Slow SSDs?

Using an Inferno 100GB SSD on Win 7 x64 with 8GB RAM and an i3 350 w/Optimus nVidia 310M platform, I too fail to see the difference the SSD has made on my (new) system compared to the exact same setup on my Win 7 x64 Core 2 Duo P8600 8GB with a Seagate Momentus XT hybrid drive except that it is quieter and hopefully more robust (re: bumps/vibrations).

As an example: merely surfing last night (about 2 hours) with all available updates applied hours earlier including a reboot (so they can't be used as a reason for it being so slow) the i3/Inferno based system took over 3 minutes to shutdown. Yeah - SSD do not rock.

SSD's, over the last (almost full) year that I have been testing/using/exposed to them have left me underwhelmed. When they're fast they cut seconds and even minutes off certain tasks, but when they're slow... you may as well be using floppy drives to drive the O/S.

I'm hoping that a real SSD with all of the benefits of current HD's and none of the drawbacks of all SSD's I've had the displeasure to use will be out before 2012 (and for a reasonable price too).

So far, I'm not holding my breath for that to happen - even in two years time.

Why? Because people have fallen hook line and sinker for the marketing the SSD makers and I'm guessing mostly the noise the reviewers (Anand...) are making about them. So what incentive (to the manufacturers) is there to make them actually better than the mediocre stuff we're seeing for the last three years or so?

Sure, I agree that Vista/Win 7 is needed for optimum performance out of this new storage medium, but if you can't feel it being faster than a mechanical HD - Always - then all we buy SSD's for is robustness and silent operation - not performance.

Two orders of magnitude better access times, insanely faster 4K random r/w's and other nonsense benchmarking means nothing when at the end of the day; the same amount of work gets done by 'old tech' mechanical HD's and 'bleeding edge' SSD's that not only cost 10x or more than the proven tech, but also just randomly 'die' or worse perform like 1990 hardware with 2025 prices.

I know that SSD's are the future; but... the future is not here yet.

 

Wrong. Not all update installations are finished until a full reboot of the machine is done.

 

Mugenski,

You quote me saying that I did a reboot after the updates - hours earlier.

What's your point?

 

Phil,

I agreee with your analysis (it's the laptop) on a surface level, but in my opinion it is not the laptop but SSD's in general.

They simply are not ready for prime time.

I broke down and and downloaded both Crystal Disk Mark and AS SSD benchmarking utilities and the results are horrible. And this is on a new i3 platform, clean Win7 x64 install with 8GB DDR3 RAM and the latest Intel Rapid Storage Technology driver.

In Crystal, I actually saw one pass of the 4K reads drop below 0.7 MB/s! Yeah, I'm so happy about this Inferno (not!).

Well, at least it confirms that it is not just my imagination - this (as all other SSD's I've tried) are dog slow in the end - with hiccups/pauses and inevitably longer times (compared to mechanical HD's) to do anything - even just reboot sometimes (3+ minutes).

The logical explanation is that TRIM is not working (obviously), but what hoops and walls of fire do I have to jump through to make one single SSD work in just one of my systems?


Zaz, I'm not trying to hijack your thread - just giving some more information to maybe 'prove' that it is not your laptop but the SSD's in general.

I have seen this effect on a Mac Book Pro, a couple of VAIO's, a Dell M6400 an Asus U30Jc and on a couple of desktops too. All used with a mix of Intel, Patriot, Samsung and Vertex SSD's (basically the top 5 controllers in the SSD world).

In each case, loaded with my normal programs (about 60GB), a clean install of Windows 7 x64 and just a subset of the data I would normally use, the SSD's were no better than regular HD's in actual use and on the desktops they faired much worse than my Raptor's - eventually.

What it comes down to is that SSD's need to be babied too much at this point to make them worth my while - I may sell the Inferno off and go back to my Seagate XT or even my 7K500 if the drive continues to slow down with normal (and on this system, very light) use.

The worst part of the Inferno is that there is no utility (from Patriot) to do a manual TRIM to see if the performance can be restored to normal.

Remember, I don't care what the benchmarks say - the drive seems slow to me and it doesn't matter if one metric or another is 1000 x faster than a mechanical drive. Even when it was brand new and I had simply installed Win 7 and the drivers on it, I could see it get slower with each reboot and each program I installed, but this is the first time I benchmarked it to see if it was 'normal' for an SandForce controller or not.

