Samsung F1 series hard drives w/1TB model

[nrm]

I have gone through two drives, other had only ecc errors and the replacement also has sector errors. I did some digging and theory of what is going on:


Here's pretty good article that goes to explain HDD theory and error correction. Suggest seeing the part about System Area and Bad block tables.

http://www.myharddrivedied.com/presentations_whitepaper.html

My theory of this is that the HUTIL 2.10 is actually working perfectly besides the Check MC thing.

quotes from the article:

"When the drive is manufactured it is known that there is going to be errors in every drive. Drives use ECC to correct most errors and if ECC can correct the error then the sector is never marked as bad."

Why other tools do not show errors besides HUTIL? Most obvious explanation is that they use a higher level read (similar to reading from Windows) that can only tell errors of such magnitude that drive should be replaced (eg. unable to correct the read data and reallocate the bad sector). But of course there's varying degrees of success/errors since we are talking about analog device. Different manufacturer and command may expose different data, some of which is only meaningful for factory/recovery business.

Hutil likely uses a lower level read command that passes over every sector, even those already in the "P-LIST" as referenced in the linked article (scan: SECTOR ERROR). These sectors were marked as bad in the factory and won't be used by the OS but are not skipped by this low level read.

ECC Errors:

http://episteme.arstechnica.com/eve/forums/a/tpc/f/24609792/m/879007293831
http://www.hydrogenaudio.org/forums/lofiversion/index.php/t50484.html

Depending on the hdd manufacturer, they may or may not expose the ecc data in SMART values for example (Maxtor, Samsung seem to do this). However all drives would use this prml+ecc method to get better data density so you see the values change fast in SMART if you read maxtor or samsung disks. Why does HUTIL show ecc error for only some sectors then? I believe there's some sectors that aren't quite broken but maybe take extra effort to read, this can be seen in HDTACH and other tools by looking the Tranfer rate, it's not very steady, there's lot of tiny drops. Still there's no audible seeks or huge drops which would happen if it was bad sector that would have to be read from other part of the disk. This doesn't happen with all drives though, depends on manufacturer mostly how steady their read speed are.



Summary:

I believe HUTIL 2.1 is working better than other tools, you get low level information you can't get from other drives!

Which drive to RMA?

ECC/Undefined errors, these don't affect perf in a noticeable way in my testing and so no reports have been seen that data written in these areas would be lost.

Lot of SECTOR ERRORS? It seems that if you have sector errors, these areas are reallocated to outer edge of the disk or somewhere very near: On my disk Sector errors, the performance GOES UP when reading the part of the disk where the errors are (perf better at outer edge of the platter). However there is risk that the problematic area spreads to previously good sectors when used, seen this happen. So I would personally return a drive that has more than a few hundred of these. I have more than few hundred so I'll probably replace.


New HUTIL version? I believe they just change it to use higher level read so that drive turns up "ALL OK" up to the point when you lose data.

edit: Based on some perf tests, I think the Sector Error areas are factory reallocd with a large margin: eg. 500 KB of errors in position x: realloc: 500 KB + a whole lot around it. This would decrease chance of new data being near the bad sector area and would make most sense to lower risk of data near the bad area.

[nrm]

On my 3rd drive now, already regretting replacing again as there's a pulsating mid-pitch sine wave noise, essentially a constant "beep beep beep" sound. And sector errors, though at this point I'm ready to give up on this test util as long as I get a silent unit that passes more traditional write+verify test in windows...

If you get a silent unit that works apart from the HUTIL issues I'd keep it or switch brand as finding a perfect no error unit could take quite a few tries it seems.

[zzombi]

"To make it more efficient data is written on the top of a platter and the bottom of the platter and the next platters and so on, at the same time."

Why don't hard disks with more platters, but same platter capacity, write/read faster than single platter versions?

[whiic]

zzombi: "Why don't hard disks with more platters, but same platter capacity, write/read faster than single platter versions?"

