[saltines17]

I'm looking to add a second HD to my G5. On OWC, I see:

1) 250GB IBM drive with 8.5 ms seek time and 8 mb buffer

2) 300GB Maxtor drive with 9 ms seek time and 16 mb buffer

I could search Google for an explanation on seek times and buffers and the respective benefits of each, but I'd rather, if no one minds, get some quick input on how much of a factor these things are in the real world and whether the $80 extra for the Maxtor is justified.

Thanks

[Catfish_Man]

If it's one of the new Maxtor drives it support native command queuing as well. Dunno if the G5 does though.

[Detrius]

Originally posted by saltines17:
I'm looking to add a second HD to my G5. On OWC, I see:

1) 250GB IBM drive with 8.5 ms seek time and 8 mb buffer

2) 300GB Maxtor drive with 9 ms seek time and 16 mb buffer

I could search Google for an explanation on seek times and buffers and the respective benefits of each, but I'd rather, if no one minds, get some quick input on how much of a factor these things are in the real world and whether the $80 extra for the Maxtor is justified.

Thanks

Seek time:

7200rpms = 7200 revolutions/minute

1/(7200r/m) = 1m/7200r = 1m*60sec/(7200r*1m) = 60sec/7200r = .0083sec/r

Therefore, 7200rpm = 8.3ms seek time

Quick answer: the seek time is the average amount of time it takes to locate a single piece of information on the surface of the disk.

If both drives are 7200rpm, then the different companies probably rounded differently.

The buffer is RAM that is physically located on the drive. It's the disk cache. It stores recently accessed data. The larger the buffer, the better.

[saltines17]

Originally posted by Detrius:
Seek time:

7200rpms = 7200 revolutions/minute

1/(7200r/m) = 1m/7200r = 1m*60sec/(7200r*1m) = 60sec/7200r = .0083sec/r

Therefore, 7200rpm = 8.3ms seek time

Quick answer: the seek time is the average amount of time it takes to locate a single piece of information on the surface of the disk.

If both drives are 7200rpm, then the different companies probably rounded differently.

The buffer is RAM that is physically located on the drive. It's the disk cache. It stores recently accessed data. The larger the buffer, the better.

Ah, I see. It didn't occur to me that seek time is just a simple calculation prone to rounding differences.

So, since the buffer is double, do you think the $80 extra for the Maxtor is worth it?

Thanks

[kupan787]

Originally posted by Detrius:
Seek time:

7200rpms = 7200 revolutions/minute

1/(7200r/m) = 1m/7200r = 1m*60sec/(7200r*1m) = 60sec/7200r = .0083sec/r

Therefore, 7200rpm = 8.3ms seek time

Quick answer: the seek time is the average amount of time it takes to locate a single piece of information on the surface of the disk.

If both drives are 7200rpm, then the different companies probably rounded differently.

I think you just calculated latency, which IIRC is different from seek time.

Sometimes latency is not even mentioned specifically at all, but the average can always be calculated using this formula:

(1 / (SpindleSpeed / 60)) * 0.5 * 1000 (Latency Calc)

With regards to seek time, almost all IDE/ATA 7200 RPM drives shipping after 2000 have an average seek time specification of 8.0, 8.5 or 9.0 milliseconds. You must realize that of the four components that comprise the drive's access time, if you are comparing two drives of the same class and spindle speed, only seek time will differ much between them. (Seek Time)

So it is not a "rouding difference". Those drives actually have different seek times.

[saltines17]

Originally posted by kupan787:
[B...[/B]

Ah, the plot thickens, I see.

Is there an appreciable difference among 8, 8.5, and 9 ms seek times?

As of now I'm eyeing the Maxtor drive, which is $240 at newegg.com.

Anything specific I should know about the drive, or is it generally accepted as a good thing?

Thanks again

[kupan787]

Originally posted by saltines17:
Ah, the plot thickens, I see.

Is there an appreciable difference among 8, 8.5, and 9 ms seek times?

Honestly, for a consumer to prosumer system, probably not. The hard drive is still about 1000 times slower than accessing RAM (not an exaggeration either). If seek time was vitally important, NewEgg sells a SATA WD Raptor (74GB) with a seek time of 4.5ms. But here you are sacrificing storage space for speed.

I don't know exactly how much buffer comes into play as far as perceived speed goes. It can only help with opening the same thing multiple times. Otherwise if it is something new, it has to go read it from disk, and buffer doesn't matter. So I guess a larger buffer is generally better. $80 is getting you 50 GB more, and double the buffer space.

[saltines17]

Originally posted by kupan787:
Honestly, for a consumer to prosumer system, probably not. The hard drive is still about 1000 times slower than accessing RAM (not an exaggeration either).

I don't know exactly how much buffer comes into play as far as perceived speed goes. It can only help with opening the same thing multiple times.

Thank you for the help so far.

Forgive my ignorance, but I'm a bit confused as to what kinds of files are in the 8-16MB range and are accessed so often that the buffer helps at all.

Thanks again

[Catfish_Man]

Originally posted by saltines17:
Thank you for the help so far.

Forgive my ignorance, but I'm a bit confused as to what kinds of files are in the 8-16MB range and are accessed so often that the buffer helps at all.

Thanks again

OSX has huge numbers of little files it uses (app caches, fonts, colorsync profiles, daemons, etc...) so it seems like a large buffer could help with those.

[chabig]

Kupan is right. Seek time is the time it takes (on average) for the head to move to the correct cylinder. Latency is the average time it takes for the data on a cylinder to rotate under the head.

