Advice on internal disks

[Tesselator]

I have to agree, super fast and quiet lovely drive, inexpensive too

I was just about to suggest the same thing. There's four different ones tho so be aware. And the UI model just got a 1.5TB added to the line. 🙂

http://www.tomshardware.com/reviews/hdd-terabyte-1tb,2077-4.html

The part name for the 1.5TG is HD154UI. The UI is the low power model. I suggest the RAID Class UJ model. part name: HE103UJ. It's like $15 or $20 extra (which brings it to about $110) but it's the fast one that is making so much ink! Good stuff! Currently it shares 1st place with the WD Black. The Black is a little faster at random small files but the Samsung UJ has faster throughput - for like digital photography, DD Sound recording, and video. The 1st 15% of the 3 plater stack will do 130 MB/s too! I'm impressed!

In another thread i said I hadn't seen a drive that was speced at 60C max - well this one is. 🙂 This means you can run your drives a little hotter which means you can turn fan speed down which means a MUCH quieter system!

Here's the maker page:
http://www.samsung.com/global/busin...p=&type=58&subtype=70&model_cd=233&ppmi=1162#

 

[Chilz0r]

Thumbs down to Western Digital.
Thumbs up to Seagate.
7200RPM is fine.

Nice troll there bro.

 

[Mac Husky]

Asked it in another thread already, but maybe better here:

Using RAID 0/1 only, is it usefull to go for "RAID optimized disks" as there are the Western Digital RE3 instead of Caviar Black or (for friends of Samsung 😉) new Samsung RAID class modell H103UJ instead of Spinpoint F1 😕
Or are they worth it for "higher RAIDs" only 😕

 

[VirtualRain]

I know it's expensive, but have you considered SSD's?

Even a single drive for your OS/APPS would make your machine seem light-years faster... 2 in RAID0 and you couldn't work off a magnetic disk ever after! 😱 Combine that with a large WD Black drive for storing your project work and you would be in heaven. 😀

 

[Tesselator]

Asked it in another thread already, but maybe better here:

Using RAID 0/1 only, is it usefull to go for "RAID optimized disks" as there are the Western Digital RE3 instead of Caviar Black or (for friends of Samsung 😉) new Samsung RAID class modell H103UJ instead of Spinpoint F1 😕
Or are they worth it for "higher RAIDs" only 😕

Yes! RAID Class drives or "Enterprise" class drives are typically faster and made for 24/7 operation. In the case of the H103UJ it's only $15 to $20 more and very worth the $120 total!

IMO SS Drives aren't there yet. Maybe in 2 years. Currently you pay a lot more for about the same or slightly better performance and get way less space at the same time. 🙁

 

[NoNameBrand]

I thought RE drives were meant for use in RAIDs? There's a feature of consumer drives designed for data protection that slows disk performance*, that's disabled on the RE drives as RAID provides that sort of protection.

*though less than 10% on the Black models compared to RE3, IIRC.

I went with WD Black disks last week, because there were no fast Hitatchi 1TB models in the city, and I'm going to give Seagate a pass for a little while due to the disk-bricking issue and their awful handling of it.

I've had no problems with 7200.9/.10/.11 disks in the past, though, and am likely to buy Seagate again in the future. The Seagate 7200.12 disks should be pretty fast.

Tom’s Winter 2008 Hard Drive Guide (from the fall of 2008!)

 

[Tesselator]

I thought RE drives were meant for use in RAIDs? There's a feature of consumer drives designed for data protection that slows disk performance*, that's disabled on the RE drives as RAID provides that sort of protection.

*though less than 10% on the Black models compared to RE3, IIRC.

From what I've read that's totally incorrect, yes. 🙂 But I'm not designing or manufacturing them so all I know is what I read. If you read something (seemingly crazy) like that I would be very interesting in reading the original context. Where did you read that? Got a link?



PS, that link is the same link I provided above. It's a nice article huh?! For example it says:

"If you want your system to run 24/7, to be available for downloads or other services, then you should make sure that your hard drive was designed to support continuous operation. That’s definitely the case for the Samsung (and other) RAID edition drives."​

.

 

[NoNameBrand]

From what I've read that's totally incorrect, yes. 🙂 But I'm not designing or manufacturing them so all I know is what I read. If you read something (seemingly crazy) like that I would be very interesting in reading the original context. Where did you read that? Got a link?

http://www.tomshardware.com/forum/244141-32-raid-edition
http://techreport.com/articles.x/15588

I believe it's the "TLER" stuff mentioned: "If the [RE3] drive can't recover the error within this limited time span, it simply gives up, deferring error recovery to the RAID controller.", whereas the non-RE3 Caviar Black drives will do error recovery for longer.

