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

Yes, try the Intel SSD Toolbox and do a manual TRIM.

I would also try the IRST drivers 9.5.4.1001 and see if you get a further speed boost.

As for your battery life issue; try calibrating the battery again. Also, try it without Battery Bar Pro (I found it decreased my battery life during my trial with it).

 

Would probably be more meaningful if you used the 1000MB test size, just to ensure that the cache was not being tested (although, it shouldn't be for an Intel SSD).

Thanks btw, for posting that!

 

I cant wait for Micron SSD in Q1 2010, it looks hot.

 

Phil, I dunno then. That's really odd. Maybe there's a program very weirdly and annoyingly running in the background that's preventing the SSD from idling?

Just editing to wish all of you a Merry Christmas

 

Ok after running Tony Trim on my Samsung 256gb SSD (with VBM19C1Q) and experiencing severely degraded performance, I cleared the nand and started over. Well I'm happy to report things are back to normal, here's my new Winsat results:

Code:

NV Cache not present.
> Run Time 00:00:00.02
> Disk  Sequential 64.0 Read                   224.70 MB/s          7.4
> Disk  Random 16.0 Read                       124.65 MB/s          7.3
> Responsiveness: Average IO Rate              1.70 ms/IO          7.3
> Responsiveness: Grouped IOs                  10.93 units          7.0
> Responsiveness: Long IOs                     6.25 units          7.6
> Responsiveness: Overall                      68.24 units          6.9
> Responsiveness: PenaltyFactor                0.0
> Disk  Sequential 64.0 Write                  173.47 MB/s          7.2
> Average Read Time with Sequential Writes     0.316 ms          7.9
> Latency: 95th Percentile                     1.243 ms          7.9
> Latency: Maximum                             10.821 ms          7.9
> Average Read Time with Random Writes         0.550 ms          7.9
> Total Run Time 00:01:11.90

 

Yeah, I was finally able to restore the performance on mine too but I had to disable TRIM with the fsutil behavior set disabledeletenotify 1 elevated command prompt, reboot, AS Cleaner, rebooted and left at logon screen for 10 minutes. Apparantly the TRIM command was blocking PRF or GC whatever from kicking in.

 

I really don't see point of bothering with trim when "GC" is doing it all so good user/OS independent ?!?!

 

I re-enabled TRIM and rebooted, ran an Iometer test file leaving disk size at 0 so it completely filled the drive and deleted the file. Re-ran the winsat disk benchmark and it seems to indicate that TRIM is not working on my system and since it also seems to be blocking GC from kicking in, probably going to keep it disabled. Here is the winsat disk benchmark after deleting the Iometer test file.

 

Totally agree on this.

Got 2 X VBM19D1Q RAID0, once a day or once a week, I just reboot and leave on password screen for 30 minutes and I'm good to go. I don't need TRIM, I got ITGC and, usualy, when you've got 2 tools aimed at doing the same job, it's rarely a good idea to use them at the same time, think about anti-viruses which try to compete against each other for finally ending screwing up the whole system... ...been there, done that, already got the T-Shirt...

How do I know my ITGC works?

Well, sure, I can benchmark to get an idea, but I got another way. Say I know I got a file called whatever.file, say I delete it. Now, I run Restorer Pro, and if this guy finds my file (which is deleted, but not really as you know) AND if it can recover the said file, I'm in trouble. Curiously, after leaving the laptop on idle at loging screen, if I run Restorer pro, this guy doesn't find the file anymore, therefore, proving it has really indeed been deleted, meaning GC works.

Why would I want/need TRIM then?

Especially in my case where a RAID pile is used, so TRIM would be useless anyways...

Bottom line, if ITGC works, why bother with TRIM ?

 

Because TRIM is supposed to work instantly when you delete something without any need to reboot and sit at logon screen for 10 minutes. I would prefer TRIM but since it doesn't work on my system and prevents GC from working, I'll keep it disabled.

Also, with TRIM enabled my system would lockup sometimes when right clicking on things. It came to my attention because another user was complaining of this with the new firmware.

 

I get your point about TRIM being faster than GC.

However, for myself, unless I would notice that ITGC would be so slow to kick-in (kinda days or so), and if this would really affect my system's performance (severe degradation, easily noticeable), then, yes, I would put a lot of energy trying to find a better way, TRIM or else.

