder which is a reason for wanting fast disk access. For my storage I had two 1tb seagates in Raid 0.

I WAS also going to store work files that I access frequently on partitions on the three raptors to speed up my access to them. The three raptors are 150 gigs each and there's around 80gigs or so for the XP partition and the Vista partition and the rest is available for storage.

My question is, will the X25 do anything for me in terms of speed and if so, WHAT would the estimated speed increase be?

The system has 8 gigs of DDR2 ram although the Xp will only access 4 of it because it's 32 bit. The Vista is only there IF I need to access 8 gigs for some reason as it is 64 bit.

My graphics card is a Palit with 1gig on it and the processor is an 8400 overclocked to 4 gigs.

Also, I am wondering about the X25 as a single unit because the builder had problems making Acronis work on my system. Presumably because of the P45 chipset but the Raid 0 could be another problem so i was thinking of possibly avoiding that but at the same time, he highly recommends it for speed so the question is also, what difference would ONE x25 and TWO x25's in raid o be compared to my current (untested) setup.

Here's a link to another forum where they have discussed my Acronis issue at length.
http://forums.dpreview.com/forums/read. ... e=29333476

Thanks for any advice

George

[DMB2000uk]

I read half of that acronis thread on your dpreview link, but it got mind numbing and I couldn't bring myself to read the other half.

Was there any outcome apart from people bashing raid 0?

Back to your question of which setup works best; the outcome is going to depend on if you need the write speed advantage, or the read speed advantage.

The X25M is great at read speeds, and will about match your current 3x150GB Raptor setup, but falls to just single drive performance at write.

X25M in RAID 0 would give you about twice the read speed of your current raptor setup, but only 2/3 of the write speed of your current raptors. But at nearly $1200, is it really worth it?

Dan

[guiri]

Dan, that's the thing, I'm not sure if I need read or write speed. I DO save files all the time when I work so i would assume that I need more WRITE speed than read speed but I have never sat and analyzed what I actually do. I was hoping that you guys could tell me from what I wrote but I guess not enough info huh?

Let's see if I/we can analyze this. What the X25 does since it's a boot drive is access info very fast. That is, it'll load anything fast that's on the drive including booting up fast and opening all the programs, correct?

IF I have files on it, it'll find and open them up fast too, right?

What speed difference in READ speed is there between my three raptors in Raid 0 and ONE X25, and TWO of them in Raid 0 which I assume is what you meant when you said it was $1200?

How about the same question but for WRITE speed?

How much faster would three raptors in raid 0 be vs a single VELOciraptor?

As for the Acronis thread, it DID get long and way too complicated for me
and as for the outcome, I don't think so but frankly, I never really made it through OR understood most of it either.

THanks

George

[DMB2000uk]

Let's see if I/we can analyze this. What the X25 does since it's a boot drive is access info very fast. That is, it'll load anything fast that's on the drive including booting up fast and opening all the programs, correct?

Correct.

IF I have files on it, it'll find and open them up fast too, right?

Correct, and the finding part will be even quicker (near instantaneous [NB. The seeking of files, not the opening]) as it has no platter to spin round to look for.

What speed difference in READ speed is there between my three raptors in Raid 0 and ONE X25, and TWO of them in Raid 0 which I assume is what you meant when you said it was $1200?

I answered most of this in my last post, but to re-iterate:

One X25m vs 3 Raptors: Pretty much the same read speed (give or take a few MB/s).
X25m: 200-220MB/s
Raptors: 200-220MB/s

Two X25s vs 3 Raptors: X25M RAID0 will be twice as fast at reading.
X25m: 400-440MB/s
Raptors: 200-220MB/s

How about the same question but for WRITE speed?