What I read about previous RE disks (a few years ago, when I was spec'ing out servers, trying to convince the pointy-haired bosses that SAS wasn't practical, given that we needed the capacity of SATA disks) made me think they were perfect for servers running hardware RAID or MD software RAID in Linux, but I'd rather have the non-RE models in non-RAID situations.

 

[nanofrog]

http://www.tomshardware.com/forum/244141-32-raid-edition
http://techreport.com/articles.x/15588

I believe it's the "TLER" stuff mentioned: "If the [RE3] drive can't recover the error within this limited time span, it simply gives up, deferring error recovery to the RAID controller.", whereas the non-RE3 Caviar Black drives will do error recovery for longer.

What I read about previous RE disks (a few years ago, when I was spec'ing out servers, trying to convince the pointy-haired bosses that SAS wasn't practical, given that we needed the capacity of SATA disks) made me think they were perfect for servers running hardware RAID or MD software RAID in Linux, but I'd rather have the non-RE models in non-RAID situations.

WD has a utility that allows you to change the TLER values. Quite handy, as you can use it to add the GP drives to a RAID, or to up the value of an RE2 or RE3 for single drive use. 🙂

 

[Tesselator]

Thanks for looking those up NoName! And for the tip too Nano!

I wasn't aware of this. But I wonder how much difference it makes in actual practice. In the one thread TeraMedia said:

"The RAID edition HDDs from WD have a shorter retry period when there's a disk fault (Time Limited Error Recovery, or TLER for short). So with a normal disk, if it detects a problem while reading a file, it will retry for a while until it gets the data or determines that it cannot. With a RAID edition disk, if it detects a problem while reading a file, it will retry for a short interval (I think 5 secs or less), and then tell your computer it has a problem."​

And two posts down Joex444 says:

"...RAID Edition is rated for a higher MTBF, which should mean that it will last longer. "​

Have you ever had a drive sit there and reread a sector or block? I have on an ancient WD 1GB drive. Or maybe it was 100MB I can't remember. But since then no, nothing like that. I set up and maintained VOD servers for a friend of mine who owns a hotel chain and just chucked any drives that failed with no need to think about why anything failed other than environmentals like air conditioning. I'm babbling but 5 seconds seems like a VERY long time. Just one of those and I'm chucking the drive! It's a good soon-to-fail indicator where S.M.A.R.T. isn't available I think. Hmm, well, actually that IS the failure point isn't it? 😉

Anyway I can't see how this is more important (in actuality) than a non-fudged and much longer MTBF/MTBR rating. Right? If the drive is made to be faster, to withstand harsher environments, and be run 24/7 then when the drive is ready to hit the dust-bin (platter error) who cares if it retries for 5 sec. or for 20 sec.? As soon as it does either I'm throwing it away.

 

[DoFoT9]

*stuff*

Anyway I can't see how this is more important (in actuality) than a non-fudged and much longer MTBF/MTBR rating. Right? If the drive is made to be faster, to withstand harsher environments, and be run 24/7 then when the drive is ready to hit the dust-bin (platter error) who cares if it retries for 5 sec. or for 20 sec.? As soon as it does either I'm throwing it away.

well if its a platter error then yea, throw it away.. but it mightnt be a plattor error!! it might be fixable!! so you could save yourself buying a new one!

 

[Tesselator]

well if its a platter error then yea, throw it away.. but it mightnt be a plattor error!! it might be fixable!! so you could save yourself buying a new one!

I might be missing the point of the TLER function and the REx ratings but I'm throwing my drive away at the 1st sign of trouble anything like an error as is being described here - unless I know I caused it myself somehow or can guess how the error might have come to be.

An error that can't recover in 5 sec. of retries? What is that anyway, between 50 and 5,000 read attempts (depending on what the drive does on an error condition). If the file is splat and the OS reports it as an erred file you just trash it. I don't want the disk taking 20 seconds (certainly not longer that 2 or 3) to tell me about it. It I'm getting many of those the drive is dead. After just 2 or 3 unexplained broken files I'm testing the integrity of the media and will act on the results from that.

I dunno, like I say... maybe I'm missing the point. <shrug>

 

[DoFoT9]

I might be missing the point of the TLER function and the REx ratings but I'm throwing my drive away at the 1st sign of trouble anything like an error as is being described here - unless I know I caused it myself somehow or can guess how the error might have come to be.

