GullLars' SSD discussion/rant corner

**Anvil** · 05-27-2010, 09:43 AM

One Hertz,

On what controller?
NCQ?

**One_Hertz** · 05-27-2010, 09:54 AM

Originally Posted by Anvil

One Hertz,

On what controller?
NCQ?

Tried on ICH9R, ICH10R, LSI 9211. The results are all the same.

Yes, NCQ is on. Otherwise, my read results would suck as well and they do not. It seems like my NSQ is off only for writes? Makes no sense how that is possible to me.

**Anvil** · 05-27-2010, 10:28 AM

One Hertz,

Don't think that's possible wrt NCQ, either it's on or it's off.

I just did a 4KB rw IOmeter test on a single E on the 9211 ~40.000 iops (random write 4KB)
QD1 12740
QD2 22590
QD4 32803
QD8 38356
QD16 40748
QD32 40222
QD64 39396

Can't blame the 9211

, the ICH test you downloaded was from my UD7. (ICH10R)

**GullLars** · 05-27-2010, 10:36 AM

It COULD be you have turned off drive write cache... That could result in a major drop in random write performance, but i have no idea how much.

**~~onex~~** · 05-27-2010, 10:40 AM

Tried on ICH9R, ICH10R, LSI 9211. The results are all the same.
Yes, NCQ is on.

that's one of the weirdest things i heard,
maybe the MB is problematic..

so actually, it gives the same results with QD of (4KB random read) and QD of 32?
if it is, it is only NCQ/AHCI related, extremely doubtful if the drive is getting QD 1, less then 16K IOPs..
let's see.

**~~onex~~** · 05-27-2010, 10:40 AM

Tried on ICH9R, ICH10R, LSI 9211. The results are all the same.
Yes, NCQ is on.

that's one of the weirdest things i heard,
maybe the MB is problematic..

so actually, it gives the same results with QD of 1 (4KB random write) and QD of 32?
if it is, it is only NCQ/AHCI related, extremely doubtful if the drive is getting QD 1, less then 16K IOPs..
let's see.

**Anvil** · 05-27-2010, 10:43 AM

I disabled the "Write caching" Policy and iops are down to ~2300 so it can't be that. (~9MB/s)
(still on the 9211)

**One_Hertz** · 05-27-2010, 11:25 AM

I will show my iometer when I get home tonight...

From my memory - QD 32 4kb random read - ~135k, which is fine. QD 32 4kb random write - 67k, which is low apparently.

Two of my Es are VERY early drives. I got them the week they came out. Perhaps something is gimped with them. All 4 are running latest firmware though,

**Computurd** · 06-01-2010, 08:25 PM

hey one hertz...did you ever figure out what is the deal with your e's?

**One_Hertz** · 06-02-2010, 05:37 AM

Originally Posted by Computurd

hey one hertz...did you ever figure out what is the deal with your e's?

Nope... Mainly because I really do not particularly care if I can do 70k random 4kb write iops or 150k, especially since my reads are fine.

**GullLars** · 06-02-2010, 06:18 AM

Good point

As long as your sequential write is OK for you too, it shouldn't matter.

**Computurd** · 06-02-2010, 04:36 PM

good perspective. if it aint broke, dont fix it

**GullLars** · 06-02-2010, 06:08 PM

Rather: If some aspect is partially broken, but fixing it won't make a difference, don't bother.

**GullLars** · 06-03-2010, 11:28 AM

I have a project i'm working on to show scaling of different storage sollutions (both total performance and compared to itself) by Queue Depth and block sizes 4-32KB for 100% random 100% read. This allows for a deeper look into the effects of parallellity, and the granularity.

Different from earlier suff i've done, this will be with a linear QD stepping of 1 from QD 1 to 32. This is low to medium "parallell load", and will capture the relevant range for desktops and most kinds of workstations.

I'm looking for results from single HDDs and SSDs, and arrays of HDDs and SSDs. Basically any storage type. The more diversity i get the better. Anything from a 10 year old 5400RPM PATA drive, to a first gen JMicron SSD, to a velociraptor, Intel SSD, and 2-8(+) SSD/HDD arrays.

Results will be posted later as nice sorted graphs and diagrams in a dedicated storage/SSD benchmark thread I'll make.

Here's the config, i've included a txt file with info:
project random read linear QD scaling.zip
It will take about 20-25 mins to run and give me 128 datasets/points.
Post result files (csv) with the naming described in the info file in this thread (for now). If you do runs on multiple configurations, you can include multiple csv files in a single zip file with your username (and possible a general description of content).

Any results are greatly appreciated, since this won't be possible without collaboration, and nobody else are doing anything simelar of this detail and work size and sharing results (to my knowledge).
EDIT: it would be great with a comment on what you think, and if you're in, so i can see if there's any interrest in this or if i should just leave it.

**GullLars** · 06-03-2010, 06:03 PM

OCZ shows off HSDL (High Speed Data Link). A 20Gbps cable interface using a PCIe 2.0 x4 host card. They also have an x16 version with 4 HSDL ports.

