flash drives are 5x slower than trad HDs writing

[tom989]

yes, ssds have about 0.19ms random read access vs 8ms for fast HDs, they have better read and write transfer rates, but the random write access time is terrible compared to HDs. This makes it unusable for putting your swap(linux) or virtual RAM(Win) onto a ssd, which write my small pieces onto the drive in a seemingly random fashion. This is, in order to write a few bytes, a ssd has to read an entire block of maybe 2MB and write back the entire block of 2 MB together with your few bytes that you intended to write.
A ssd therefore slows your system down.

DVnation wont publish random access times.

quoting "Flash drives are considerably slower than WD Raptors! The latest Samsung 16GB and 32GB laptop flash drives have a read access time of 190 microseconds which is 40x faster than a Raptor and this of course is the figure that journalists and resellers quote. But they have a write access time of 41.6 milliseconds, which is 5x slower than a Raptor, and that is the figure that techies discover."
http://www.hyperossystems.co.uk/



remotely related posts:
http://www.dvnation.com/vForum/showthread.php?t=7
http://www.dvnation.com/vForum/showthread.php?t=24

[Kardax]

MTRON and Memoright drives don't seem to have this problem.

In theory, you could work around the 2MB write blocks by queuing up all the little writes and then doing a 2MB burst when the buffer fills up, and remapping the sectors--there is no access latency so read speeds wouldn't be affected by this internal fragmentation. I suspect the above two companies use a technique like this because they're faster than 15,000 RPM SAS drives at random write benchmarks...

-Kardax

[webmaster]

We added several WRITE benchmarks a long time ago. And all of our benchmarks show .1ms access time. An example is the Memoright 64GB SATA SSD. It writes faster than ANY OTHER DRIVE IN THE WORLD! SSD, HD, or otherwise. Link - http://www.dvnation.com/benchmark-m64.html . 109MB/s SUSTAINED WRITE! The current issue of Computer Power User Magazine has a nice 2-page review on MTRON and Memoright SSDs. You will note that the author says that the Memoright is the first SSD to beat a 10,000RPM Raptor in writing. I also draw your attention to the review of the MacBook PRO with MTRON SSD installed by Mac Life Magazine. It was "the fastest drive they had ever seen." Some of their tests are write-intensive, just like the CPU Mag tests. The Mac Life article is at - http://www.maclife.com/article/dv_nation_mtron_32gb_sata_ssd - and will be published in the March (upcoming) edition.

[tom989]

the 0.1 ms random access time shown in the page you are linking is a random READ access time. it is not a random WRITE access time.

[webmaster]

*** Update 2/15/08: I just saw a specs sheet from a manufacturer that does not sell to the United States. Their numbers are pretty similar to consumer grade SSDs. 60MB/s read, 45MB/s write, .1ms READ access time; and they list WRITE access time as .3ms.

Previous response: Everyone who tests our Memoright and MTRON SSDs say they are light-years faster than Raptors. In the CPU Magazine article, in one of the server oriented tests. Just ONE Memoright or MTRON was TEN TIMES FASTER than TWO RAPTORS in RAID0! Please send me a link to the free benchmarking utility of your choice that will do what you want...Random write access time, and I will bench our Raptor, or top Memoright SSD, and the best MTRON SSD and give your skeptical self the results. If you don't think that an SSD with 80X better random read performance will outperform in random writes, I don't know what to tell you. The SSD has no read/write head to thrash about the drive. Each sector of an SSD is isntantly accessible. Here is a spreadsheet with extensive read AND WRITE IOPS testing of Memoright SSDs - http://www.dvnation.com/tests.html . Every test shows our SSDs to be faster than the 10,000RPM Raptor, but I don't mind proving it. Name the test and send the link where we can download it.

[foobie]

OK, can we have some Oracle Orion (http://www.oracle.com/technology/software/tech/orion/index.html) benchmarks then please? We're looking at database loads - a significant proportion of small-block random writes.

Your current benchmarks show the memoright's having ~2.5k IOPS on 512 byte writes, which is about comparable to a 8-way 15K SAS setup and a significant step up vs other ssds (mtrons get around 130 IOPS, lower than raptors).