This is the FAIL of SSD's for me: normal use will eventually get you below the performance of mechanical HD's. And, you get the pleasure of paying a premium for that honour too!

Anyway, here are my benchmarks and any more information you or anybody else can shed on this problem is much appreciated:

Edit: Added IOPS screen and copy screen benchmarks from AS SSD.

 

Tiller your sequential speeds and 4K random reads look way too low too.

Are you running Intel Rapid Storage driver?

FYI: I've never noticed any of the hiccups/pauses you're talking about. Both my C300 and Vertex LE do everything much faster than my 7K500 did.

 

tilleroftheearth, what chipset do you have?

 

Please look at my updated post (a couple of more screen shots).

Phil and stamatisx,

I have the Asus U30Jc i3 with Optimus Graphics and 8GB RAM. Clean Win7 x64 install and the IRST drivers (9.6).

If this is the only system I had seen this on (SSD getting slower just using it) I would agree its the system, but this is standard on every SSD based system I have used so far. Mechanical HD's, while not ultimately as fast (when the planets are all aligned for SSD's) are much more consistent in their performance with my usage.

Thanks for trying to help.

 

12-14MB/sec 4K random read is not what it should be performing. Most reviews show over 20MB/sec.

I don't think it's your usage, it's your system.

If an SSD is not doing everything faster than your HDD, there's a problem in the setup.

 

Tiller I don't know it that will make you feel any better but I have the Intel X25-E 64GB and we have the exact same 4K random reads/writes. My SSD costed me $800 and uses SLC based NANDs, which means that I don't need TRIM since performance stabilizes after a while and remain the same no matter what. Another thing we have in common except from the very low 4K random reads/writes is the Intel based chipset which in your case is the HM55 and in my case the PM55.
I am afraid that there is a pattern here with the Intel based chipsets and the 4K performance.
The exact same SSD on a laptop with nvidia chipset performs as advertised or even better.

*EDIT*
I still need to find out if the BIOS is responsible for that, or this performance hit is due to the power saving features of laptops (implemented in the BIOS by each manufacturer)

 

That's kind of ironic isn't it? Intel chipset, Intel drivers, Intel SSD = poor performance?

I'm a bit on side with Tilleroftheearth though. I bought an Intel 80GB G2 for my gaming desktop, coupled with a WD 1TB Caviar Black. I was impressed by the responsiveness at first. Now there's a handful of apps that take 5-10 seconds to load, when they start up immediately on my sager with 500GB Seagate Momentus (not XT). Yes, latest drivers, TRIM enabled, aligned, etc. Responsiveness isn't there like it was when it was new either.

I opted for an SSD in my netbook and notebook mainly because mechanical drives tend to put one or the other or both of my hands to sleep. Sager isn't too bad, the M11x I had was horrible. I couldn't type more than ten minutes and my left hand just went numb. Problem is I need lots of storage but am kind of regretting spending the money on the 256GB Samsung that's in my Envy. It's not much faster than the Seagate Momentus in my Sager.

It seems too many people are afraid to say SSD's aren't worth their price at all. Initial use is impressive, but after that, it's all downhill, performance issues, and worrying about TRIM or background GC routines. These are most likely the growing pains of solid state storage, but more people use them, the less impressed they will get.

I've decided not to worry about benchmarks. If my system is acting sluggish then maybe I'll run some to see what's going on. Otherwise, just use your machine and don't worry about it.

 

Thanks guys for your responses!

When I posted a year ago that I was less than impressed with my first SSD encounter in a Dell M6400 (Samsung SSD) I was met with 'are you insane?' comments (well, not literally, but you get the point).

In the year since, I've played/used/tried them extensively I am no more impressed even though the benchmarks have gone up markedly.

Phil, it is not my system - as I mentioned this is on every system I have tried it on - it is the SSD's (all of them).

stamatisx, I'm sorry to hear that even an SLC drive like the (still) highly regarded X25-E gives similar performance - I've always wondered if just 'getting the best' would solve these issues but it seems not. Thanks for sharing that.

htwingnut, just to clarify - I'm currently using an SandForce based Inferno and getting the same results as you (general slowdown). But keep in mind that the other systems had a mix of chipsets and SSD controllers thrown at them and they all suffered too (eventually - about a week of casual use or a day or two of production type work).

So the rememdy is to use my mechanical HD's then?

Okay, but don't tell the Inferno yet...