Chris

[Oneota]

16 MB buffer, eh? Hm..if that weren't a Maxtor drive, I'd be really tempted, too. I've just had bad experiences with Maxtors. Maybe they've gotten better lately.

[saltines17]

Originally posted by Oneota:
16 MB buffer, eh? Hm..if that weren't a Maxtor drive, I'd be really tempted, too. I've just had bad experiences with Maxtors. Maybe they've gotten better lately.

I've tended to drift towards IBM/Hitachi myself, but since my G5s existing HD is apparently a Maxtor, I figure, why not?

I'm going to go with the Maxtor drive.

Now, since (I assume) my existing internal drive has an 8MB buffer, the second drive has a 16MB buffer, and the only thing the buffer helps with is small OS cache files, would it make more sense to run the OS on the 16MB-cache-drive?

Thanks again, everyone

Edit: Sigh, I don't know. A link from barefeats.com brought me to these tests, which basically say it's not as fast as the IBM drive (more so for server use than for single-user use). http://www.storagereview.com/article...B300S0_sp.html

I'll wait a little longer, then I'll see what I feel like doing

[kupan787]

Originally posted by saltines17:
Now, since (I assume) my existing internal drive has an 8MB buffer, the second drive has a 16MB buffer, and the only thing the buffer helps with is small OS cache files, would it make more sense to run the OS on the 16MB-cache-drive?

Correct, since the buffer can only effect files on that particular drive. I don't know how the buffer works exactly, if it is drive controlled or OS controled. So I don't know if it is just a buffer of recent files, commonenly used files, or if the OS can say "put this in the buffer".

[saltines17]

Originally posted by kupan787:
Correct, since the buffer can only effect files on that particular drive. I don't know how the buffer works exactly, if it is drive controlled or OS controled. So I don't know if it is just a buffer of recent files, commonenly used files, or if the OS can say "put this in the buffer".

All right. Forgive me for all the questions, but more seem to come to mind as soon as one is answered.

Question #1:
I originally intended to have the OS, Applications, and whatever on the stock 160GB drive and use the 250GB drive for extra files and as a scratch disk for Final Cut, Photoshop, and for Cinema4D render files.

Having the OS, Applications, and wahtever on a 300GB drive with a 160GB drive for a scratch disk seems kind of odd and inefficient. Perhaps a 160GB scratch disk is sufficient?

Question #2:
Do the "hundreds of little files" that would go in the buffer ever get stored in RAM? My question is really: if I have 2.5GB of RAM, would these "small files" even go into the buffer often enough to even consider buffer size?

Thanks very much, again...

[Catfish_Man]

Originally posted by saltines17:


Question #2:
Do the "hundreds of little files" that would go in the buffer ever get stored in RAM? My question is really: if I have 2.5GB of RAM, would these "small files" even go into the buffer often enough to even consider buffer size?

Thanks very much, again...

OSX will cache everything it can in ram, and it doesn't really get a say (as I understand it) about what the hard drive buffers (I think it buffers at a much lower level than files. Possibly individual reads, or small chunks of files or something). Presumably 2.5GB is enough to avoid touching the HD much, but I'm not really sure.

[saltines17]

Originally posted by Catfish_Man:
OSX will cache everything it can in ram, and it doesn't really get a say (as I understand it) about what the hard drive buffers (I think it buffers at a much lower level than files. Possibly individual reads, or small chunks of files or something). Presumably 2.5GB is enough to avoid touching the HD much, but I'm not really sure.

Okay, thanks.

I ordered the Maxtor drive. We'll see how it goes. Thanks everyone.

[awrc]

Maybe I'm just way out of date but I always thought that seek time was the average time it takes the drive head to get to the correct physical track on the platter.

So the time taken just to get the drive head to the right place is the seek time, plus whatever rotational latency is involved in waiting for the drive to spin to the right position, then you have to add the time it takes the head to move over the data. The drive's rotational latency is dependent on the spin speed (i.e. 7200rpm). This also has an effect on the time taken to actually read the data and you also have to factor in the number of platters that make up the drive.

If you look at the online specs for a typical drive you'll see that the actual data rate between the drive and the drive buffer can be a heck of a lot faster than the supposed drive speed. Example - WD Caviar SE, 250GB, 7200rpm according to the specs on Western Digital's site can transfer data between drive and buffer at 737 Mbit/s.

Average seek time is skewed a bit by the fact that drives have a constant rotational velocity (wasn't it old Mac floppies that varied the drive speed so that the medium had a constant data density?) so the outer tracks store more data than the inner tracks, so the average seek time isn't just half the maximum seek time.

I don't know to what extent typical modern drive controllers make use of techniques from the days when drives were the size of washing machines, but other clever tricks, like reading the closest bit of data in a read (rather than waiting for the drive to rotate to the correct position) and then rearranging stuff once it's in the buffer can help, and although I doubt it's of much benefit in consumer drives, re-ordering drive access requests to minimize seek time can improve overall drive performance at the expense of potentially slowing down individual read requests. However, once you get into that stuff you need to start with queueing theory to predict performance.

Given two drives with the same spin speed and same buffer size, I'd look at the number of platters (or the platter density) next. More platters generally means faster transfer between drive and buffer, but more platters also means there's a higher risk of drive failure. If anyone remembers those 80GB WD drives from four or five years ago that had such a lousy reliability record, yet people kept buying them because they were the fastest drive around, that's what you get when you make an 80GB drive from 4 20GB platters.

Al

[saltines17]

Hey again - Just installed the new Maxtor HD. The installation was rather shockingly simple, and everything seems to be running fine.

Thanks for all the help!