But, for the time being, it's pretty flying fast, so I really don't see the point of chasing TRIM. Ever defragmented an old HDD? Might take more than minutes in some instances... I consider, while this is NOT exactly true, that time taken for ITGC to do it's work is like when I used to defragment my HDDs, just a disk self-hygiene task... like me brushing my teeth...

Merry Christmas to All !

 

well, trim is the proper way, as it lets the os and the ssd work together perfectly.

but i life since over a year without trim without any issue. so it's no drama to not have it. even on ssds that don't really gc or anything (the mtrons).

best would be all working, obviously but even without it, hope everything works in one way or another for each of us. that's what's important.

merry christmas to all of you, too!

 

merry xmas to all from me 2
and I wish all of you find 512GB slc ssd beast under the xmas tree delivered by some geek santa

well, for trim, the way I see it, anything that OS or user need to do to to keep disk running nice and fast is not good...
I really don't wanna some crapy windows OS to handle it, or, even worse, user himself...
all I wanna to have is disk that is behaving like any other hdd, just very faster...
so, for me, "GC" is the way to go, not trim

and, as for GC, I really see no point of leaving computer on logon screen couse GC kicks in even when you reed forums on internet and every other stuff that is not hdd related
works on my slc and also on tomy b. mlc

 

well, the os HAS to communicate with the disk. to inform about any data that has to be stored. trim equals that out by letting it inform about anything that gets deleted, too so it's balancing out.

but yeah, a disk should work well without trim. trim should just make any "communication problems between os and disk" a thing from the past. but even without it, the disk should work well (and that's one fear i have of trim: cheap ssds won't work without 100% working trim anymore. so anything pre-win7 can't really be installed on it).

no halftera slc for me here... but i wouldn't want it. i'd prefer 1tb mlc, and more than one of them. so i could make my home server quiet

 

The point of GC at the logon screen is to ensure that the system is in an idle state. Supposedly GC kicks in when it detects the system as idle -- there is a problem with SSD's seeing many notebooks as being idle. Also, users with Nvidia chipsets report GC not recognizing their system as idle unless at logon. I know when GC is working on my system at logon screen after a minute or so after rebooting there is rapid HDD light activity for about 5 minutes, then intermittent HDD light activity for about another 5.

 

btw, my intel ssd has 2.55TB Host Writes so far or, my cells are written over nearly 17 times by now.

yeah, it will soon die. i soon reach the 10000 times overwritten



this thing is btw in use since june. so it's about half a year old. and has no trim and such and works great.

 

Merry Christmas all and well....I have to say this thread still amazes me....easily the most active on the site...

 

merry christmas to you, too. starter of that magical ssd trip thanks for the info you spread, helped me diving into the best thing for pcs and laptops. now, some thousand bucks later, i still think they where worth each cent

 

Basically, both TRIM and ITGC are aimed at preventing performance degradation occuring form having to delete data prior to write new stuff. This is because when data is "deleted" by the user, only the data in the index (the table of content of the disk) that points out to the real data on the disk gets deleted, thus letting the O/S know that the "real data" is OK to be overwriten, while still physically there, thus the garbage... HDDs operate the same, difference being that overwriting data (delete and write) on them is not affecting the performance as on SSDs.

I can't really figure how an SSD which has neither TRIM nor ITGC can maintain it's original performance once every single cell has been written to at least once. But if that's the case, it's great, just need faith, hey?

Specifically PM800 speaking, it's true that ITGC do kick-in even if my laptop is not sitting idle at logon screen; I however kind of notice cells cleaning to happen faster this way, thus me doing it. Again, I consider ITGC as I did consider defragmentation on my old HDDs: an hygiene-self-task of the disk itself, and neither with those drives nor with my SSDs do I want to "work" while this auto-cleaning task is happening, just feel like leaving it alone, I do not like neither having someone looking at me when I brush my teeth...

It's so true however that, whatever the way it's done, TRIM, ITGC or any other esotherical stuff , thing is that if it works, that's all we want, don't we?

I am foreseing a RAID10 pile of 5 X 512GB SLC's.... Dreamin' is free...

 

eye: it's simple: by writing to the full disk once, it's allready in degraded mode. sell that, and it will continue at it's sold performance.

that's the idea.

 

This is a 160GB G2 Intel with Intel Matrix Storage Manager version 8.9.4.1004
(They are WHQL certified, but not yet released by Intel)

 

my 160GB G2 SMART info 'power on hours' keeps constant as 26 from day one after firmware update and never changes so far, intel toolbox, crystal disk info and everest ultimate all shows this same number in SMART info in power on hours column, this seems not right, anybody got same issue?