 

If "the SSD's were no better than regular HD's in actual use" there's something wrong. They should be faster at every task bottle necked by the storage subsystem. Every real world benchmark will confirm this.

Now you may find that the performance benefit doesn't justify the costs, which is a point of view I can understand. Yet there is a performance benefit. In booting, launching applications and all tasks bottlenecked by the storage subsystem.

And yes I agree that the Momentus XT gives near SSD experience.

 

I have another suggestion. Take out the good old stopwatch and measure the difference between the 7K500 and the Inferno.

Boot your system, duplicate a folder with a lot of files, launch a couple of heavy applications.

The Inferno should outperform the 7K500 by a large margin.

Maybe try the 250GB. I haven't heard if it does any better though.

 

Phil,

I kindly point out that bencmarks are seconds, minutes or at most a couple of hours and they're done.

Actual use (with an actual O/S and software installation) is another story altogether and no benchmark can replicate/predict that scenario that I have seen.

The Momentus XT is amazing - actually, it is sometimes even faster (in a lower-end (notebook) platform too) than my Raptors I have. But overall, the Raptors are still the standards that SSD's have to beat before they get replaced - not for pure 'snap' but for an actual productivity increase.

To briefly compare the XT to the Inferno with a recent (small 16MB) CS5 update just the other day:

DL=Download, Ins=Install

SSD: DL:1min, Ins:4min - Total 5min

XT: DL:1min, Ins:40sec - Total: 1:40 mm:ss


Phil, not only is 12-14MB/s Random Reads below what the drive 'should' do - when I saw it hit less than 0.7MB/s on one of the CrystalDiskMark passes this explains why the Seagate is faster than the Inferno.

To put this in perspective, the Raptors are consistently much faster (3x?) in 4K Random reads, than the low (oddball) score the Inferno displayed - but all it takes is one low read to allow a mechanical (or an Hybrid, like the XT) to destroy the supposed superiority the SSD's have.

Either way, I have yet to see any system that has continued to impressed me with an SSD running the O/S and programs over time.

Hope this changes soon.


Edit: Just saw your new post Phil:


I did compare the differences and yes, initially the Inferno was better - that is why it did not get returned. But now, I would bet on the 7K500 to be faster or at least as fast. To justify the price of the SSD then is impossible!



Comparing the Inferno to the XT a bit more:

Boot up (i3) SSD: 50secs

Boot up (p8400) XT: 55secs

Launch PS CS5 (i3) SSD: 12-16secs (cold boot)

Launch PS CS5 (P8400) XT: 5-8secs (cold boot)

Launch LightRoom 3.0 SSD: 11secs (cold boot)

Launch LightRoom 3.0 XT: 4secs (cold boot)

 

So your SSD took 4 minutes to install something that your XT did in 40 seconds. Everything you are posting is pointing to the fact that there's probably something wrong with your SSD or system.

I recently compared the real world performance of two Sandforce SSDs, C300, WD Scorpio Black 500GB and Momentus XT in my own notebook. Installing CS4 was one of the jobs. I will not leak the results before it is published but as you can understand the results were completely different to what you're getting.

 

Sure, comparing is swift and easy - one wins the other doesn't.

I'm sure your results will be different than mine - but... how long did you use the computer after you tested each install?

That seems to be the theme here: long term use makes all SSD's slow down.

 

My Vertex LE had been used for six months. Performance stayed consistent at all times. Like it should be.

 

Okay.

In what usage scenario?

 

Average usage, nothing special.

Do you have any sources that confirm this for TRIM enabled SSDs: "long term use makes all SSD's slow down" ?

 

Sources?

I have at least me and two or three other posters that confirmed this in this thread.

Well, my 'average use' is a full install of all programs I use - monitor calibration software - updates of any and all programs as soon as they are available (and tested on my dedicated testing desktop system) in addition to using 4-20 programs simultaneously (PS CS5, LR3.0, Bibble5.1, Illustrator CS5, Outlook 2010, WMP 12, Capture One, Word/Excel 2010, etc...). The O/S and sofware install is around 60-80GB depending how many sample files I keep on the system and my systems are standardized at 8GB RAM and Win 7 x64.

On the Inferno, i3 based notebook, I don't actually have more than 3 or 4 programs running at once (screen too small, etc.) so if anything, it is being used less here than on my Raptor based desktops where I also tried it (and saw no performance increase over the 4 Raptors).

So what is your 'average use', specifically?

