6_6_6

NCQ: Best Upgrade For a Power User!

Recommended Posts

yes, test I suggest:

make two distinct partitions (same hdd of course)

then do a simultaneous copy of a 2 Gb file in both partitions.

See the total troughput then.

Edited by extrabigmehdi

Share this post


Link to post
Share on other sites
yes, test I suggest:

make two distinct partitions (same hdd of course)

then do a simultaneous copy of a 2 Gb file in both partitions.

See the total troughput then.

I don't know if that's a good idea, because it involves writing as well. Until now, we've only tested reading (with HD_Speed, too).

Copying to a different drive would be better. But you have to make sure that the destination is fast enough.

Share this post


Link to post
Share on other sites
I don't know if that's a good idea, because it involves writing as well.

If writing speed is similar to reading speed (the case for many hdd),

this shouldn't be an issue.

At least it would give an idea either or not it screw significantly performance.

Share this post


Link to post
Share on other sites

I am sorry I am done. My screenshots are above. They speak for themselves.

My system is blazing fast weather it is 2 applications or 10. That is because even with 10 applications reading the disk, the disk is about to push 90 MB/s while without AHCI, its is at 15 MB/s. It is at complete standstill without AHCI and those numbers are not for decoration purposes.

I am not willing to do more. You did not bother even to post a screenshot.

"He who wishes to give little should not ask for more." Anonymous

Share this post


Link to post
Share on other sites

You are afraid that I could be right, aren't you? <_<

What I suggested isn't much work, and it would be interesting for literally everybody reading this thread.

My results look exactly like yours, and I have put so much effort and money into this stinker that I am really fed up now. It is unnecessary to make, upload and post more screenshots that do not show anything new.

The Seagates are packed and will go back to the shop.

My conclusions are:

1. NCQ does not work or does not have any significant impact.

2. All tested hard drives are equal.

3. The Intel ICH9R (and probably other ICHxRs as well) have a terrible bug with AHCI or RAID mode enabled.

My new RAID controller solves the problem because of which I came here in the first place. So it seems obvious that the Intel ICH9R is the culprit.

Edited by FAT_Punisher

Share this post


Link to post
Share on other sites
What I suggested isn't much work, and it would be interesting for literally everybody reading this thread.

Installing Vista isn't much work too. It install faster than XP.

Just set your other drive as the boot drive (set in bios), install Vista on it.

When you are fed up of vista, you can just set original hard drive

as boot drive and format other.

Edited by extrabigmehdi

Share this post


Link to post
Share on other sites
You are afraid that I could be right, aren't you? <_<

What I suggested isn't much work, and it would be interesting for literally everybody reading this thread.

This has already been done at the VERY FIRST POST in this thread.

My results look exactly like yours,

No they don't. You are full of contradictions. Here is your previous post:

SEAGATE ST3640323AS 7200.11 (640 GB, 7200 RPM, SATA-II)

1 instance: ~110 MB/s

2 instances: ~53 MB/s (avg. 25,3 + 27,3) (0% and 50%, 256K blocks)

Whereas it is 105 MB/s in my screenshot. So your result looks nothing like mine.

My new RAID controller solves the problem because of which I came here in the first place. So it seems obvious that the Intel ICH9R is the culprit.

Didn't you just say few pages ago:

I have tested my two WD2500YDs and two WD6400AAKSs with the Dell Perc 5/i (LSI MegaRAID SAS 8408E).

NCQ is not working with all drives (terrible throughput with two concurrent disc accesses).

Still, the situation is very unsatisfactory. Even a quite expensive SAS/SATA RAID controller couldn't improve much.

Dude, stop trying to invalidate the effort with put into this with your inconsistencies. Your one word is not matching with the other. That is why I asked screenshots. And I knew you wouldn't be able to produce any.

Post some screenshots. Both for your measurement in XP Management Console and your RAID controller. Till then, further conversation is futile.

Edited by 6_6_6

Share this post


Link to post
Share on other sites
Whereas it is 105 MB/s in my screenshot. So your result looks nothing like mine.

After that, I removed the jumper (as I have mentioned already) and the results looked like yours after that. The numbers you are referring to were gained with the jumper still in place which seems to put the drive in SATA I mode.

Dude, stop trying to invalidate the effort with put into this with your inconsistencies.

There are no inconsistencies.

The WDs at the Dell Perc 5/i with "not working NCQ" show the following behaviour:

1. Total throughput is shitty indeed, BUT:

2. every process accessing the disc gets exactly the same bandwidth. No application starves.