**~~onex~~** · 06-18-2010, 09:30 AM

it's not about you lars, and never been, actually u and anvil where the ones that actually really helped out since the beginning, answering every question, especially you, and it is really worth mentioning,
i would have never understand SSD's today the way i am without you'r guys help, though you can watch this thread falling back already to the second page in the forum, after only few weeks and it shows you the amount of caring here at the forum, where everyone is occupied with himself, this is how it looks.
when you fired up this thread, first thing which has came up mind was, heck, this would never be the way it was at the Norwegian forum, simply because people don't seem to care and look where it is...
some people do mind, others just keep to they're own, forming groups etc, this is also happening in the WCG forum, by people's 'ranking' etc.
maybe for you it is o.k, you take it, though i think it is just showing the crippleness and actual lack of commitment and caring of people here, this is not how a forum should behave with all due respect to the technical level which is indeed quite high.
that's the point, the main issue.

**Computurd** · 06-18-2010, 06:59 PM

onex, i think that your complaint could be construed as a basic dislike of society in general

things in the real world are going to happen just the same as that. it doesn't matter if its work, school or the playground man. that just life. that is how society itself is, and this is just a little cross-section of society here as well. i do not feel that as a whole this forum is worse than any other forum out there. dude there is no lack of commitment here in this sub-forum you have a group of guys who come and collaborate together every day. I challenge you to find a forum that has this nice little utopia that you speak of....it doesnt exist in real life or in forums.
sorry, just my take on it bro.

**~~onex~~** · 06-18-2010, 07:55 PM

i can't accept it though, people at anandtech are much different, MUCH different.
yet i'm still investigating it though.
last time with the same happening at a different forum, around 15 people were coming to the thread, posing the same issues, and it was after 2 weeks or so at that site..
here everyone fall silent, why?
because they are just comfortable with it, where people treat them like they are nothing, casually.
i know, this is society, this is what is ugly in society, and every sane person would say that,
where people have to change in order to fit themselves with others, always be sorry, step back, not be too much like that or too much like that, you know, conforming into a pattern,
where others set the tone, where there are no individuals, free thinkers, there is the crippleness of intelligence and freedom of the individual, so lack of creativity.
we have to understand that we as individuals create society, we create all this mess, and it is our responsibility to change that, by stepping out of the line, be different,
not conforming into any pattern, not being socialist or communist or any other form of organization, but people who are free in they're mind and heart,
only then there would be real collaboration, where there is no, me and you, I or him, there is only us, as human beings in this world, working and living,
as one.

**Anvil** · 06-19-2010, 04:15 AM

hang in there onex

I'm not saying things will change, there will always be cliques.

**GullLars** · 06-19-2010, 06:04 AM

One of the reasons I haven't gotten much done lately is because there was zero response to the post i'll quote (actually, there was one from audienceofone).

Originally Posted by GullLars

I have a project i'm working on to show scaling of different storage sollutions (both total performance and compared to itself) by Queue Depth and block sizes 4-32KB for 100% random 100% read. This allows for a deeper look into the effects of parallellity, and the granularity.

Different from earlier suff i've done, this will be with a linear QD stepping of 1 from QD 1 to 32. This is low to medium "parallell load", and will capture the relevant range for desktops and most kinds of workstations.

I'm looking for results from single HDDs and SSDs, and arrays of HDDs and SSDs. Basically any storage type. The more diversity i get the better. Anything from a 10 year old 5400RPM PATA drive, to a first gen JMicron SSD, to a velociraptor, Intel SSD, and 2-8(+) SSD/HDD arrays.

Results will be posted later as nice sorted graphs and diagrams in a dedicated storage/SSD benchmark thread I'll make.

Here's the config, i've included a txt file with info:
project random read linear QD scaling.zip
It will take about 20-25 mins to run and give me 128 datasets/points.
Post result files (csv) with the naming described in the info file in this thread (for now). If you do runs on multiple configurations, you can include multiple csv files in a single zip file with your username (and possible a general description of content).

Any results are greatly appreciated, since this won't be possible without collaboration, and nobody else are doing anything simelar of this detail and work size and sharing results (to my knowledge).
EDIT: it would be great with a comment on what you think, and if you're in, so i can see if there's any interrest in this or if i should just leave it.

I would prefer if people say they don't have the time to participate, or have set up their rig in such a way that testing is difficult. I also understand people are thinking of vacation.
I've got a lot of stuff going on IRL ATM, but if i had some result files i could work on them an hour now and then when i get time.

**Anvil** · 06-19-2010, 06:12 AM

I'll see what drives are available later this evening, should be able to do a few tests.

In which thread did you post this request?

**SteveRo** · 06-19-2010, 09:51 AM

@GullLars - my apologies for not responding, I have been trying to find the last 120 pnts or so in pcmv to overtake Mike one more time -

http://service.futuremark.com/compare?pcmv=333089

**Anvil** · 06-19-2010, 10:45 AM

Steve,

Hmm, 120 points, not much but not that easy as we know.
SS cooling or what?

@GullLars,
Here's a few runs using the LSI 9211, 1x Seagate 5900LP 2TB and 1x Intel X25-E.
I might be able to do a few more runs later tonight or tomorrow.