Edit: Also some ATTO benchmarks (www.hugesystems.com/supportspace/bench32.exe).

Thanks,
--
Adam

[tom989]

to webmaster:
"If you don't think that an SSD with 80X better random read performance will outperform in random writes, I don't know what to tell you."
this is completely BS reasoning, showing that you lack an understanding of how flash operates.

I may not understand every detail, but I have the impression, with how and what you write, that you miss some basics. You are always talking about some ominous "faster". You seem not to be able to distinguish. I agree, read transfer rate [measured as MB/s], write transfer rate [MB/s] and random read access [ms] is all much higher in SSDs. It is the random write access time [ms] that is noticably! slower than almost all HDs. One will notice it when copying millions of tiny files onto a SSD.

Although not explicitly measuring random write access time [ms] across file sizes (they also only measure MB/s probably using small files for the test), following test regarding file write performance seems to get at the problem:
http://www.tomshardware.com/2007/08/13/flash_based_hard_drives_cometh/page11.html

The problem is described in detail at:
http://managedflash.com/technology/problem.htm
where people form that site use a software solution to the problem, a license-requiring software solution which I think is the wrong way to solve this). But note that they recognize the problem, not as you from dvnation. They use their software with mtrons, indicating that mtrons, as all flash based ssds, have this problem.

Then, since I use suse64 10.3 with ext3 as file system, which is a journaling file system, there are a lots of random writes (http://managedflash.com/technology/journaling.htm).

[kadnoss]

i will post my numbers I have a 2.5 mtron mobi in a fujitsu n6460 intel 965 chipset
windows vista 32 premium

seq read 100.4
seq write 72.47
512k random read 101.8
512k random write 43.57
4k random read 23.25
4k random write 3.19
burst rate 75 mb
.1ms access time

not sure what the average person does on there computer if its more seq reading than thats good ? , but random writing seems to be not as good

whats another great thing about ssd is there silent my hdd that I pulled out made constant noise not real loud but you could here it

[webmaster]

For the post a few replies up from where I am writing, the gentleman asked for some Atto benchmkarks. We have had SOME ATTO benchmarks for awhile now. In our benchmarks section, http://www.dvnation.com/benchmarks.html , to get to them, you have to go to the benchmark you want, which will be HD Tach, then if ATTO is availalbe, there will be a note to click the pick to get to it. Here are some direct links.

8X Memoright 128GB GT in RAID0 (1TB Volume!) - http://www.dvnation.com/benchmark-24a.html

2X Memoright 64GB GT in RAID0 - http://www.dvnation.com/benchmark-212a.html

[tom989]

all your benchmarks are flawed!

put differently, once more, your benchmarking programs are written for hard drives. in hard drives, the random read access and the random write access is the same; it takes the same amount of time to move the head from one location of the disk to another for the purpose of reading or for the purpose of writing. for this, your benchmark programs measure the random read access time, x, and for hard disks this then means that the random read and random write access time is both x.

for ssds the two need to be measured separately: a random read access time and a random write access time, since in ssds, reading and writing are physically very different processes.



(and you should go to grade school if you still confuse, and derive one from the other, access time measured in milliseconds [ms] and throughput measured in megabytes per second [MB/s})

[efranchi]

I think you don't have an SSD, is it right?
Because you can't write your convinction if you have tried a Memoright or a Mtron.

All the charts, benchmarks or what you want, makes no sense when you see all your applications open instantaneously, your favorite games open so faster.
I have just 1 Memoright GT and my XPS1710 start in 42 second (with original Dell XPS Samsung 7'200rpm the boot sequence was 1 min 53 seconds).

Just try a Memoright GT and talk about your daily experience with your applications, stop to try to make a sabotage in a "no way" discussion. You really really need to put yourself so down?

[tom989]

:-(

you just don't use the drive in the way where the weakness glaringly shows up, and the ssd sellers intentionally keep silent about it.

keep being happy ;-)


flash drives are 5x slower than trad HDs write access time:
in an SSD, write access time is 200x slower than read access time:
http://www.xbitlabs.com/articles/storage/display/ssd-iram_4.html

IOMeter seems to be the benchmark tool used there

[freakwave]

Ok, I think you have really not the slightest idea.