One X25 vs 3 Raptors: The X25M will be 1/3 the write speed of your Raptor Raid0 (so one X25m is about one 150GB Raptor in write speed).
X25m: 70MB/s
Raptors: 200-220MB/s

Two X25s vs 3 Raptors: X25M RAID0 will be 2/3 the write speed of your Raptor Raid0 (and over twice as expensive)
X25m: 140MB/s
Raptors: 200-220MB/s

How much faster would three raptors in raid 0 be vs a single VELOciraptor?

A lot

As for the Acronis thread, it DID get long and way too complicated for me
and as for the outcome, I don't think so but frankly, I never really made it through OR understood most of it either.

I think a lot of people went off on a tangent about one thing or another, so if there was something useful in there, it got buried.


You should take the numbers in this post with a pinch of salt as I had to scour the internet for people with 3x150GB Raptor results, and there weren't very many so I can't vouch for how accurate their benchmarks were. So think of them as on average results.

The other advantage the X25M has in read speeds is that it is a sustained 200-220MB/s and will keep reading that fast till the end of the drive, where as the Raptors will slow down the further they get into the middle of the platters (they start on the outside of the disk which is spinning faster, so they can read it quicker and faster).

The X25M will also be silent, compared to what will probably be quite a noisy raid0 setup with all the raptors.

Dan

[guiri]

Thanks Dan

George

[sitnduck]

So... any further info on the 80% thing? BTW Kudos on finding this out!

This is HUGE! Looking at the dropoff in the Windows Media test, this means write performance just DROPS OFF THE CHARTS when the drive is full!!! Planning on using it in an enterprise setting with it mostly full, that's a game changer potentially!

Any firmware fixes?!

[Apoptosis]

well... When I filled the drive up to 80% I took a folder with a couple applications and copy/pasted it a dozen times to fill up the drive... Intel said this about my methods:

I also assume this was an operation of copy file folder 1 concurrent with 2 and 3 and ... This is a VERY random workload at the SATA interface, which will chew up the SSD system very similar to an IOMeter random workload, not containing ANY of the typical client usage sequential writes which are self healing. If you simply copied a bunch of user data, say from a USB drive, the performance drop would have been significantly less. Similarly, if you took your current drive, deleted the bogus content, and then use a usb drive to copy files sequentially, similar to real user content creation, you will see smaller performance drops.

If you run PCmark a bunch more times (as is), you should see the performance creep back up as the drive heals itself with typical client usage.

We will soon be releasing a tool, which performs this "defragmentation" of the internal ssd, called the ssd optimizer.

I've since removed the data and formatted the drive, but the performance isn't what it used to be.

intel_before_after.jpg (133.33 KiB) Viewed 4983 times

I figured just deleting the data would due the trick, but no... Then I figured a format would do the trick, but no dice... As you can see above the burst speed went from 140MB/S down to 90MB/S and the transfer rates are down as well. The test system is the same.

After I formatted the drive and removed the drive letter and ran HD Tach with the Long benchmark and tested the write speeds... The Read scores were okay, but the write scores were flat.

intel_ssd_ich10.jpg (121.38 KiB) Viewed 4981 times

I sent Intel an e-mail last week asking to RMA the drive as I think my copy/paste of ~20GB killed the drive. This was the first SSD drive I used and I didn't know copy/pasting a large chunk of data would ruin it.

[DMB2000uk]

Unless expressly told, I don't think anyone would realise that would kill the drive.

Dan

[guiri]

Well, hell, if you can't do that, then the damn thing is useless isn't it?

I mean, I do that every so often..

[dicecca112]

well... When I filled the drive up to 80% I took a folder with a couple applications and copy/pasted it a dozen times to fill up the drive... Intel said this about my methods:

I also assume this was an operation of copy file folder 1 concurrent with 2 and 3 and ... This is a VERY random workload at the SATA interface, which will chew up the SSD system very similar to an IOMeter random workload, not containing ANY of the typical client usage sequential writes which are self healing. If you simply copied a bunch of user data, say from a USB drive, the performance drop would have been significantly less. Similarly, if you took your current drive, deleted the bogus content, and then use a usb drive to copy files sequentially, similar to real user content creation, you will see smaller performance drops.