So my original problem is solved by the Perc: some applications starving, because they can't read/write from/to the drive at all. Comparing to my last computer, based on the same drives and a nVidia nForce4 chipset (incl. onboard RAID controller), the behaviour with the Perc is identical. With the Intel, it was worse by far!

No inconsistencies.

What you call "working NCQ" is actually causing severe trouble for me, as described above and 100 times before.

The reason why I don't post any screenshots has been explained: even more work, and nothing new to see.

As I said: the Seagates are already re-packed for shipment. No more benchmarks and tests from my side, because for me, everthing is crystal-clear.

@extrabigmehdi: see above + installing Vista doesn't take much time?!? Dream on. I've installed it tens of times before, and with all drivers, it takes some time indeed.

Share this post


Link to post
Share on other sites

@FAT_Punisher

So to summarize your experience,

before you got your "Dell Perc 5/i" controller , you were almost unable to do simultaneous access with applications (full access for one & it freeze for others).

Now you have a shitty "throughpout" with simultaneous access, but at least nothing freeze (like me with ICH9)

You are able to get same results as 6_6_6 using "hd test", total throughout remain almost the same

when experimenting with this software.

However, when you do real stuff, you see this NCQ thing doesn't work at all.

Moreover the tests made with "hd test" only work with some block size.

I think you should use block of 4kb, because it's the default cluster size used by the system.

@extrabigmehdi: see above + installing Vista doesn't take much time?!? Dream on. I've installed it tens of times before, and with all drivers, it takes some time indeed.

Ok, here's how to install faster :

* Vlite Vista, and make it in full unattended mode (no intervention required)

* take some beers, I've heard it helps to make you feel things are faster

.

.

.

Seriously , after all the time you spend to gather information, and to buy the seagates,

you could afford a last effort.

Edited by extrabigmehdi

Share this post


Link to post
Share on other sites
I think you should use block of 4kb, because it's the default cluster size used by the system.

Bigger block size is used in order not to pull the stuff from cache... and right off of the drive.

Share this post


Link to post
Share on other sites

@6_6_6

Bigger block size is used in order not to pull the stuff from cache... and right off of the drive.

Really ? Results are the same for me anyway (whatever is block size)

Edited by extrabigmehdi

Share this post


Link to post
Share on other sites
The reason why I don't post any screenshots has been explained: even more work, and nothing new to see.

You are a very impolite person. I had to install linux for you people to test as per your requests before. And you can't post a screenshot.

I am glad it didn't work for you after all .

Share this post


Link to post
Share on other sites

IOMeter single to establish base:

100% sequential read, 256K block size used for single instance. The rest with 2 instances (managers) seperated by half of the drive's sectors (0% and 50%). 3 mins runtime.

iometer_NCQ_single.png

BIOS: AHCI

iometer_NCQ_2.png

BIOS: Disable

iometer_noNCQ_2.png

Share this post


Link to post
Share on other sites
@FAT_Punisher

So to summarize your experience,

Yeah, that's it. :)

Again, I'll stress the point: with the Perc 5/i RAID controller, the system behaves just as it behaved with an old nVidia nForce4 onboard RAID controller. I call it "normal". :D

- I tested all possible block sizes. Smaller block sizes resulted in very small overall throughput, but equally distributed between all instances of HD_Speed (yeah, similar to real-world behaviour indeed!). Starting with 256k, the behaviour described by 6_6_6 began (alternating between 0 and full-speed).

Ok, here's how to install faster :

Thank you for your advices, but I just won't test Vista. It's over for me, really. The drives are ready for the post office, and they'll go there tomorrow.

Even if it works under Vista: I won't use Vista. It's of no use for me. I am using XP, and I will continue to use it, for different reasons (no drivers available for some hardware, slower with certain games, picky about signed drivers (64 bit version), and more).

@6_6_6: sorry, but from the very beginning I in fact considered you to be quite impolite. Others who followed this discussion but did not comment chose even more harsh words to describe your behaviour.

I do appreciate your efforts, but you should simply try to believe me, even if I don't show you 129578125 screenshots which we've all seen already. I've spent a shitload of money, I've spent enormous amounts of my time, and I've shared my results. I think this is quite fair. I do not want to waste even more of my resources on this topic, although I do want others to be able to benefit from what I found.