 

If modern TRIM enabled SSDs would really loose their performance over time it should be easy verifiable. I would think a major techsite would have found it out by now but maybe you can inform them.

My 2 cents: if your SSD is performing like it should (which clearly isn't the case) it will blow your 7K500 out of the water at every task bottled necked by the storage sub system. This is also what all the reviews done by major tech sites will confirm.

 

Frankly, I wouldn't worry too much about transfer rate. It certainly will vary from one case to another. The only thing that should be constant for SSDs is the snappiness. If it's not snappy, then something is amiss.

 

f4ding,

I agree - I don't care about absolute transfer rates - I only care in relation to my last/old drive.

The snappiness though - that is what everyone is touting about on SSD's!

But the reality is: in the last year, over a dozen systems and 5 SSD's later and still not one SSD based system for me to say 'that is worth it'.

Yeah, something is amiss - but the only thing common to these systems is that they had an immature SSD in it.

Pretty sad.

 

I don't know how easy it would be to verify, but I do know that all 'techsites' have a vested interest in promoting the products they are supposed to be 'unbiasedly' reviewing - not slamming them.

With your interest in concurrent benchmarking of so many different SSD's, why don't you set them up for long term use (I'm sure a month would be enough) and see how they perform compared to the same computer using a mechanical HD?

Don't give me before/after benchmarks - just (really) use them and then go back to the mechanical HD based systems and simply do the same stuff to both and see which is performing better then. That is what I have done and each time the mechanical HD has always been the better alternative.

You still didn't mention what your average use is though, so I'll assume it is very, very low intensity (to the SSD). Like I mentioned elsewhere, when I first install Win 7 on an SSD I am left sitting there with a silly grin. But that soon wears off as I install more and more apps that are needed to take the raw hardware and transform it into a tool I can use to make money with.

As to the point you make that every tech site confirms the superiority of SSD's, well I believed them enough to buy one that fooled me just long enough to miss the return period - they are doing their job.

But that doesn't change the fact that what they report is nowhere close to what I have seen in my own and others systems this past year.

I can see hints and glimpses of the promise of SSD's - but they disappear soon after (very soon after) installing the programs that make my computers usable.

 

I wonder which 5 SSDs you had that were not worth it for you.

I've had an old school Samsung with 100mb/s read with no TRIM in my system for nearly two years without a slow down in real world usage. It was a 32GB and was filled to about 28GB with XP, AutoCAD, PS 2, Illustrator, Rhino,..... and I used it heavily for architecture school. After all the usage I felt no slow down in the system over time and this is taking into account I bought the SSD USED, so I'm not sure what kind of usage the owner did before.
The boot up time stayed the same, opening up programs stayed the same. I did bench it with ATTO or CrystalDiskMark and the write number did drop but very little.

My point is that there are some early SSDs and maybe current SSDs that were horrible but they have matured and should not degrade as before.

I replaced the Samsung with an Intel, about a year ago and I notice no lost of performance. And I recently got a Corsair Nova which so far has been excellent too.

 

jessea510,

All the SSD's I've tried are not worth it. I didn't say I tried only 5 of them; I've tried all the top 5 controllers.

I'm glad that your systems do not exhibit these symptoms, but I can't help but wonder if you simply do not know what a fast mechanical HD was before you got your SSD's?

When I tested the Intel G2 (finally - it took a while to get a hold of one), it was out of the computer before the final piece of software was installed (had it two days). I was stunned that this SSD could perform so badly after waiting so long to try it. When I got the Inferno - the performance increase was very noticeable over the Intel G2 - for about two weeks. sigh... just past the point of being able to return it.

In a desktop setting the differences were even more apparent against 4 Raptors (not in RAID).

If I simply open/close programs and also shutdown and bootup my computer constantly all day then I agree the SSD's are slightly better - but I actually need to process a few hundred to a few thousand image files in-between and in that scenario the SSD's add not one lick of performance compared to the Raptor based desktops and even to the Momentus XT based notebooks. Worse, they perform badly in inane tasks like updating programs (which is where they should excel at).

I'm curious how much you push your computers? I'm on my 'digital notebook' Inferno based computer for at least 12 hours a day - this does not include any time I might spend on a workstation to do any 'real' work to process the images I shoot daily.

 

SSD's in general are not like HDD's as far as not caring if you fill them up. The C300 seems to be the first one I've seen that is giving a glimmer of hope for high usage of the storage area and still giving close to new performance.