 

Yepp, my power on hours are nonsensical...don't change...
Or change very slowly...

I have 49 though...

 

Well, my point exactly: if every single cell gets written once, and providing there's no such subsequent "cleaning", therefore, performance will already be "degraded" from that point, but will not degrade more, no?

If you buy it already "used", you should then not notice any further degradation, correct ?

Still, performance of a such drive will be worse than it was prior to any writes, yes?

But sure, I get your point: If you buy it "used", you shall not notice any further degradation because the "job is already done"...

 

Supposedly, you can run HDDerase on drives without TRIM or GC to recover performance. I imagine one would want to do that anyway if selling a harddrive. I can't report whether or not HDDErase works because it doesn't recognize my drive on this system.

 

several companies delivered NEW ssds in that "degraded" mode. it's much more important to deliver an ssd that performs reliable than one that performs well. this was, before ssds entered the consumer marked. where numbers count.

 

Has anyone seen this article?

http://www.fudzilla.com/content/view/16960/1/

Q4 2010

Intel's new X25-E SSDs will be based on 34nm technology and is MLC. Code name Lyndonville. Capacities of 100/200/400GB. My speculation is that X25-E MLC drives will be based on 2-bit MLC and X25-M MLC drives using 3-bit MLC.

(The capacities are interesting. If we assume the 100GB version has 20GB reserved space with 120GB equivalent in flash chips, does it indicate a 15-channel controller? I wish. 510MB/s read 500MB/s read here we come?)

Transition from SLC to MLC must mean significant controller improvements. It's looking really exciting in SSD land. Their presentations say the focus will be on stable performance which is a departure from previous mentality of "peak performance".

I think we'll also see PCM-based Intel SSDs in the near future. 2011 maybe?

On the Intel drives the problem is compounded by the fact that the high IOPS/random write speeds result in faster degradation. It's why the G1 drives have much more robust GC than the G2. G2 can sacrifice that GC for even higher peak and IOPS since it has TRIM.

 

I wonder where we'll be in about 2-3 years time... that's for when I plan my next laptop upgrade... or maybe later if my SZ lasts and works well enough - but I might upgrade to whatever is current then.

 

I tried to install those rapid storage manager drivers, but it says that my "system does not meet the minimum requirements" ;;. That's... odd... How doesn't it meet the minimum requirements? I'm using win 7 x64 and I do have an Intel SSD.

 

They require at least ICH7 and a chipset newer than 965... I think its 975...
That's why.

 

Mine's an Intel ICH8/ICH8R chipset in a D901C... but... I think this platofrm is PM965. D'oh.

 

Same for me, PM965... but in the end, it doesn't matter that much, does it?

 

Wow... Intel is abandoning SLC flash. I find that confusing not from a price or capacity sense: would any type of MLC flash be suitable for use in a high use server or high end workstation? Maybe Intel figured out a way to keep the cell lifetime up. Let's not forget their marketing stuff specfically says the MLC x-25 M isn't viable for long term server use. It's not like the hefty price tag for the E series would allow businesses to replace the SSD after like 2 years of use. Plausible but ludicrous?

Besides, I thought Seagate had already cornered the market on disposable hard drives. (much in the tradition of those late 90's Fireballs and Hitachi GXP Death Stars)

 

ah, jeah, I agree, and, afaik, samsung is the one that sells them once they have every cell used, right ?!?

BUT
and this is BIG BUT



if all of this is true, how come I'm able to degrade it even more with some IOmetre scenarios ?
and I'm not talking only about mlc, slc is affected to ?

that's what I can't understand

 

you can try to force it install by uncheck compatible box.

 

just tested 8.9.4.1004 with AAS and winsat, it seems worse than 9.5.4 in my case:

 

Well, the "8 series" is the matrix storage driver, and the "9 series" is the Rapid Storage driver.

Anybody with a 965 chipset can only instal the "8 series" while anybody with a newer chipset can instal the "9 series" which also requires ICH7 or higher.

You're possibly better of with the Rapid Storage Drivers - aka "9 series".

 

Curious:
Are you from Germany? (German software in your screenshot )

And you are running your drive in IDE mode? Else the Consecutive Read speeds seem low... my Vaio did that with AHCI off...

But then, with AHCI off you couldn't instal the matrix drivers... confused...

