Formatting a new Samsung F4 HD204UI 2tb 4k sector drive Win7 x64?

[bankster55]

I think it gets worse because AHCI is better (faster), but your system cant handle it
Kinda like driving an old car at 90mph, drives fine at 55
Maybe up a notch or two on CPU voltage?

I didnt want to mention this now, because you will forget about my testing post, but the simple way to find out is to put in a powerful AMD CPU, and or, putting your F4 in another PC. As I said in my testing post, your drive IS working at 133, just not constantly. I'm leaning towards thinking the drive is fine

Edit: you could also chnge the NCQ settings in AHCI controller in Dev man properties

 

[airmark]

I think it gets worse because AHCI is better (faster), but your system cant handle it

But then how come it can handle the SSD? Its performance remains high and in fact the windows experience index went up to 7.7 (vs 7.3 with IDE) - maybe a coincidence as I had only ran it once before, so maybe 7.3 was lower than normal, but at the very least it proves that the system can handle AHCI throughput.

 

[bankster55]

hmmm back early today I see
Guess what - I just bought coupla of these drives today at ewiz $90 ea shipped free. Never thought I'd buy a green drive - lol. I want to use parted (text based) to see what happens when you set both offset and sector size at 4096, no emulation, nada 512. Windows is supposed to never see the 4096 sector.
But they wont come for 5-7 days

There is another post today in this section from a guy who loves SP1 as far as HDD "snappier". I am really beginning to think SP1 (RC1) has something for HDD performance

As far as your argument - SSD drive has no moving parts, no wands, no electric motor (they do go bad), no magnetic domains or no servo motor like that on a speaker. Its little different than a ram stick

But we are off topic (as usual).
Lets get back to the grind of eliminating the obvious with testing stuff
Lets not waste the effort that went into that post. Right?
Do 45 min of memtest.org (3.0) run from floppy or CD ISO
Download and run HD Regenerator (2011) trial. If you have a bad sector its usually in the beginning of the drive where the O/S is and you only have a 45GB partition. If you had a bad sector in the middle of the O/S that would ruin the I/O. 2TB would take a long time even tho the old vers 1.71 was alot slower, but your 45GB wont take much time, since Win 7 is about 11.2GB as loaded, thats all you really have to worry about
And once again - are you overclocking you sempron, and if so to what?

 

[airmark]

I just bought coupla of these drives today at ewiz $90 ea shipped free.

You lucky bastards over there. I paid £73 (i.e. $117) 😛

There is another post today in this section from a guy who loves SP1 as far as HDD "snappier". I am really beginning to think SP1 (RC1) has something for HDD performance

Thanks. I'm gonna try it.

As far as your argument - SSD drive has no moving parts, no wands, no electric motor (they do go bad), no magnetic domains or no servo motor like that on a speaker. Its little different than a ram stick

I understand. I was just making the point that the system could easily cope with the data throughput. In any case, I tried again the other mechanical drives (F1 and F3) and they are as before, i.e. faster than the F4.

And once again - are you overclocking you sempron, and if so to what?

No overclocking at all

But we are off topic (as usual).
Lets get back to the grind of eliminating the obvious with testing stuff
Lets not waste the effort that went into that post. Right?
Do 45 min of memtest.org (3.0) run from floppy or CD ISO
Download and run HD Regenerator (2011) trial. If you have a bad sector its usually in the beginning of the drive where the O/S is and you only have a 45GB partition. If you had a bad sector in the middle of the O/S that would ruin the I/O. 2TB would take a long time even tho the old vers 1.71 was alot slower, but your 45GB wont take much time, since Win 7 is about 11.2GB as loaded, thats all you really have to worry about

I'll do these tests tomorrow, as tonight I want to use the PC actually.
Btw, I ran the long MEMTEST when I bought the RAM, about 3-4 months ago. Fairly new RAM, 2x2GB DDR3. So, do think it's likely to have gone bad in the mean time and I should still do it again? Bear in mind this is not a gaming or folding rig, just a Media Center PC, used only evenings and weekends.