Because what you quoted is not true: HDDs do not read/write with each head simultaneously but instead with only one head at a time. Where did you get that bullshit quote, btw?
_________________
Antec 1200 | HX520W | Commando | Q6600 G0 @ 3.15GHz | Noctua NH-U12F | 8GB of RAM | HD 4670 (passive)
7 TB of storage: 1x 1st gen GreenPower (1TB), 1x 2nd gen GreenPower (1TB), 1x 3rd gen GreenPower (2TB), 1x 7200rpm F1, 2x 5400rpm F2 EcoGreen

[DragonOptical]

[nrm]

[nrm]

[bkh]

> Now of course the even more interesting question on this off topic subject that has always bothered me is why would all HDD manufacturers only choose to read only single platter at a time. All the platters are spinning and heads are tracking same position after all.

Tracks are quite narrow, and they aren't mathematically-perfect circles. Moreover, the comb assembly (to which the heads are attached) isn't perfectly rigid, so the heads do suffer sideways motion with respect to one another. The servo dynamically positions one head over one track to get an "optimal" signal, which moves the other heads to suboptimal positions with respect to their tracks.

Decades ago there were a few disk drives that did read all the heads simultaneously for higher performance, but this approach was abandoned both because of the cost and complexity of multiple complete head amplification circuits, and because of tracking difficulties as feature sizes decreased.

[nrm]

That certainly explains it well thanks

[SileX]

[zzombi]

"Tracks are quite narrow, and they aren't mathematically-perfect circles. Moreover, the comb assembly (to which the heads are attached) isn't perfectly rigid, so the heads do suffer sideways motion with respect to one another."

I could see non uniform thermal expansions playing some role. But didn't some not so old hdds use a dedicated platter face for servo data alone?

"The servo dynamically positions one head over one track to get an "optimal" signal, which moves the other heads to suboptimal positions with respect to their tracks."

Indeed, according to those articles tracks aren't well defined nowadays, but overlapping; kinda like a subparticle isn't in a defined place, but all over with various probabilities.

[DragonOptical]

[Lich]

Just got a 1TB model (103UJ) but how should I test it when Hutil isn´t reliable ?

[Luminair]

[andyb]

Try HD Tune, I ran my drive through a full test and it was a sea of green

http://www.hdtune.com/

It also reads the SMART data on this drive, if you want other utilities that work with this drive, you can also use HD Tach (SMART data + benchmark), and Speedfan (SMART data).

Utilities that dont work, Hutil 210, Seagate's test tool under windows (DOS version not tested), and SMARTmonTools.

Please let us know your findings with the Hitachi tool.


Andy
_________________
ANTEC SUPPORT TOTALLY USELESS - READ MORE IN THE LINK BELOW
http://www.silentpcreview.com/forums/viewtopic.php?t=55327
Updated 2009-11-03 with more bad news

[Lich]

Thanks to both of you for the tips

Drive Fitness Test 4.11 (in dos mode) gave the message

"Opertaion completed successfully
disposition code = 0x00"
after a whole advanced test

so I assume that means it didn´t find any errors

Here are results from Hd Tach 3.0.4 and HD Tune Pro 3.00











So do these results look normal? I would like to add that Windows XP SP2 (no other updates) was installed on the disk itself as was to benchmark programs.

The only odd thing i noticed was the "UDMA MODE". The little pamplet that came with the disk says I can change it with HUTIL. So does this have an effect on performance or should I let it be?

edit: and I´ll run the Error Scan in HD Tune tomorrow sence I don´t have the time to do it tonight.

edit2: oh sence there has been discussion about the chipsets fyi I´m running the hd on a DFI LANPARTY LT X38-T2R (Intel ICH9R)

[zzombi]

For my hdd speedfan reports a raw read error rate >= hardware ECC recovered, which seems logical.

[Lich]

Here are also the SpeedFan 4.33 results didn´t post them before because it caused the imidiate reboot at startup but I have managed to find the problem to that HERE.

Looks good to me

[Mech0z]

Wonder if they fixed it now :/ anyone else bought any recently?

[winguy]

320GB review.
http://www.hartware.de/review_791_1.html

I'd like to see this vs WD's 320GB.

[SileX]

[LTG1]

It's really frustrating that people went through so much trouble with these drives when people started seeing the problem as soon as they came out.