The Sandforce 2 has a perfomance limiter with excessive writes. This is KNOWN as "life time throttling" and can be seen in SMART values if active. I am not sure of exactly why others slow down but it seems there are various reasons.

For O/S's with other than TRIM support I'd say just get a Hybrid and be done with it. Windows 7 and other TRIM enabled O/S's are fine with SSD's as a primary device. With a SF2 you may have to wait a while after all the installs for the LTT to disable before seeing your numbers back to normal.

I can say my Callisto DX 120GB at 25% rocks without the LTT enabled but I've had to modify my workflow to avoid writes to the SSD. Once more realistic FW addresses this I will be able to use it normally, at least this is my hope.

So in the end, other than the C300, you have to look at the drives intended workflow. If you are doing 99% reads and 1% writes SSD's are for you. If you plan on heavy writes get a C300, hybrid or HDD. Other than the C300 I right now wouldn't even consider for myself a SSD for my single bay U81a. So the hybrid will stay there for now.

As always you have to take the hardware eviews with a grain or more of salt. Opinions and agendas are almost always hidden between the lines. If I'd have known of the LTT the SF2 would have never made it to my laptop. But this is what I have for now and have to make good of it. Hindsight is almost always 20/20..................

 

TANWare,

'Life Time Throttling', huh?

Thanks! This little bit of knowledge you gave me brought me back to this side of senile too!

Like I mentioned before, I don't care about the numbers (benchmarks), I just care to have the performance of at least the 7K500, if not the Momentus XT.

I'll be leaving this computer on for the next few days and see if any snap returns to it. (Having just read that SF controllers are notoriously slow at actually performing a TRIM command even when left alone for hours).

If this 'trick' works - I'm uninstalling this Inferno from the notebook and will be putting it into a desktop instead (I can't leave the notebook running all the time like I can a desktop).

Thanks again and +rep!

 

tilleroftheearth - the posts you are making in this thread (and elsewhere) are those of a troll. I'm surprised Phil took the bait.

The comments you are making are subjective and not based on anything factual; I'm tempted to delete them because they are borderline false information. What you are seeing with your SSDs is either a figure of your imagination, something you are saying because you don't like SSDs (and want to spread it on others), or a combination of both. Either way, I don't want to see any more comments from you about the topic in this thread or elsewhere. If I do, I'll ban you in less time it takes an SSD to read a 512kb file.

 

Chaz, obviously I default to your troll-o-meter, since it's the most powerful within The Force.

However, I have to say that while SSD benchmarks do show significant performance, a lot of the day to day stuff is quite subjective. Sometimes it's difficult to assess the real world difference unless you use two identical machines side by side doing the same tasks.

I do thik a lot of it is perception. While an SSD's performance may degrade slightly over time, the more you use it, the more you're used to its performance and may not seem quite as snappy, that is unless you go back to using an HDD for an extended period again. Same thing for HD video. First time I saw it, I was like "Holy COW!" it was awesome, not it's not so much. But if I go back and watch video in SD, forget it. It's like watching through a murky window.

Anyhow, my $0.02. That's all I'll say. I've been following this thread and got a lot of useful info. But I also see a lot of people worrying too much about benchmarks than the drives actual real word usefulness.

 

I agree with you, htwingnut; the difference with an SSD is a matter of perception. However, comments like this:

. . . are simply dribble. He is proceeding to impress his single poor experience with an SSD on everyone else in a subliminal and indirect manner. Sorry, that's trolling at its finest and we state in the forum rules that trolling is not allowed. Comments like the above help no one and waste time.
Your experience != experience of everyone else. Also, your experience is not a testament that everyone who thinks otherwise is wrong. Take your crusade somewhere else.

I'm going to quietly close this thread.

Edit: I removed these posts from another thread ( http://forum.notebookreview.com/har...arket-upgrades/507173-why-my-ssd-so-slow.html) and placed them in a separate one since the original thread was going fine until this stuff.

 

Phil,

I agree that an XT should be easily beaten, but that is not what my systems have shown.

 

If an SSD takes 4 minutes for a job that your HDD finishes in 40 seconds, I'm pretty sure there is some kind of problem.

I recommend you start a thread at the Patriot forums and get some help from the manufacturer.

And TRIM does work on Sandforce drives as explained here by Tony from OCZ:

http://www.ocztechnologyforum.com/f...trim-recover-speed-on-a-Sandforce-drive/page4