 

[bankster55]

Btw, I ran the long MEMTEST when I bought the RAM, about 3-4 months ago. Fairly new RAM, 2x2GB DDR3. So, do think it's likely to have gone bad in the mean time and I should still do it again? Bear in mind this is not a gaming or folding rig, just a Media Center PC, used only evenings and weekends.

Well, if thats the case just run HD Regen, I am really curious about bad sectors here

You lucky bastards over there.

Where are you?

 

[airmark]

Across the pond.

 

[bankster55]

Heh heh

Well I didnt know you had both an F1 and an F3 and both work just fine on your wussy little system. The SSD was irrelevant since its a dif animal.That really brings it back to the drive. Cant be mobo - we moved to another chip and port - sometimes tightening a HSF too much crimps a trace, especially one with a bare steel back bracket. If you pass the HD Regen with zero, as a last effort we will run OCCT while the drive is doing HD tune while in the 4 min idle period, and see if the 12V voltage graphs have little sags

Passing that the only logical move then is RMA. I mean thousands and thousands of folks have bought the drive, installed it and boogied on down the road.

>What firmware shows in HD Tune info tab? I want to compare it with mine when they come.
>What chunksize do you have set for benching?
>Do you have drive set for speed or low noise?
>Please verify 1024KB offset with diskpart (can see in Win 7 cmd)

Hey, just found 4K aligned tests in HD Tune
Wow, version 4.60 is one up to date sw
Also in random access
Now I have a way of REALLY testing my new drives
Kewl

Edit: Just had another thought. download trial vers of Perfect Disk 11 and defrag C. Of course, if you have a bad sector, this is irrelevant.
http://www.raxco.com/products/home-perfectdisk11-home-premium/learn-more

 

[airmark]

Cutting my losses and returning the drive. I may try again if the technology graduates from beta stage or get an old and trusted 1.5TB drive (which is at the sweet spot in terms of GB/£ anyway).
Thanks for your help bankster and hope your drives turn out good. If you find any additional clues after you've worked with them, let me know.
I realized that the firmware wasn't on the drive itself AFTER I took the drive out of the system, closed the case, and wrapped and packaged the drive to ship it back! 🙁 I sent you a pm with a pic of the drive itself with S/N on. Would that help you?

 

[airmark]

Heh heh

Well I didnt know you had both an F1 and an F3 and both work just fine on your wussy little system. The SSD was irrelevant since its a dif animal.That really brings it back to the drive. Cant be mobo - we moved to another chip and port - sometimes tightening a HSF too much crimps a trace, especially one with a bare steel back bracket. If you pass the HD Regen with zero, as a last effort we will run OCCT while the drive is doing HD tune while in the 4 min idle period, and see if the 12V voltage graphs have little sags

Passing that the only logical move then is RMA. I mean thousands and thousands of folks have bought the drive, installed it and boogied on down the road.

>What firmware shows in HD Tune info tab? I want to compare it with mine when they come.
>What chunksize do you have set for benching?
>Do you have drive set for speed or low noise?
>Please verify 1024KB offset with diskpart (can do in Win 7 cmd)

Hey, just found 4K aligned tests in HD Tune
Wow, version 4.60 is one kickass sw
Also in random access
Now I have a way of REALLY testing my new drives
Make my day.

Edit: Just had another thought. download trial vers of Perfect Disk 11 and defrag C. Of course, if you have a bad sector, this is irrelevant.
http://www.raxco.com/products/home-perfectdisk11-home-premium/learn-more

I believe HD Tune tests the drive itself irrespective of partitioning etc. so defrag wouldn't make any difference. Unless you refer to C: as the SSD in my normal installation, but then SSD has practically zero access time and is idle during the test anyway, so can't see how it helps.

I did play around with fast vs accurate and different block sizes without any meaningful change in the results.

 

[bankster55]

Hey, I dont blame you.
All that screwing around for a dumb hard drive install
The only thing I regret is that you didnt test for bad sectors - that was my best guess on your problem after getting your whole situation
The firmware is no biggie
The fact is WDC and Samsung put out new tech drives on an unsuspecting public waaay before these are ready to be run properly. And theres NOT ONE STINKING WORD on the label to tell you its AFT!!!
They CAN be used to good effect, but those guys that run em and love em are just using them for cheap server data drives. Try that drive with CAD CAM and see what happens, but dont dare have your O/S on the HDD. Meh.