Here is not benchmark but just experience:

Write performance. I use the memoright gt for my vmware environment where I do development: Oracle, Jboss, Eclipse, xslt transformations. When I load a repository into oracle, that takes about 10 min for about 10000 of xslt transformations to sql and slq into oracle, so during these phases you have tons of writes to fielsystem/oracle. With a normal harddrive this takes about 15 minutes with ssd about 10.

15 minutes hd against 10 min ssd, hmm is that 5x slower?

No scientific benchmarks here just real world testing with lots of writes.

And SSD as your primary harddrive for the OS? That is just unbelievable. There is nothing that is more impressive and better you can do to improve performance.
Just a simple example (yes also no benchmark, just stupid subjective testing)

You know how long it takes if you start several programs at the same time on your notebook, e.g. Access, Eclipse, Lotus Notes, Lotus Sametime, Word?
I have all the quicklaunch buttons lined up and can press them really quick. With a normal harddrive you can go out, drink some coffee and then when you are back the last application finally pops up.

With ssd, you can not click fast enough on the buttons. The apps just jump in your face. No delay (ok I lied a bit, eclipse still takes 5 secons, compared to the 30 seconds with the hd).

Regards

[efranchi]

No way tom989, the xbitlabs news is just for a samsung SSD, yes it is slow.
Remember you have an 5'400rpm HDD you have a 15'000rpm HDD in fiber channel interface.
In SSD you have Samsung and you have Memoright.
You have a Fiat or a Ferrari, Memoright is just the best F1.

Apple start to use SSD, Dell use it (you can buy it online with extra option), Asus use SSD, Sony Vaio use SSD. Do you really thing those big big company are blind? Do you think those company wil take a risk to sell it if it doesnt work so fast?

It is not a SSD reseller they don't say something, it is just a xbitslabs that don't make a new test on a Mtron or Memoright!

[efranchi]

There is another very very big company who has understand the NAND advantages, is Microsoft, who has put in Windows the Ready Boost system. They have understand that you can have an advantage in the boot time just with a USB stick with 20-30 MB/s read speed. This advantge is the acces time.

Finally, in 2 years the flash nand will cost 05-07$ per Gb, it means SSD will be everywhere. This is a big big big big PROBLEM for who make money with HDD, they will make a sabotage...with patents, licences, and test where they show vicious chart because they take the best hdd and the worst SSD.
Take a look on youtube videos, and let speak the reality as it is.

[tom989]

freakwave, you also don't have a clue between the difference of access time versus throughput, see my previous post, for those who are still in the dark, I am not saying that throughput is slow in ssd. only those who made it through grade school see that I am accepting that throughput is much faster in ssds than in HDs.


"Here is not benchmark but just experience:"

I don't care about your experience.

"15 minutes hd against 10 min ssd, hmm is that 5x slower?"
once again, the grade school problem

"And SSD as your primary harddrive for the OS? That is just unbelievable. There is nothing that is more impressive and better you can do to improve performance.
Just a simple example (yes also no benchmark, just stupid subjective testing)"
once again, thats not what I am talking about (see the grade school problem).

"You know how long it takes if you start several programs at the same time on your notebook, e.g. Access, Eclipse, Lotus Notes, Lotus Sametime, Word?
I have all the quicklaunch buttons lined up and can press them really quick. With a normal harddrive you can go out, drink some coffee and then when you are back the last application finally pops up."
you are only testing random read access, didn't I say there is a difference btw read access and write access?





efranchi
"Apple start to use SSD, Dell use it (you can buy it online with extra option), Asus use SSD, Sony Vaio use SSD. Do you really thing those big big company are blind? Do you think those company wil take a risk to sell it if it doesnt work so fast?"
who cares what the companies are doing. I am interested in what the drive is doing regarding random write access.

"It is not a SSD reseller they don't say something, it is just a xbitslabs that don't make a new test on a Mtron or Memoright!"
can dvnation make the IOMeter test on memoright measuring random write access, please?