An error that can't recover in 5 sec. of retries? What is that anyway, between 50 and 5,000 read attempts (depending on what the drive does on an error condition). If the file is splat and the OS reports it as an erred file you just trash it. I don't want the disk taking 20 seconds (certainly not longer that 2 or 3) to tell me about it. It I'm getting many of those the drive is dead. After just 2 or 3 unexplained broken files I'm testing the integrity of the media and will act on the results from that.

I dunno, like I say... maybe I'm missing the point. <shrug>

no i dont think your missing the point, im just saying that just because one tiny bit of data cant be read (for one of many reasons) doesnt mean that the WHOLE HD is dead. run a few programs over it to mark the sector/whatever as bad and your sweet to go...

 

[nicolasmasset]

Could somebody explain this please? I'm reading the tomshardware review roundup but I don't understand:

- sequential data transfer rates
- access time and I/O performance

What are those and what do they mean in real world usage?

 

[DoFoT9]

Could somebody explain this please? I'm reading the tomshardware review roundup but I don't understand:

- sequential data transfer rates
- access time and I/O performance

What are those and what do they mean in real world usage?

sequential data transfer data rates are the constant data rates a HD can copy at, say you are trying to copy a very large file to an external HD.. the data transfer will be how many MBps it can constantly copy.

access time is a combination, of seek time and the head actually moving into position. with a HD it is normally a few milliseconds.

I/O performance is the amount of operations IN and operations OUT that are performed per second.

these all change depending on what you want to do. if you want to constantly copy files fast, then go for a drive with high transfer rates.

if you want a HD that can read files faster, then you go for the lower seek times..

 

[nicolasmasset]

sequential data transfer data rates are the constant data rates a HD can copy at, say you are trying to copy a very large file to an external HD.. the data transfer will be how many MBps it can constantly copy.

access time is a combination, of seek time and the head actually moving into position. with a HD it is normally a few milliseconds.

I/O performance is the amount of operations IN and operations OUT that are performed per second.

these all change depending on what you want to do. if you want to constantly copy files fast, then go for a drive with high transfer rates.

if you want a HD that can read files faster, then you go for the lower seek times..

Wow! So correct me if I'm wrong here...
A boot drive or a drive to read and load drumsamples into logic for instance doesn't benefit from high transfer rates, but needs a low seek time? And a time machine drive DOES benefit from high transfer rate since it has to copy alot of stuff from the other drives?

 

[DoFoT9]

Wow! So correct me if I'm wrong here...
A boot drive or a drive to read and load drumsamples into logic for instance doesn't benefit from high transfer rates, but needs a low seek time? And a time machine drive DOES benefit from high transfer rate since it has to copy alot of stuff from the other drives?

yes thats exactly right. current SSD drives are perfect for booting!! they have EXTREMELY fast seek times (literally none), but some really lag in sequential write speeds (read speeds are quite high). you have to pay more for the better write speed ones...

were you after any particular drive?? or just browsing (i sound like a sales assistant haha)

 

[nicolasmasset]

yes thats exactly right. current SSD drives are perfect for booting!! they have EXTREMELY fast seek times (literally none), but some really lag in sequential write speeds (read speeds are quite high). you have to pay more for the better write speed ones...

were you after any particular drive?? or just browsing (i sound like a sales assistant haha)

DoFoT9, I'd really apreciate it if you use your salestalk on my thread here 😀 That's my whole situation, but I didn't really get the answers I was looking for.. maybe you can help?

 

[nanofrog]

I might be missing the point of the TLER function and the REx ratings but I'm throwing my drive away at the 1st sign of trouble anything like an error as is being described here - unless I know I caused it myself somehow or can guess how the error might have come to be.

An error that can't recover in 5 sec. of retries? What is that anyway, between 50 and 5,000 read attempts (depending on what the drive does on an error condition). If the file is splat and the OS reports it as an erred file you just trash it. I don't want the disk taking 20 seconds (certainly not longer that 2 or 3) to tell me about it. It I'm getting many of those the drive is dead. After just 2 or 3 unexplained broken files I'm testing the integrity of the media and will act on the results from that.

I dunno, like I say... maybe I'm missing the point. <shrug>

TLER* of a consumer drive (single disk) use is 7sec, and RAID is set to 0sec. The latter falls on the RAID controller for error recovery.

As for tossing a drive that quickly, it might be worth testing the disk with a utility. It can actually verify if the drive is bad. Otherwise, it may just be corrupted data, which (hopefully), was backed up previously for a file restoration.

If there's a bad sector, remap it. Things to try at any rate, rather than just toss it immediately. 😉 😛

*WD's factory specs, and typical of most drives I've encountered.

sequential data transfer data rates are the constant data rates a HD can copy at, say you are trying to copy a very large file to an external HD.. the data transfer will be how many MBps it can constantly copy.

access time is a combination, of seek time and the head actually moving into position. with a HD it is normally a few milliseconds.