Rather coincidentally there was a review out today by the excellent Techreport site about the new 3TB WDC green drives (also 4096) and the add in card Highpoint Rocket Raid that comes with it. Yet more tech not ready for prime time. What are these companies thinking? Do you need a degree in Compter Science to build your own PC?
http://techreport.com/articles.x/19871
The first page is the one with the info, thats all you have to read.

This brings out the coming switchover to GPT partitioning and that whole mess. From what I have learned so far is that the GPT bios chips need to be a lot bigger and code is written in C (cant use regular bios chips), and first actual Intel GPT native will be Sandy Bridge - mid 2011. However MSI has been way ahead of the curve on GPT, using a bios 32MB chip that you could toggle back and forth (MBR/uEFI), and this with old school P45 mobo. I was totally unawares of this. Dunno if this is the new C coded chip.
http://www.msi.com/html/popup/MB/uefi/applied_model.html
http://www.msi.com/html/popup/MB/uefi/download.html

The only reason I bought the drives is to get answers just for my own personal understanding of the tech, and I would never use the things in any PC of mine.

Have a nice day fella

 

[airmark]

When I first installed it and formatted it, I did the long full format, not quick. It took several hours (~7-8). So, wouldn't that had revealed any bad sectors? Also, afterwards I copied 1.3TB of data on the drive without any problem. So I doubt bad sectors are to blame, esp because the spikes in the HD Tune graph are throughout the hard drive, not in one or two areas.
But anyway, enough time wasted on this. Have a good day too!

 

[bankster55]

Well, you might keep an eye on this thread, after the drives come and i get a little free time, I'm gonna start doing some testing with the drives in my physical possession.

 

[bankster55]

O.K. The drives came this morning from ewiz (or superbiiz)
First of all, dont like the way they were packed.
They had hard black styrofoam endpieces on ea side and were in boxes exactly the size of the styrofoam, This means they take the full brunt of any shock as they were tossed around.
They need to be in a second separate larger box so that as they were hit, there would be no direct impact transfer.
You will soon see why this might be important here.

The drives I received were Sep 2010 Korea
3 platter 667G/platter, 333GB per side(head)
Firmware 1AQ10001

Note the drive has all kind of tweaking features - its an eco friendly drive.

So as a baseline I decided to load up the drive just as i got it - bare naked. The only free PC I had at the momemt was an XP Win 7 dual boot, which had also 2 cloned backup drives.
So I formatted the Samsung with XP diskpart in my other drive, which makes the offset to the beginning of the first partition at 32K, or 512B X 63 sectors = 32K rounded. Wanted to see if "normal" 512B sector 32K align would work.

These drives were on an Intel SATA controller set at IDE mode Intel ICH10R dr 9.1.1.1027 inf
Drive was recognized at boot no prob
Booted into Win 7

Things get real interesting real quick
I fired up HD Tune 4.60 and did the read benchmark
The drive came as set to low acoustic mode (128)

SHAZAAM!!
The drive started out real good, then at the 750GB mark (base 10) it went into the same crap mode as airmarks whole HDD, then amazingly it recovered at around 1440 GB and went on its merry way.
THANK GOD I MADE A SCREENIE!!
Why?
Because it fixed itself!
I would have been so pissed if I couldnt verify what happened.

So I then went to HD Tune AAM and changed to performance mode (254). Re ran the bench and gess what - went perfectly.

http://i54.tinypic.com/sxn041.jpg
then I changed back to low noise and ran bench again - also worked perfect. Note the dif access times for per and low noise.
http://i56.tinypic.com/5lz8zk.jpg

So what have we learned so far?
Airmax probs had nothing to do with IDE SATA or AHCI, partitioning, or aligning, or bad sectors or weak CPU etc.
Its the friggin drive - something going on there.
Run of sticky motors? Misaligned wands?
Drive dropping into low power mode?
I initially thought that a pair of heads and its platter were borked, but the drop area doesnt coincide with platter/cylinder barriers, it should have been the whole drive random, not in one contiguous bunch.