Please have a look at your screenshots from your last post: in the AHCI shot, you see a maximum I/O response time of 1530ms, that's 1,5 seconds. It's ten times! the maximum I/O response time with AHCI (and NCQ) disabled (155ms)! And that is precisely what I described! Applications react worse, not better, because the response time becomes higher! I think we agree that higher is worse in this case. And if the response time becomes too high, the application freezes (for some time).

I really wanted this NCQ story to be true, because I wanted to profit from the increased performance, but at least for me (using XP), it does not work out at all. :(

Share this post


Link to post
Share on other sites

By the way, one more reason why I won't do more tests is that the Seagate drive seems to remember for how many hours it has been running (see "Seatools"). And like most other S.M.A.R.T. capable drives, it remembers the number of power cycles and spinups it went through. I wanted to keep it as "new" and "unused" as possible in order to avoid problems with the shop where I bought it when I return it. So I kept the tests to an absolute minimum, but it was still sufficient.

Share this post


Link to post
Share on other sites

@FAT_Punisher

I do appreciate your efforts, but you should simply try to believe me, even if I don't show you 129578125 screenshots which we've all seen already.

Taking a screenshot take 2 second, give it to him :lol:

At least it would avoid all the typing you made to explain results are the same.

Please have a look at your screenshots from your last post: in the AHCI shot, you see a maximum I/O response time of 1530ms, that's 1,5 seconds. It's ten times! the maximum I/O response time with AHCI (and NCQ) disabled (155ms)! And that is precisely what I described!

you are right but ...

but if you see the TOTAL I/O per second,

and total Mb/s exchanged they are increased.

Which means: each application might be less responsive, but overall, more things are done.

At least that's how I understand this "cryptic" tool.

I'm unable to launch to it from two different locations by the way,

not enough familiar with it ... (tutorial ?).

I wanted to keep it as "new" and "unused" as possible in order to avoid

all right .... I 've heard that low level format allow to reset smart settings,

but I was unable to do this to my own hard drive ... <_<

Edited by extrabigmehdi

Share this post


Link to post
Share on other sites
Taking a screenshot take 2 second, give it to him :lol:

At least it would avoid all the typing you made to explain results are the same.

But now it's impossible, because I won't unpack and install the drive again. ;) I keep saying it, but nobody listens: it's already packaged. :P

That's why explanations instead of screenies.

you are right but ...

but if you see the TOTAL I/O per second,

and total Mb/s exchanged they are increased.

Which means: each application might be less responsive, but overall, more things are done.

That is true, and it's generally a good thing, but not if applications start to freeze and even to fail because of it.

Like it is for me: when my TV recording app can't access the hard drive for some time, the recording fails. In that case it doesn't help me that the file copy was x times faster when I even can't record a movie with some hundred kB/s. :blink:

Adding to that: as my "real world" test shows, there isn't even more work done. Overall throughput decreased from ~113 MB/s to 20-30 MB/s with two apps accessing the drive (file copying and backup archive verifying).

all right .... I 've heard that low level format allow to reset smart settings,

but I was unable to do this to my own hard drive ...

Didn't know that, otherwise I would have tried that. :)

I wrote all zeroes to the disc, so that you at least can't restore or see anything.

Share this post


Link to post
Share on other sites

@FAT_Punisher

But now it's impossible, because I won't unpack and install the drive again

I think you've done everything in a hurry, and that's not wise.

I don't see the point of getting a seagate, and not experimenting with it.

Anyway your budget must be quite restricted, because I think I would have kept the drive (60euro the 500 GB drive, that's fair).

I've done some experiments, and the "maximum I/O response time" is not relevant at all,

and should discarded. The maximum value occurs only when your start the test, and after, everything becomes normal.

I'm not sure I've understood well this IOmeter tool.

But anyway, I've performed a test with this setting :

iometer-setting.png

First I created a manager, with a single worker, and choosed one target partition.

Here's the result:

one-worker.png

Then I created a second manager, with a single worker, and choosed an other partiton.

Always same setting.

Here's the result:

two-worker.png

Note: the drive I tested is where my system is installed.

Edited by extrabigmehdi

Share this post


Link to post
Share on other sites
I think you've done everything in a hurry, and that's not wise.

I don't see the point of getting a seagate, and not experimenting with it.

Anyway your budget must be quite restricted, because I think I would have kept the drive (60euro the 500 GB drive, that's fair).