If you run PCmark a bunch more times (as is), you should see the performance creep back up as the drive heals itself with typical client usage.

We will soon be releasing a tool, which performs this "defragmentation" of the internal ssd, called the ssd optimizer.

I've since removed the data and formatted the drive, but the performance isn't what it used to be.

intel_before_after.jpg

I figured just deleting the data would due the trick, but no... Then I figured a format would do the trick, but no dice... As you can see above the burst speed went from 140MB/S down to 90MB/S and the transfer rates are down as well. The test system is the same.

After I formatted the drive and removed the drive letter and ran HD Tach with the Long benchmark and tested the write speeds... The Read scores were okay, but the write scores were flat.

intel_ssd_ich10.jpg

I sent Intel an e-mail last week asking to RMA the drive as I think my copy/paste of ~20GB killed the drive. This was the first SSD drive I used and I didn't know copy/pasting a large chunk of data would ruin it.

Intel's gotta realize there is no such thing as typical usage. Usage is usage. You should plan for all contingencies and if your drive can't handle it, fix it. Not a fan of companies now adays saying its the users fault all the time

[Apoptosis]

notice Intel acknowledged the performance drop when the drive fills up though:

If you simply copied a bunch of user data, say from a USB drive, the performance drop would have been significantly less

[sitnduck]

Thanks so much for the update! So if I get this straight, the slowdown was caused by your copying of a lot of files in a certain way (ouch!!!), but that the problem will be solvable with their future "SSD optimizer" tool??!

Can they tell you when that tool will be online, and it would be great if you could test it to confirm that claim... it sounds like even reformatting doesn't help that much! This really puts a dent in this one for us!!!

[Apoptosis]

Thanks so much for the update! So if I get this straight, the slowdown was caused by your copying of a lot of files in a certain way (ouch!!!), but that the problem will be solvable with their future "SSD optimizer" tool??!

Can they tell you when that tool will be online, and it would be great if you could test it to confirm that claim... it sounds like even reformatting doesn't help that much! This really puts a dent in this one for us!!!

The poor performance seems to be how i copied and pasted files from the drive back to itself, which is how I filled the drive up to 80% as it was the quickest way for me to fill it up. I tried to 'repair' the poor performance of the drive by formatting with Windows Vista Ultimate 64-bit with the drive as a secondary (e:) with no success. The performance was this after a full format:

intel_ssd_ich10.jpg (121.38 KiB) Viewed 4903 times

I almost gave up and asked Intel to RMA the drive, but while I am waiting for the response I used HDDERASE to see if that would do anything versus a windows format. I did an enhanced secure erase...

secure_erase.jpg (73.07 KiB) Viewed 4900 times

After a few minutes the erase of the 80GB drive was complete... I fired up the system and went straight to HD Tach after the system sit for idle on the desktop for 15 minutes and got this for performance.

intel_hdderase_3.jpg (92.79 KiB) Viewed 4901 times

After the benchmark completed I ran it a few times and the results were the same across the board with the burst speed being identical every time even. Looks like the performance of the drive is back to normal, so I didn't 'ruin' it. Whatever I did do to the drive by copy/pasting data to it wasn't able to be fixed with a simple Windows format though and it took HDDErase to do the trick.

[sitnduck]

Mmm... Thank you again for the update. Any word from Intel as to when they would release such a "defragmenter"?

I guess all this fancy footwork is how they got MLC to be so fast, at least when the drive is empty. I never thought we would go back to the days of defragmenting

[sitnduck]

...although by "defragmenting" I'm sure it would mean more like "wear management algorithm fixing", not physically repositioning data so it is contiguous.

[martini161]

...although by "defragmenting" I'm sure it would mean more like "wear management algorithm fixing", not physically repositioning data so it is contiguous.

it doesnt need to make the data contiguous, since it has no moving parts it takes the same amount of time to acses it no matter where it is on the drive