I/O performance is the amount of operations IN and operations OUT that are performed per second.

these all change depending on what you want to do. if you want to constantly copy files fast, then go for a drive with high transfer rates.

if you want a HD that can read files faster, then you go for the lower seek times..

The Server vs. Workstation use issue. 😱 😀

Of course, one could choose both, if desired. Fast Random Access for something like an OS drive, and High Transfer rates for Data. Prefferably for RAID, as a considerable boost in speed and redundancy (depending on array type) can be had for a reasonable fee (relatively speaking of course). 😉 😛

 

[DoFoT9]

*TLER stuff*


The Server vs. Workstation use issue. 😱 😀

Of course, one could choose both, if desired. Fast Random Access for something like an OS drive, and High Transfer rates for Data. Prefferably for RAID, as a considerable boost in speed and redundancy (depending on array type) can be had for a reasonable fee (relatively speaking of course). 😉 😛

yes both would be the best option... however it would be much much more expensive lol!

 

[Tesselator]

TLER* of a consumer drive (single disk) use is 7sec, and RAID is set to 0sec. The latter falls on the RAID controller for error recovery.

As for tossing a drive that quickly, it might be worth testing the disk with a utility. It can actually verify if the drive is bad. Otherwise, it may just be corrupted data, which (hopefully), was backed up previously for a file restoration.

If there's a bad sector, remap it. Things to try at any rate, rather than just toss it immediately. 😉 😛

*WD's factory specs, and typical of most drives I've encountered.

Oh, 0 seconds and the controller takes responsibility. OK So, what are the pros and cons of the controller handling such things? And out of curiosity does the Mac Pro's chip-set RAID also function in this way?

 

[nanofrog]

yes both would be the best option... however it would be much much more expensive lol!

Performance requires compromise. With you're wallet. 😱 😀 😛

Oh, 0 seconds and the controller takes responsibility. OK So, what are the pros and cons of the controller handling such things? And out of curiosity does the Mac Pro's chip-set RAID also function in this way?

In the case of the controller, if it determines if a drive is problematic, it elliminates it from the array, and will rebuild it (assuming the array type is equiped with some level of redundancy). In the case of a single drive, the drive becomes unresponsive until the drive is corrected (remapped). This can take a bit, up to a few minutes. Not acceptable for a system meant for high availability.

TLER values are meant for RAID only, and help keep the array up and running with fewer instances of a drive being determined as "problematic". It's useless in a consumer/single drive system, as there's no controller to handle the processing, or software/driver to place the load on the CPU. Hence the default value of (0s for both read and write). The 7s is for reads BTW. Writes are set at 0s. Forgot to distinguish that. 😱

In the case of a software RAID, such as OS X, the error recovery can be handled the same way. The functions are built into the software, not the SATA controller chip soldered to the board. The CPU does the work. Eats clock cycles, but is cheaper than a hardware solution.

 

[byrgenar]

Re: noise

Someone mentioned noise a while back. As far as my experience extends, new drives are not normally the components to worry about in Mac Pro's. The PSU etc will drown them out. Older drives are another matter (ugh, seek noise)

Regards

 

[Tesselator]

Performance requires compromise. With you're wallet. 😱 😀 😛

In the case of the controller, if it determines if a drive is problematic, it elliminates it from the array, and will rebuild it (assuming the array type is equiped with some level of redundancy). In the case of a single drive, the drive becomes unresponsive until the drive is corrected (remapped). This can take a bit, up to a few minutes. Not acceptable for a system meant for high availability.

TLER values are meant for RAID only, and help keep the array up and running with fewer instances of a drive being determined as "problematic". It's useless in a consumer/single drive system, as there's no controller to handle the processing, or software/driver to place the load on the CPU. Hence the default value of (0s for both read and write). The 7s is for reads BTW. Writes are set at 0s. Forgot to distinguish that. 😱

In the case of a software RAID, such as OS X, the error recovery can be handled the same way. The functions are built into the software, not the SATA controller chip soldered to the board. The CPU does the work. Eats clock cycles, but is cheaper than a hardware solution.

Wow, thanks Nano! That was nice (and thorough) explanation. It seems to put all the technicals in place. Of course that leaves me with the general question: When would you recommend using RAID class drives?

Regardless of the wimpier self-recovery are they going to act pretty much the same for the same length of time (or longer) than non RAID class drives in a single drive system and/or in a RAID array. I mean other than being faster of course. 🙂

 

[jons]

I buy all my drives from Newegg.com

Same here. 640GB AAKS or black FTW!