External Firewire 800

[LunarMist]

I have seven internal hard drives as well as one internal removable bay and one external removable bay to fit another nine drive trays. With the amount of data I have, shuffling files and drives around it getting ridiculous, so I am considering a 4-drive external 1394b case. Initially, it would be used for two separate sets of 200GB drives; 400GB data and 400GB backups. If these were all backups I could live with 1394a. Are there any known problems with the Oxford 922 1394b-based systems? Top-notch performance is not neded, but I want better speeds (esp. writes) than 1394a provides. I am also looking for a FW PCI card that has both external FW800 and internal FW400 if that is possible, or does the whole bus slow to the lower speed?

Thanks.

 

[Pradeep]

I have an OWC 64 bit FW800 controller, currently in a 32 bit slot. Gives me three external ports. I have another controller for FW400. I believe there are cards that also include a seperate FW400 chip, so both types of devices can be enjoyed at max speeds.

Hooked up to it is an OWC FW800 3.5" enclosure, using some kind of Maxtor 160GB drive. IIRC I was getting a good 50-55MB/sec when the disc was empty.

 

[Santilli]

I've been looking into this a bit. Problem with 1394B is it requires a 64 bit slot, IIRC.

I really wonder if cost wise, a SATA hotswap system wouldn't be the way to go.

http://www.granitedigital.com/catalog/pg53_satahot-swapbackup.htm

I wonder if this system uses IDE drives? Can't tell from the posting.

Further down they offer complete systems with ide drives, so I think it's time to call Ken on this one.

I did see card with internal and external ports.

http://www.firewiredirect.com/firewire/products/adapters_pci800.shtml

looked pretty cool.
Hope this helps..
s

 

[LunarMist]

I ordered a bunch of 1394b gear today, which should be here later in the week. The performance loss in a 32-bit slot is curious. I'll let everyone know how it works out.

 

[CityK]

Problem with 1394B is it requires a 64 bit slot, IIRC.

1394b: 800Mb = 100MB .... so 32bit would suffice, but not a lot of head room for other devices on the bus....see P's post above

 

[Santilli]

Odd part is, I think it's OWC that had the information about the 64 bit slots...

http://eshop.macsales.com/Catalog_Item.cfm?ID=5777&Item=OWCFW800PCI3

FireWire 800: Macintosh with 33-MHz/64-bit PCI slot. Power Macintosh Blue & White G3* or later, all AGP Power Macintosh G4 systems fully supported. Compatible with the Power Macintosh G5! FireWire 800 requires Mac OS X 10.2.3 or later, Mac OS X 10.2.5 or higher recommended. FireWire 800 is NOT supported under Mac OS 9.
FireWire 400: Must use 9 pin to 6 pin adapter cable (not included). Macintosh with 33-MHz/32-bit PCI slot. FireWire 400 requires Mac OS 8.6 or later. If used in a 33-MHz/32-bit PCI slot, FireWire 400 speed will be maximum attainable.
*Blue & White G3 / Power Macintosh G4 "Yikes" will not see maximum performance with this card due to an architecture flaw in the logic board design. There is nothing that can be done to rectify this situation.

Windows System Requirements: FireWire 800: Windows PC with 33-MHz/64-bit PCI slot. FireWire 800 requires Microsoft Windows 2000 or XP.
FireWire 400: Must use 9 pin to 6 pin adapter cable (not included). PC with 33-MHz/32-bit PCI slot. FireWire 400 requires Windows ME, 2000 or XP. If used in a 33-MHz/32-bit PCI slot, FireWire 400 speed will be maximum attainable.

 

[LunarMist]

1394b: 800Mb = 100MB .... so 32bit would suffice, but not a lot of head room for other devices on the bus....see P's post above

Of course that would be the case, but I was concerned about something worse. The 1394a write speeds that max out at a sucky 29MB/sec are the main problem.

 

[Santilli]

Keep us posted.

s

 

[CityK]

That's awefully noble of OWC to be straight up about the requirement. Its a shame other manu's aren't as forthright and don't spell out the performance oddities affecting the wares they peddle.

If used in a 33-MHz/32-bit PCI slot, FireWire 400 speed will be maximum attainable...... "Yikes" will not see maximum performance with this card due to an architecture flaw in the logic board design. There is nothing that can be done to rectify this situation.

Sounds like some sort of quartz crystal limitation in the design. I wonder how many other cards are similarly affected.