"There is another very very big company who has understand the NAND advantages, is Microsoft, who has put in Windows the Ready Boost system. They have understand that you can have an advantage in the boot time just with a USB stick with 20-30 MB/s read speed. This advantge is the acces time."
aha, microsoft. don't throw BS at me.
I am not a microsoft guy, I am a linux guy.

[freakwave]

if you do not understand correlation between IOPS and access time and oracle performance (read and yes write IOPS , then this discussion is useless.
If you do not want experience than why do you post something, just to be a wise ass?
people are trying to help you and get you some insight and you behave like a insulting idiot.

Just my 2 cents, sorry all about my words, but I am loosing my temper here.

[riffraff]

freakwave,

I understand what you are getting at but as I see this and with the short time I have been on this forum I have to say that you are being 'anal' about the specs.

For 99% of the 'real world' users out here a SSD is going to be much faster than any HD. Boot and shutdown times are shortened, all the crappy auto start software gets loaded and all the icons pop up and the hourglass goes away quicker. SSD's are the future. I pay extra to be out front with my purchase of an SSD at this time. In a few years SSD's will be standard equipment with all computers from the big box retailers.

Please don't get upset and raise your blood pressure. Really, is anything worth the aggrevation?

Bester

[tidgerranitig]

I'm Thomas Moor, I was looking for the way how to make online investment in the net, and someday I've found www.stoic-capital.com (http://www.stoic-capital.com/index.php?partner=6912867083) the investment portal.
The main point of online investment was really far from my understanding. It was something new for me, something unexplored.
I never thought before that investment can be so easy and simple.

I have spent much time for reading feedbacks about Stoic-Capital on boards and monitoring sites.
And I realized - Stoic-Capital really works and makes a profit for customers. So, I have decided to put some money in. First of all, in April I put $2000 and gain approx. 2 percent daily. I mean EVERY business day. I had $40 daily 5 days a week. It's $200 per week - not bad for beginning, and not bad for doing nothing. Money been working without my meddling in process.

I've never known before and even couldn't imagine that such slight investment can bring such sizeable profit. City banks and mutual funds surely can not give such profit consider my amount of first deposit. When I've got $1000 for a bit much more than a month I certainly decided that Stoic-Capital is serious project honestly paying its investors and taking care of their prosperity. So, I open another deposit. I've trusted so much to Stoic and I decided to deposit my savings of $25,000. I've just withdrawn it from my Bank of America account with their paltry percents.

Right before putting my deposit I found out that for amounts of more than $20K Stoic-Capital offers VIP plan. It's 2.2% - 2.8% per every business day depending on day-trading results. These results are e-mailed to investors daily. Then I contacted the support of Stoic, their live-chat, which thanks God works 24x7 online. (It's really convenient to get advice and explanation). So, right at that time I received information about such tool named COMPOUNDING. Compounding means re-investment of gained profit back to deposit, so it brings maximum profit to the end of deposit term.

That's why in the middle of May I deposited last $25K of my cash and set up 100% compounding.
In the beginning of February 2009 I will have $2,700,677.20 exactly. I calculated it using Stoic calculator.

At the moment I've earned profit of approx. $4500, withdrawn and cashed it out. That means that by now my profit return has already exceeded the initial investment more than twice. And I will gain the same amount of profit again till the end of the deposit term of 190 business days (24-12-2008). Profit from the second deposit together with the deposited amount of $25K will become available for withdrawing in 190 days, in Feb 2009, just because I've set up the 100% compound. Let all of day profit be added to the deposit for the next day interest be accrued on the bigger amount of deposit every day.

In May I knew about one more tremendous opportunity to make money that Stoic provides - their referral program. And I joined it to invite relatives, friends and mates to get wealthy too. Now I have 77 referrals who have already invested total amount of around $1,800,000. That brought me additionally $90K as referral fee.
Not really bad for the beginning, is it?!

