Lengthy questions about PM G4 Hard Drive options

[aphetica]

My machine is a PM G4, 2GB ram, Dual 1.44ghz, FW800.

ATA/66 - Maxtor 5400rpm, 20gb less than 2mb cache (secondary)

ATA/100 - Hitachi 7200, 120gb 8mb cache (main)

A question for now:

I'm having an issue with my secondary hard drive. Whenever I try to open ANYTHING off of it, even a small JPG, it lags the entire system and I can't do anything else on that drive until it finishes with its little spasm. They last between 10 seconds and 10 minutes.

It's only about 50% full so I don't think it has anything to do with capacity.

I know it has a small cache, but it was never this slow in any of my other machines.

Could it be because I have it on the ATA/66 bus? If I moved it to the ATA/100 bus, would it be quicker? Could it end up lagging my main drive?

A question for later:

Also, when I get around to replacing these drives, would this model be able to make efficient use of a 16mb cache on an ATA HDD, or would the hardware restrict it from fully utilizing that much cache?

I've thought about maybe getting an SATA card also. Are the pci slots even fast enough to utilize SATA300? Can I boot from SATA?

I don't know. I'm sure I will end up with separate drives for data (atleast 500gb) and OS when I can afford it, but I am not sure what configuration to use.

I need speed.

What would you do?

 

[Kami]

Do you have the drives set up as Master (primary) and Slave? Or Cable Select on both drives?Do you have any other devices (e.g.CD-ROM) on the ATA 66 and ATA 100 buses?

 

[aphetica]

Do you have the drives set up as Master (primary) and Slave? Or Cable Select on both drives?Do you have any other devices (e.g.CD-ROM) on the ATA 66 and ATA 100 buses?

The CD-ROM is on the ATA33 bus.

I tried half a dozen different jumper configurations, including master/slave, and cable select on both drives.

I tried moving the secondary drive to the ATA100 with the primary drive, with no change other than that it locks up BOTH drives now.

Tried a spare 30GB drive on the ATA66 bus and it worked beautifully, moving a GB of data in under a minute, to and from the main drive.

The secondary (20GB) seems to have trouble moving data FROM the drive. Getting it on there is no problem. Getting it OFF takes 15 minutes for only 3mb.

I'd like to transfer its contents (a mere 11GB of data) to the 30GB drive, but at the speed it's moving, it'll take 3 days.

 

[Franklinstein]

Your Maxtrash hard drive sounds like it's nearing death. Backing it up and replacing it would be the best course of action.

With Ultra ATA, the transfer rate will match that of the slowest drive, so your ATA/100 drive would clock itself down to ATA/66 speeds to match your old drive.

Caches speed up disk accesses, especially for non-sequential writes. The caches are under the complete control of the hard disk's microcontroller. The host controller can issue a command to disable caches in certain scenarios (such as when used as a removable disk), but that's about it - the computer's host controller doesn't interact directly with the cache and therefore doesn't really have anything to do with recognizing or using caches.

Another aspect to consider is the drive's transfer rates. Usually, only the fastest hard drives have transfer rates that can saturate a SATA/150 interface, let alone a SATA/300. The enterprise-class 15,000 RPM Seagate Cheetah 15k.6, for example, has a sustained media transfer rate of 164 MB/sec, where SATA/300 can handle 300MB/sec. (the Cheetah is used only as an example; it doesn't use SATA). As you can see, with a consumer 7200RPM drive such as the 1TB Seagate Barracuda 7200.11 (with its 105MB/sec transfer rate), SATA/150 would be more than adequate.

As for the PCI transfer speed, standard PCI (32-bit, 33MHz) maxes out at 133 MB/sec. 64-bit 33MHz PCI caps out at 266MB/sec.

If everything worked perfectly, unless you went with the enterprise-class drives, your bottleneck would remain the hard drive, at 105MB/sec with the Barracuda example.

I don't know about booting from an add-in SATA card in a Mac, you'd have to check the manufacturer's website for that. I think Sonnet makes an SATA card...

 

[MacJunky]

There are a few SATA cards that support booting if I recall. I have noticed that a number of SATA cards also have the Mac driver in their firmware as well. It is not hard to find them online.

 

[aphetica]

I named this drive Pepito, and in the end, it really did end up being the spawn of Satan.

(see: http://en.wikipedia.org/wiki/Squee#Pepito for details)

It took 32 hours to move 11gb of data off the drive.

The entire time, my system was nearly comatose. I couldn't even use Adium.

At least it afforded me plenty of time to work on music and clean my apartment.

The drive is a Maxtor D540X-4K, 5400rpm, 2mb cache, 12ms seek.

I've read in a couple places that this particular drive is known for this sort of thing.

What a pain in the butt.

Thanks for all the info, Franklinstein.

Like you said, I think it's failing. I'm going to check it with some kind of diagnostic tool later to make sure... then pull it apart for the metals. Some things just deserve vengeful disassembly.

 

[coius]

I named this drive Pepito, and in the end, it really did end up being the spawn of Satan.
(see: http://en.wikipedia.org/wiki/Squee#Pepito for details)

It took 32 hours to move 11gb of data off the drive.

The entire time, my system was nearly comatose. I couldn't even use Adium.

At least it afforded me plenty of time to work on music and clean my apartment.

The drive is a Maxtor D540X-4K, 5400rpm, 2mb cache, 12ms seek.

I've read in a couple places that this particular drive is known for this sort of thing.

What a pain in the butt.

Thanks for all the info, Franklinstein.

Like you said, I think it's failing. I'm going to check it with some kind of diagnostic tool later to make sure... then pull it apart for the metals. Some things just deserve vengeful disassembly.

Yeah, I have had the same issues with other Maxtors, and a few with some others (early Seagates and Fujitsus) and it's pretty much really slow access. It's actually common that what is dying is the controller board, not the actual platters. It keeps screwing up the data transfer and locks up the bus when it keeps making errors (like constantly requesting the bus and sending bad data) and makes the system sluggish to the point that sometimes it takes several seconds for the mouse to redraw. I have seen it happen in windows and Linux as well. So what you really aren't seeing is a failure of the platters or read head, what you are seeing is the controller board.

Anyways, that bit of info is not going to help. What WOULD help is if you can get another drive like it, and replace the controller board, you would get the data off easily.

 

[aphetica]

Anyways, that bit of info is not going to help. What WOULD help is if you can get another drive like it, and replace the controller board, you would get the data off easily.

There's not much point now really, since it's all done and over with. I'll keep that in mind though if I'm ever in this situation again.

 

[Franklinstein]

What WOULD help is if you can get another drive like it, and replace the controller board...

I always keep the PCB from crashed hard drives; I've got a few boxes full of them, from brands such as Conner, DIGITAL, IBM, Micropolis, Quantum, Seagate, and WD. Useful for converting drives from IDE to SCSI (when possible, such as between similar versions of old Quantum Fireballs) and for resurrecting drives that have blown their original board. I've seen several drives with charred microchips that work fine when the boards are replaced with one from the same model.

It's possible that the controller could be failing, but since it's a Maxtrash, it's more than likely a problem within the HDA. Often, as a drive ages, the platters or heads degrade to a point where they can no longer properly register the magnetic signals. The controller attempts to figure out what the signals it's receiving are supposed to be (1 or 0), and if it can't readily determine that, it'll reread the same sector or cylinder over and over until it gets some consistent results. Then it'll accept the result and do whatever it's supposed to do with it, and move on. If the drive is failing, the read/reread cycle will continue as it encounters even more sectors that it can't get a positive reading for. PRML drives can be pretty bad about that, especially the cheap ones.