Since HDD tune shows 512 sectors as does the label, this means anything that looks at the samsung will only see 512, never 4096.
So what happens when you use Linux to partition the samsumg with 4096B sectors - is that a waste of time?
I am now wondering if the WDC Green EARS also shows as 512 sectors in HD Tune?????

More to come, next up Win 7 only on a drive with correct offset running HDTune on the samsung one big partition

 

[Emulex]

they probably do alot of sector reallocs the first time around.

do a full format (not quick) and then fill it up a few times so it can do a bunch of bad sector remapping then it should stabilize.

you want cheap AND good 😉 lol

 

[bankster55]

Already tried that with HD Regenerator in scan only mode
0 sectors reallocated on entire drive, tho I do recall airmark did a quick format.

However a strange thing happened I have never seen with HD Regen - at 45 degrees it put out out a cool HDD immediately blinking message. 45 is warm but not overly so, I think the drive itself has some kind of temp monitor - maybe lowering power? Running the bench is intensive for the drive, but never had that happen with a WDC. Have SMART off in bios.

But I think whatever fixed it is permanent

So I'll just do remote benching of Samsung from Win 7 SP0 on a WDC FEAX with 1024 offset, SATA AHIC, one big partition, to slaved no O/S Samsung, full format (4 hours down the drain), and I'm gonna also try the new 9.12.1008 Intel inf drivers. Additionally want to see if SP1 changes things
Then I move over to Linux tools.

 

[bankster55]

Hmmmmm. Just noticed on the pic airrmark took of his drive has a P/N with a Z4 and thats also showing in the newegg pic of the drive
My P/N has no Z4. Both 9-2010, both Rev A
And my P/N on top is also dif - the last 5 letters
(**sigh**)
Geez, hope there's only one 2TB F4
Anybody know?

 

[tweakboy]

I recommend going 4 sectors which is 2kb per cluster

The default will be 4090KB

If you go 512k bytes since its a large hard drive that might hinder it a bit. Also make sure sata cables are firmly tightly fit in both sides and also you can try a different SATA slot on your mobo.. GL ,,,

BTW you can also use Partition Wizard to change cluster size without losing any data on the drive. but after doing that you must do a re image or else its gonna be fragmented and slow. gl

 

[bankster55]

Gotta be careful here. Not talking bout a 512B drive.
There are physical sectors and virtual (emulated to 512BYTES) sectors to the O/S. Both XP and Win 7 default to 8 X 512Byte 4096Byte clustors during their formatting. But on an O/S that can see 4096B sectors (4Kn), 4096 would be the sector and if you made the cluster 4096, then one cluster would be one sector.
Thats why Win 7 diskpart aligns ITSELF to 1024KB by default, or 1 MiB
1Mib could hold up to 256 4096B sectors
Thats why WDC calls their 4096BYTES sector drives 512e and Samsung uses 2 powerful ARM processors to do all that emulation back to 512B on the fly.

Quote:
Advanced Format
Generation one categories:
* 512 emulation (512e): 4K physical sectors on the drive media with 512 byte logical configuration translation to the host.
* 4K native (4Kn): 4K physical sectors on the drive media and 4K configuration reported to the host.
* 4K-ready Host (Client devices only): A host system which works equally well with legacy 512 as well as 512e hard disk drives.

-----------------------------------------
This is why I want to try Linux formatting which can do 4Kn, but if the 2TB F4 drive I have is not 4096, then testing is useless. The WDC 2TB green has 3 dif models and only the EARS is 4096


Quote:
Will it break Windows XP compatibility?
Windows XP and other older OSes work by accessing specific disk sectors, not by using 'atomic writes' that work on byte-level data placements that we see in Windows Vista, 7, Server 2008, MacOS 10.4, 10.5 and 10.6 (Tiger, Leopard and Snow Leopard) and recent Linux kernels from 2.6.31 onwards.