Edit:
You seem to be in IDE mode... you can get higher sequential speed with IDE mode off and AHCI on

 

ah i didn't extract language file for that AAS program but i guess that program somehow cannot detect my AHCI connection. i do use AHCI as i also use intel turbo cache for temp directory here. btw, you can always force to install newer version for a test by unchecking 'show only compatible device' when manually updating the newer driver. i have 965 chipset (X3100) but can use 9 series driver with no problem.

 

I don't get a "show only compatible device" error... I execute it and get an immediate error that it didn't find a supported device... that's it - closes...

I'm happy with the matrix storage manager drivers though

About your AHCI - the sequential Read speed is about 15MB too low... can you try CrystalDiskMark?

 

yeah, you have to choose update driver from device manager, then choose "browse my computer..." then choose "let me..." then uncheck "show compatible hardware" then click "have disk..."....

here is the result from CDM:

 

I'm not really sure of that benchmark...it seems rather... hmm... varying...

I ran it twice, Intel Matrix Storage Manager - 8.9.4.1004 - once just "like that" then after a manual TRIM via the toolbox.

(and it does detect my AHCI correctly)



Not sure why my sequential write dropped by 17MB/s after a manual TRIM...

 

Ok, let me try that
Thanks

Edit:
Looks more like it on the read speed.

 

Hey everyone! Merry Christmas and Happy Holidays to one and all!


sgilmore (and others responding to this topic),

I agree that first and foremost the most important thing is getting the drive optimized one way or another.

But, on a 'purer' and/or more theoretical level, I do not like GC at all, mostly because of what Anand said:

Quote:
" Presumably this isn’t without some impact to battery life in a notebook. Furthermore, it’s impossible to tell what impact this has on the lifespan of the drive. If a drive is simply reorganizing data on the fly into a better (higher performing) state, that’s a lot of reads and writes when you’re doing nothing at all. And unfortunately, there’s no way to switch it off.

While Indilinx is following in Samsung's footsteps with enabling idle garbage collection, I believe it's a mistake. Personally, real TRIM support (or at least the wiper tool) is the way to go and it sounds like we’ll be getting it for most if not all of these SSDs in the next couple of months. Idle garbage collection worries me."

From this article (linked to the page I took the above quote from):
See:
http://www.anandtech.com/storage/showdoc.aspx?i=3631&p=14



I know, we shouldn't be worried about 'extra writes' to our SSD's - we will probably upgrade them way before we hit any life-expectancy or capacity reducing limits show up. However, I would still like to limit as much as possible the writes to the current drives - GC takes this out of our hands and effectively 'uses' the drive up for us.

Maybe, (if I was being cynical - but not on Christmas Day! ) the manufacturers implemented GC like this knowingly, so current SSD users had to upgrade in the 'nearer' future?

Anyway, my thoughts on GC and why O/S initiated TRIM is the only SSD option I'll consider - with everything I've learned to now.

Cheers!

 

Les,

Your original thread was what made me join NBR!

A little confession: I thought the SSD thread was NBR for a long time.

Merry Christmas to you and yours as well.

Cheers!

 

lol

The controller is what defines the SSD. The memory used is just 'dumb' flash. I would not be surprised that Intel managed that, and, that they are taking the next year to test them before releasing them to the general public (unlike some companies we know of...).

 

In a word? Controller.

With a relatively unsophisticated controller, even in a fully used state, the controller is not tuned (nor capable) to keep up even to the degraded performance levels when pushed very hard - like with IOMeter scenarios.

The number of channels plays a part. The fact that DATA is cached, instead of Intel's method of caching the management of the nand chips, plays a part.

Most importantly though, I feel that Samsung set it's SSD target sight too low. They aimed at mechanical HD's and although in benchmarks they do outperform them considerably, for people like myself (and other's who are used to fast storage subsystems), they do not epitomize what SSD's are all about.

Except, in the very important area of low power consumption which they still rule (even over King of SSD's Intel).

 

OK... force installed 9.5.4.1001
...not sure what to make from the results...



Edit - and that one missing value - read access time... it returned an error?

 

you could try to close other problem to run again, i got same error once when i have command prompt window opened, but the results seem not encouraging, you could try 9.5.0 which in my case has better numbers sometime than 9.5.4.

 

Hmmm...

Have to agree 'not sure what to make of the results'. (Compared to your post with the 8.9.4 drivers).

Can you run 'winsat disk' and see compare them to the 8.9.4 driver? I think you may find the latency on many of the tests to have gone down a bit?

As for the read access time - too fast for the programs 'expected' values?