That's exactly the reason. My account is overdrawn already. I'm only a student, so my budget is very limited, and a RAID controller including cables and stuff for ~180€ and two Seagates for ~150€ is quite much money. Originally, I wanted to buy only a single WD6400AAKS. Then came a second ("Perhaps the drive is defective, let's compare it to a second."), then the controller, and now the Seagates. :lol:

If the Seagate had given me the performance gain I expected, I would have kept both and have sold the WDs.

I tested enough under XP (on the Intel ICH9R, with AHCI and RAID modes), and it did not work at all. As I said, Vista is irrelevant for my personal use. So that's it. :(

And as you perhaps can understand: I want this testing stinker to end. My time is limited as well, with my diploma thesis beginning soon. :P

Edited by FAT_Punisher

Share this post


Link to post
Share on other sites
Thank you for your advices, but I just won't test Vista. It's over for me, really.

I don't find your efforts genuine. You will spend so much time following forums, posting responses, ordering drives... And not spend 30 mins to install an OS to test. You have shown nothing tangible. All talk.

I've spent a shitload of money, I've spent enormous amounts of my time, and I've shared my results.

Full of conradiction with still nothing to show. One button... Press Printscreen, paste and attach. Duh!

Please have a look at your screenshots from your last post: in the AHCI shot, you see a maximum I/O response time of 1530ms, that's 1,5 seconds. It's ten times! the maximum I/O response time with AHCI (and NCQ) disabled (155ms)! And that is precisely what I described! Applications react worse, not better, because the response time becomes higher! I think we agree that higher is worse in this case. And if the response time becomes too high, the application freezes (for some time).

Please do not fabricate stuff or pose as an expert without having any clue about anything. Have you ever used IOMeter? Do you know what it is? Maximun I/O increases with QUEUE DEPTH. Which is apparent when NCQ is working.

You look for Total I/O's sec which shows how many times in a sec the hard disk can move the head to read/write a file. You look for Total MB/s which shows how much data can the drive push to the OS in a sec and average I/Os which show the median of how fast drive can do the seeks.

By the way, one more reason why I won't do more tests is that the Seagate drive seems to remember for how many hours it has been running (see "Seatools"). And like most other S.M.A.R.T. capable drives, it remembers the number of power cycles and spinups it went through. I wanted to keep it as "new" and "unused" as possible in order to avoid problems with the shop where I bought it when I return it. So I kept the tests to an absolute minimum, but it was still sufficient.

Yet more contradictions. Almost all drives bigger than 8 GB do record all the info about the drive. Yet you do get consumer protection and you can return it within a month almost everywhere in the world since ages. Nobody checks the drive's power on or off time provided that it matches the store's policies. Did the store tell you its spindle start/stop count must be below a certain threshold to return it? What if the store does not accept the drive? Didn't you just pay 180 pounds for a used controller? More for the motherboard and all? And you will go bankrupt all of a sudden with a $80 drive?

I don't find you genuine at all. You post nothing about your specs. You posted no screenshots of any form. And you requested no help or assistance. You just came up with "i got it, i vomit, i return". Duh. Why did you order the drives in the first place if your Perc was satisfactory? And now all of a sudden, it is satisfactory? Do you know what you are saying even?

Like i said, if you indeed have the drives, i am glad they did not work. You did not deserve to benefit anthing.

Share this post


Link to post
Share on other sites

That won't do. You need to access Drive's Service Area, modify firmware, etc. There are specialized and very expensive hardware for that. HRT, PC3000, etc. This is not something to be done alone with software.

You are doing IOMeter right.

all right .... I 've heard that low level format allow to reset smart settings,

but I was unable to do this to my own hard drive ... <_<

Share this post


Link to post
Share on other sites

@6_6_6

Didn't you just pay 180 pounds for a used controller? More for the motherboard and all? And you will go bankrupt all of a sudden with a $80 drive?

yes that's a bit weird. Moreover, I don't have the bug he describes with the controllers of my motherboard .

@FAT_Punisher

and two Seagates for ~150€ is quite much money.

two, but why ? A raid0 does nothing to improve "snappiness" of system and increase risks of loosing data. So unless you constantly write / read big files, I don't think building a raid0 is interesting.

My time is limited as well, with my diploma thesis beginning soon.

Get your diploma first, then spend your time at building an enthusiast rig after, provided you'll earn more money after getting

a job.

Share this post


Link to post
Share on other sites
I don't find your efforts genuine. You will spend so much time following forums, posting responses, ordering drives... And not spend 30 mins to install an OS to test. You have shown nothing tangible. All talk.