All Samsung had to do was look at NewEgg reviews or some other source online and investigate immediately.

Then their HD division should have a blog where the post the exact status of what's happening with the update.

I guess this is wishful thinking for now - small companies are already tightly tied to their customers like this but so many big companies don't seem to get it yet.
_________________
old greeting card guy

[zzombi]

But the mysterious MC error doesn't seem to affect these drives at all. They'll probably doctor the newer hutils so everything will look green. Who knows how many other hdds behave the same, but report everything OK.

And now we can choose: WD announced 320GB platter drives, apparently 7200rpm this time, and quite silent (anandtech). Although a bit slower than F1s and with largish seek times, their biggest problem is that the buyer has to be lucky to get one, since they carry exactly the same name as the older generations.

This 320GB F1 tests nicely, however is it really worth getting one for 75 when for 105 you can get the 750GB F1? This is the question.

[LTG1]

500 vs. 750 pricing:

The biggest problem I have is that if you want to do RAID then 500 GB drives can easily not be enough.

RAID5 means at least three drives, with 1TB of usable data.

1TB holds my music and documents but not much home video at HD res.

If everything was free I'd run only one RAID card in the system to make it quiet, and have a cable run to an external enclosure in a closet with 8 1TB drives in a raid 5 array.

This would give you excellent speed, 7TB of usable data, and of course be extremely quiet.
_________________
old greeting card guy

[nrm]

Here's the news:

HD130UJ
Date: 2008.01
sn s13pj1kq....
firmware 1AA01108

Same HUTIL 2.10 that's been on my flash drive since this mess started.

No M.C. error, no errors in full surface scan in HUTIL.

Here's the kicker though: Simple Read / Write test Error : Command timeout. Even tried switching cables and all the usual along the erase/low level format and this stays there. No big deal, I wouldn't replace it over this, M.C. or the ECC error as Windows tests showed no issue. Only sector error are worth worrying about.

So finally no errors worth worrying about but but: This is the 2nd of 4 units so far along this journey now with medium (along 1-3 khz I estimate) very very annoying sine wave noise. Given all the Samsung marketing speak about how silent their drives and the fact that earlier units I had with sector errors didn't have this noise I can't convince myself to keep this as I'd have to go beyond my usual methods of HDD dampening which work for hiss, vibration but not this sine wave (escapes from where the cables go and then easily audible up to 5 meters...).

[zzombi]

"HD130UJ " you probably mean the 1TB F1, HD103

"escapes from where the cables go" try to decouple the hdd board, the least loosen its screws.

For noise concerns I only consider 1 platters, but the 750 F1 is priced too tempting. If only they'd bring the 320 closer to half its price. Anyway they're not available yet.

[KuniD]

I'm looking at ordering a couple of 750Gb drives in the UK, are these drives still experiencing problems?

Whats the platter setup on the 750's?

[sheninat0r]

IIRC the 750GB F1 uses 3 x 250GB platters, not short-stroked 334GB platters.
_________________
...with a crowbar.

[Luminair]

Samsung is pretty slow, still no new HUTIL after two months. Or GASP maybe it isn't the tool that is broken...

[flavian]

I've been followind your topic for some time. It seems to be the best one on Samsung 1TB related problems.

My experience is as follows:

Received first drive in January, tried Hutil and got the following:
1.Check MC Error
2.Full scan mode - several errors at 70-80% scan
Other than this the drive was fully operational, scanned with Windows scan tool and showed no problems, Windows running fine, copied about 200-300GB on it without problems.

Still, wanted to be sure so I sent the drive back.

Got a new one after some delays, received it today.
First thing - Simple scan with the same Hutil, it showed Simple Read / Write test Error : Command timeout. No full surface scan yet, but NO other errors, no M.C. error.
Second - partitioned and formatted drive, copied some files on it
Third - Scanned again, same result. Started full surface scan test, is 25% and running, no errors so far.

Seems to me like the same situation NRM is reporting. My drive makes no strange noise, if full scan turns out OK I'm keeping this one.

Model: HD103UJ F1_3D Rev. A
Date: 2008.02
sn S13PJ1MQ...