Windows XP (and Windows 2000 and Server 2003) use a method of accessing the hard disk that starts at sector 63. Because 4KB sectors match eight 512B sectors, sector 63 is unfortunately is one sector shy of being directly compatible, so using a 4KB data sector with these OSes causes mis-aligned writes.
--------------------
So Win 7 can write to the magnetic domain level (amazing)
If it has a HDD with hard wired physical 4096Byte sectors, the most important thing becomes WHERE does the first partition start - the alignment
In Win 7 you FIRST align, then are emulating to 512 on the fly and in XP you are stopping emulating with a jumper on 7-8 and aligning on the fly with third party sw.

Note here I formatted this Win 7 install with its diskpart, and this was on a "normal" 512 sector drive. Then I went into killdisk looking to see what happened after installing the O/S

Notice that I had the unallocated space at the beginning of the drive, also had 63 sector as start of first partition, but then theres that curious 992KB unallocated. Then I got it - "normal" offset on 512 drive is 512 X 63 = 32,256BYTES or 32KB, and win 7 made the offset 1024KB and 1024 minus 32KB = 992KB. In other words it embedded the 32K offset in a much bigger space - a space that is 4096 compliant

I know this is all just a bunch of tech crap that 99.8% of the peeps could care less about, but I am interested in the theory behind all this tech. I dont care if the F4's are fast or slow. The original post here by airmark I think has been solved by the insight from Emulex - get screwey benches - full format and put something on drive, if that fails run HD REgenerator.
Problem solved

Update: I just called Samsung and talked to "ericka", and found out (she thought) Z4 means refurbished (heh heh), so that means I do indeed have a Samsung F4 4096 drive. Great.
http://www.contacthelp.com/directory/Shopping/Electronics/Samsung?ListingID=290
You see, there is no way to tell what sector format is on a HDD unless the manuf tells you.They will all show 512, at least for now.

 

[bankster55]

Just did a write bench, before I change over to AHCI
Pretty good numbers actually - virtually same as read

 

[airmark]

they probably do alot of sector reallocs the first time around.

do a full format (not quick) and then fill it up a few times so it can do a bunch of bad sector remapping then it should stabilize.

you want cheap AND good 😉 lol

The original post here by airmark I think has been solved by the insight from Emulex - get screwey benches - full format and put something on drive, if that fails run HD REgenerator.
Problem solved

Actually, that's exactly what I first did. Put the drive in, used default Win7 settings to do a FULL format (took several hours), then copied A LOT of data in it with no error, and then noticed that when I selected all files on the drive and right-clicked and selected properties, it took longer than usual to count all the files and the bytes used. Then I ran the benchmark that confirmed the crap performance. You know the rest of the story. I have now returned the drive. If there's a real insight on how to fix it, I might buy it again.

 

[bankster55]

Hey youre back!

I know theres a ton of stuff to wade thru here, but did you catch the part where I found out by calling Samsung that F4 2TB that have Z4 in the P/N are refurbs?

http://www.contacthelp.com/directory/Shopping/Electronics/Samsung?ListingID=290

Shame you still dont have drive, was gonna suggest lightly tapping side of drive with plastic hammer while benching. Still could be QC issues I guess.
Could be too tight tolerances between wands and servo. One thing I do know - i would never use this drive for mission critical use.

 

[airmark]

did you catch the part where I found out by calling Samsung that F4 2TB that have Z4 in the P/N are refurbs?

I skimmed the posting and didn't realize it at first. I realize it now and I'm furious for wasting all this time because of a bad drive that was sold as new when it was in fact refurb.
In the mean time the retailer sent the following message:
[...] Dear ..., Following extensive tests by our Returns staff, this item was found to be faulty.[...]

So, I may buy another drive but not from them.

 

[RebateMonger]

I know theres a ton of stuff to wade thru here, but did you catch the part where I found out by calling Samsung that F4 2TB that have Z4 in the P/N are refurbs?