Lunar, what external drive bay did you end up getting?

 

[Pradeep]

That's awefully noble of OWC to be straight up about the requirement. Its a shame other manu's aren't as forthright and don't spell out the performance oddities affecting the wares they peddle.

If used in a 33-MHz/32-bit PCI slot, FireWire 400 speed will be maximum attainable...... "Yikes" will not see maximum performance with this card due to an architecture flaw in the logic board design. There is nothing that can be done to rectify this situation.

Sounds like some sort of quartz crystal limitation in the design. I wonder how many other cards are similarly affected.

I believe Yikes was a Mac mobo with horrible PCI thruput.

 

[Santilli]

Another example of apple using substandard chips on their motherboards.

Beige G3 was a double case of this: IDE channel only does 16 mb/sec,
and the grackle motherboard chipset limits the PCI bus to 75 mb/sec.

It got even worse. The Blue and whites, IIRC, couldn't even do 55 mb/sec, thanks to an even worse chipset.

s

 

[Santilli]

Another problem is insufficent room, I think it's in ROM, for compressed PCI drivers to load. The beige has this problem. It has only 3 slots, but if the drivers are compressed, their is insufficent memory to allow them to load,=frozen machine.

Until someone finally tells you the truth, you think it's a conflict between the different cards. I spent nearly 700 bucks on a pci extension box, so I could use more cards in my beige. The only cards I could use were other video cards, that used the same drivers (ATI). Talk about frustration of purpose...
Reasons I won't buy another mac...

s

 

[LunarMist]

Grrr! Why is it that nothing simply works without causing some other, seemingly unrelated, difficulties? I hate computers.

 

[Santilli]

They certainly have their moments...

Guess that's why I like taking Splash's extensive experience, both shopping wise, construction wise, etc. along with the other members here, and going with that. Better to learn from others, then pay 3 grand for a promiseless array that never works...

s

 

[Santilli]

LM:
You having problems with the 400 and 800 combination?
s

 

[Pradeep]

Speaking of Promiseless, we recently installed two of the I guess SX6000 RAID controller cards in a single machine. One RAID 5 array on each ( I think five or six 250GB Maxtors in each). Each array accessed individually was fine. But as soon as both arrays were accessed simultaneously, the server went tits-up. When the controllers were put in individual machines, they worked fine, with excellent performance that I wasn't expecting. Shame SCSI storage costs so much when you get into the TB range Perhaps one day... a few petaboxes.....

http://www.archive.org/web/petabox.php

 

[Santilli]

ROFL :excl: :excl:

Pradeep, it's so funny to read that. I fell for the same trap, when their cards came out. They advertized, in writing, and in their tech docs, that you could use two cards in the same machine. Bought two, and somethings never change...

Never did get those POS cards to work right, and spent a couple thousand trying to get it to work. They get away with it, because there are no damages. The filing fee is more then their pos card, by about 5 times when I checked. Can't really go after the tech who worked on it...

I've been having problems copying large files onto my Maxtor drive, DM9.
Keeps stopping, about 3/4 of the way through. I guessed cable, and I was right. It was about 36 inches long???? and didn't work. Don't know where it came from. I just got done throwing away about 10 cheap scsi
narrow cables, and cleaning out my box of used ide cables. It appears to be working but, during the copying, the it's using about 29% cpu utilization(during writes?), which is a lot for this machine. It's hammering the second processor, according to WTM. I just got my wallet out, and I'm going to spend the 170 dollars for a Supermicro sca/scsi from Hypermicro, via new egg.

I see so many deals on new, and used, scsi drives, 80 pin SCA's, that I would be much happier then with any other form of storage, until SATA-2 comes out, and the drives have the stability of SCSI.

Replaced the long cable with a short, way inside spec, premium cable I bought long ago, and barely got it to fit, but it fits, and the copying seems to be going well. Have to run diagnostics on the drive when the copying is done.