You have no clue concerning my efforts. Only because I don't write about every fart here, it doesn't mean I didn't fart.

I have in fact Vista installed since many months, but I will not use it as productive/main OS. So it is completely irrelevant to me.

I have even Server 2008 installed, but unfortunately I did not install the Perc driver before I migrated the system to the Perc, so it's unaccessible at the moment. I don't use it anyway, I only wanted to try out its virtualizing capabilities, which stayed far behind my expectations.

Full of conradiction with still nothing to show. One button... Press Printscreen, paste and attach. Duh!

You're a very distrustful person. I stated over and over again that my HD_Speed results look exactly like yours.

Please do not fabricate stuff or pose as an expert without having any clue about anything. Have you ever used IOMeter? Do you know what it is? Maximun I/O increases with QUEUE DEPTH. Which is apparent when NCQ is working.

You look for Total I/O's sec which shows how many times in a sec the hard disk can move the head to read/write a file. You look for Total MB/s which shows how much data can the drive push to the OS in a sec and average I/Os which show the median of how fast drive can do the seeks.

You still have not understood what is most important for a workstation (in contrary to a server).

In the very first posting where you answered my "problem" you already showed that you didn't understand, and you still don't.

The most important parameter is RESPONSE TIME!

Meaning: when I do a click, I cannot wait for 30s until the application reacts.

xerces8 tried to explain it to you as well, he understood what I was talking about.

Yet more contradictions. Almost all drives bigger than 8 GB do record all the info about the drive. Yet you do get consumer protection and you can return it within a month almost everywhere in the world since ages. Nobody checks the drive's power on or off time provided that it matches the store's policies.

Good thing you know everything about German law...

Here in Germany, you have 14 days to return a product you ordered (in an online shop or similar, not at your local dealer, in that case you got no right to return it at all). And you must not use the product, you must not damage the packaging, you may only do what you could do at a local dealer's shop. Otherwise, you might not get 100% of the money back, but less.

And I have no idea at all what this particular shop will check, and what the next buyer of this particular drive will check, and if he will complain to the shop, and therefore I'm really careful what I do to the drive!

I don't find you genuine at all. You post nothing about your specs.

I posted everything relevant. In this case, mentioning the ICH9R, the OS and my hard drives should be sufficient.

But hey, here you are.

You posted no screenshots of any form. And you requested no help or assistance. You just came up with "i got it, i vomit, i return".

Exactly. What I found in this thread seemed to be the help I needed (Seagate 7200.11 with working NCQ).

And since you praise them so much, it was only fair to report that it does NOT work for me, so that others that are considering to buy Seagates have a second and different point of view. I think this is what discussion boards in the internet are all about.

Duh. Why did you order the drives in the first place if your Perc was satisfactory? And now all of a sudden, it is satisfactory? Do you know what you are saying even?

I now know that you do not properly read what I write.

1. The Perc solves the "freezing application"-problem, because it behaves much better than the ICH9R (see above for details).

2. The Perc does not magically enable NCQ with any hard drives. So the performance with more than one application accessing the same drive stays unsatisfactory, but that is "normal", it is what other controllers do as well (e.g. nForce4 onboard controller), my PC is useable again which it wasn't before (read above, e.g. concerning TV recordings). It is now like it used to be with my previous mainboards. It behaves with the Perc like all my PCs behaved for the past decade. The Intel problem is gone, but it's not faster than a "normal" controller. So the boost of really working NCQ would still have been an enormous improvement. That is why I bought the Seagates. - It should be really clear now... <_<

I know precisely what I am saying, but you either do not read it properly, or you do not (try to?) understand.

- Since I am already repeating myself over and over again, which is quite unnecessary and a waste of time, I will not explain all this again. I'm not "incosistent", "contradicting myself" or something similar, it's all there, you just have to read my postings properly.

Share this post


Link to post
Share on other sites

@FAT_Punisher & @6_6_6

Well, please stop this pointless fight.

If someone else have something relevant with the original topic "NCQ: Best Upgrade For a Power User",

or tested one of these damns seagate 7200.11, please post.

Seems FAT_Punisher you have no more interesting information to share,

so don't try to endlessly justify yourself, that just add no value to this thread.

@6_6_6 : obviously FAT_Punisher got a problem with his original controller,

That's why he bought an other one. I've never experimented the bug he described and probably you too,

so it's useless to discuss if the controller he bought was justified or not. This have nothing to do with NCQ at all.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now