Maybe more evidence? Geeks.com listing a Samsung REFURBISHED drive with MFG Part # HM321HI/Z4.

http://www.geeks.com/details.asp?inv...-NDW-R&cat=HDD

 

[bankster55]

Maybe more evidence? Geeks.com listing a Samsung REFURBISHED drive with MFG Part # HM321HI/Z4.

http://www.geeks.com/details.asp?inv...-NDW-R&cat=HDD

Yeah, thnx for input.
Wonder if refurb means - returned but "gee it worked for us", lets send it out again.
Since "refurbing" a HDD means taking off the cover in a super clean room, and replacing parts - would be just as cheap to give out a fresh drive from the inventory I would think.
Like on a mobo, one can replace a capacitor, but all that hassle for a device that prob cost them less than $30 to make?
And what does that do to their reputation in the "forums" and comments sections. Penny wise, pound foolish?
Of course if you use a "near" high level clean room and get any microscopic dust or fingerprints on a platter, that would wreck the drive for sure. Who ever heard of refurbing a HDD anyway?
Then theres the situation that SOME vendors selling these drives make no mention of this.

Compare the pic of my label posted previously with the one airmark posted, No Z in mine
So I guess superbiiz sells FRESH Samsung drives - lol!
And even having said that, I nevertheless DID have the problem originally with the drive on first bench out of the box.

 

[bankster55]

The ongoing nightmare known as the Samsung F4 continues:

I really wanted to test to see if Win 7 X64 WITH SP1 was faster than Win 7 X64. So I formatted the drive to 25 GB and loaded Win 7 X64 SP0 ON THE DRIVE in AHCI mode and loaded with latest Intel RST 10.0.0.46 drivers.
They wouldnt load during install when on floppy, but they could be afterwards in dev man update driver.

So I rebooted and ran HDTune as a first shot baseline
Note first that WHEN THE O/S is ON the drive AAM cannot be set to 128 or 254 - defaults to 0 and cannot be changed.

Update - with drive configured as SATA IDE Win 7 X64 it can be set, but still charts bad

So I ran the HDtune and got 4 intervals with the crappy spikes

Then HDTune crashed (remember some apps dont like 4096 emulated to 512 HDD)

Ran the bench a second time and got only 1 crappy spike, then ran HDTune again, perfect trace with the usual 58-104-137MB/s min avg max results. Is the drive loosening up with heat?

So I then installed SP1 to X64 because I wanted to really know if it helps HDD somehow. Well, that turned out to be a bad idea. Immediately after reboot reran HDtune and guess what - airmark type total crappola and SAME HIGH CPU USAGE - 40% as airmark. Thats quite a jump from 1.9% up to now

So what have we learned here?
AHCI doesnt do squat, SATA IDE is fine
Make sure AAM is set to 254
Apps are dubious AS RUN ON 4096 HDD with Win 7 X64 SP1 with boot O/S on drive
Never run your O/S on a Samsung 4096 HDD as currently sold
Now heres the funny part:
Moved the samsung AHCI (with O/S) down boot order to my XP SP3/Win 7 X64 dual boot SATA IDE (changed bios back), booted to XP, and made sure AAM was set to 254 and ran bench - perfection again

So, I am now calling BS on this aligning thing
No matter what I do I get the same benches when the drive is REMOTE
XP with 32K offset, Win 7 1024 offset, diskpart XP or Win 7, Easeus, Paragon partitioning/formatting
What it boils down to ITS THE BIOS - hard wired to 512B sectors and 63rd sector startpoint of first partition
To make aligning relevant you need an uEFI bios, and HDD frmware that will show the drive as virtual 512B or real 4096B, like the MSI one click toggle bios
As on the new Gigabyte P67 mobo's
http://www.overclockers.at/news/details_zum_gigabyte_p67a-ud7_und_mehr
Whether you select LBA, Large or CHS as geometry, an MBR bios will make 63rd sector as start, no matter how far in you put it
The 2 ARM RISC chips do the virtual aligning, not the user
The next puzzle is if Win 7 can read the HDD at the magnetic domain level ignoring the emulated sectors, is that the problem??

So, I'm gonna get a WDC green EARS drive to continue my testing in Ubuntu. At least that drive works all the time - I hope!

At least they give you a little info on the label, but not a word about the sector size (as seen by O/S) - no total LBA number which would tell us, but i consider that a good sign. Maybe under right circumstances drive can be seen as actual 4096 sector HDD.

WDC also has 1--8 pins while F4 does not, only 1-2-3-4