I'm with LM, sometimes I really hate computers...

s

 

[LunarMist]

LM:
You having problems with the 400 and 800 combination?
s

Not exactly. The original plan was to move two of the old internal 1394a backup drives to an external 4-bay 1394a enclosure with (two removable drives). That would create space for two more internal primary (IDE) drives. There would also be two more backup drives added to the external enclosure (total of four drives in the enclosure). Unfortunately it is no longer possible to access more than five internal drives (combination of IDE, S-ATA, SCSI) when 4GB of RAM is installed. Then I decided to use 1394b instead so that I could add some internal primary drives. But now there are two other problems: 1) only two of three 1394b bridge boards with six total drives are recognized at one time and 2) there were some terrible read performance problems with the 1394b drives, which were ultimately traced to NAV. Apparently there is only a small delay in reading small files (~100MB) since they fit in RAM, but standard files (~1.35GB) can cause crashes. Anywho, I decided to leave all the internal drives as is for now, added a fifth (primary) drive to the 4-bay case on the 1394b bridge, added a 1394a bridge to the case for one of the backup drives and am still using one separate 1394a case with a removable drive. I don’t like having some of the primary and backup drives online together, but it works well enough and I can access up to ~2.7TB at one time, spread over one 1394b and two 1394a controllers. At least it is better than swapping removable drives all the time, yet still provides multiple bays for removable offsite backups (another ~1.1TB of drives). I’ll probably find a cheap 2-bay case for the extra 1394b controller and add a couple of drives to that eventually. The first step is to replace a couple of the 250GB internal drives with 400 giggers.

 

[mubs]

LM: Talk to Handy. He'll sell you en EMC array at a discount.

Unfortunately it is no longer possible to access more than five internal drives (combination of IDE, S-ATA, SCSI) when 4GB of RAM is installed.

Pray explain why?

 

[Santilli]

I'm with Mubs, see above..
s

 

[Santilli]

Isn't Splash getting closed to the addressing limit of 32 bit?
s

 

[LunarMist]

LM: Talk to Handy. He'll sell you en EMC array at a discount.

Yeah, but I'm too broke to buy anything like that.

Unfortunately it is no longer possible to access more than five internal drives (combination of IDE, S-ATA, SCSI) when 4GB of RAM is installed.

Pray explain why?[/quote]

I also assume there is not enough address space. It is strange that the hard drives are affected and it does not matter which controller, southbridge IDE or SATA, onboard IDE or SATA, add-in PCI IDE controller card is used.

 

[Platform]

Unfortunately it is no longer possible to access more than five internal drives (combination of IDE, S-ATA, SCSI) when 4GB of RAM is installed.

Pray explain why?

It's likely that there is now physical memory in an address region that *was* being used by 32-bit PCI memory mapping techniques for I/O. It is certainly possible run into these sorts of problems using 4 GB of RAM in combination with 32-bit PCI. Hence, the need for 64-bit technology like PCI Express (and 64-bit parallel PCI cards plugged into PCI-X) !

Otherwise, there are certain lower address regions that are reserved by BIOS early on for storage I/O (ATA) during boot-up that are not going to be affected by 4 GB of RAM.

By the way, I suspect that if you installed 3 GB of RAM instead of 4 GB, you would not be able to tell a bit of difference when doing whatever you are doing on that computer.

 

[LunarMist]

Yeah Gary, we need 64-bit computing all around, including Windows. I could probably live with 3.5GB since only ~3800MB are available anyway. Not anticipating the future problem, I sold the 512MN RAM a while ago so I can not test 3.5GB. 3GB is not always sufficient for good performances - tried that before. I'd really rather buy fewer, larger drives. Where is the damned 7K400?

 

[mubs]

It's likely that there is now physical memory in an address region that *was* being used by 32-bit PCI memory mapping techniques for I/O.

When you say "physical memory in an address region" do you mean number of drives or total storage capacity? If the latter, then replacing smaller drives with larger ones won't help, will it?

 

[LunarMist]

Whether it is a 250GB Maxtor or WD740GD, the drive size makes no difference. The sixth and higher drives detected in the BIOS are not accessible. The order in which they are detected obviously depends on the controller, southbridge, on-board, or PCI card.

 

[mubs]

So I guess number of devices that can be addressed is the problem, not the storage capacity of the devices in total.

Because NT class OSes don't use the BIOS, looks like the limitation is in the OS, not the BIOS. or maybe the hardware?

What do you do on your machine that you need 3.5 - 4GB, LM? If you have to kill me after telling me, I'd rather not know.

 

[LunarMist]

So I guess number of devices that can be addressed is the problem, not the storage capacity of the devices in total.

Apparently that is the case, but I am not sure about the SCSI drives, since there is only the one 15K.3.

What do you do on your machine that you need 3.5 - 4GB, LM? If you have to kill me after telling me, I'd rather not know.

I've already revealed too much.