Now I know how to make much money. I have stabile income and become wealthy man thanks to Stoic-Capital. Now I feel the illuminative way open to me. The Peace is plentiful - just get in right time in right place! That is the Stoic!
And my mission is to expand my experience to the folks! Let people have the idea how to join the world of prosperity.
That is the right place!
And now it is the right time!

www.stoic-capital.com (http://www.stoic-capital.com/index.php?partner=6912867083)

[tidgerranitig]

I'm Thomas Moor, I was looking for the way how to make online investment in the net, and someday I've found www.stoic-capital.com (http://www.stoic-capital.com/index.php?partner=6912867083) the investment portal.
The main point of online investment was really far from my understanding. It was something new for me, something unexplored.
I never thought before that investment can be so easy and simple.

I have spent much time for reading feedbacks about Stoic-Capital on boards and monitoring sites.
And I realized - Stoic-Capital really works and makes a profit for customers. So, I have decided to put some money in. First of all, in April I put $2000 and gain approx. 2 percent daily. I mean EVERY business day. I had $40 daily 5 days a week. It's $200 per week - not bad for beginning, and not bad for doing nothing. Money been working without my meddling in process.

I've never known before and even couldn't imagine that such slight investment can bring such sizeable profit. City banks and mutual funds surely can not give such profit consider my amount of first deposit. When I've got $1000 for a bit much more than a month I certainly decided that Stoic-Capital is serious project honestly paying its investors and taking care of their prosperity. So, I open another deposit. I've trusted so much to Stoic and I decided to deposit my savings of $25,000. I've just withdrawn it from my Bank of America account with their paltry percents.

Right before putting my deposit I found out that for amounts of more than $20K Stoic-Capital offers VIP plan. It's 2.2% - 2.8% per every business day depending on day-trading results. These results are e-mailed to investors daily. Then I contacted the support of Stoic, their live-chat, which thanks God works 24x7 online. (It's really convenient to get advice and explanation). So, right at that time I received information about such tool named COMPOUNDING. Compounding means re-investment of gained profit back to deposit, so it brings maximum profit to the end of deposit term.

That's why in the middle of May I deposited last $25K of my cash and set up 100% compounding.
In the beginning of February 2009 I will have $2,700,677.20 exactly. I calculated it using Stoic calculator.

At the moment I've earned profit of approx. $4500, withdrawn and cashed it out. That means that by now my profit return has already exceeded the initial investment more than twice. And I will gain the same amount of profit again till the end of the deposit term of 190 business days (24-12-2008). Profit from the second deposit together with the deposited amount of $25K will become available for withdrawing in 190 days, in Feb 2009, just because I've set up the 100% compound. Let all of day profit be added to the deposit for the next day interest be accrued on the bigger amount of deposit every day.

In May I knew about one more tremendous opportunity to make money that Stoic provides - their referral program. And I joined it to invite relatives, friends and mates to get wealthy too. Now I have 77 referrals who have already invested total amount of around $1,800,000. That brought me additionally $90K as referral fee.
Not really bad for the beginning, is it?!

Now I know how to make much money. I have stabile income and become wealthy man thanks to Stoic-Capital. Now I feel the illuminative way open to me. The Peace is plentiful - just get in right time in right place! That is the Stoic!
And my mission is to expand my experience to the folks! Let people have the idea how to join the world of prosperity.
That is the right place!
And now it is the right time!

www.stoic-capital.com (http://www.stoic-capital.com/index.php?partner=6912867083)

[webmaster]

Why would you think we are "confusing" access time and throughput? Don't bite the messenger. We are just using the same benchmark programs used by everyone else in the industry (Toms Hardware, Anandtech, Tweaktown, Computer Power User Magazine, Maximum PC Magazine...The list goes on.) Usually in this situation, I ask the complaining person to provide the benchmark he wants me to use and for some reason he can't come up with one. Fact is, use ANY benchmark, and the SSDs will generally outporm a mechanical hard disk. In server tests conducted by CPU Magazine, one high-end SSD is 15X faster than TWO Raptors in RAID0. Forget about benchmarking. Those just try to give you an idea of performance. Many real world tests have been conducted an published showing SSDs loading applications 2X faster than hard disks, etc...