Amiga.org
Amiga computer related discussion => Amiga Hardware Issues and discussion => Topic started by: d0pefish on July 28, 2006, 11:01:32 AM
-
Hi there,
After deciding to try the SCSI route I'm having a few problems with my new drive.
I bought a Quantum Atlas 10K II which is supposed to be 18.2 GB. I have a Foxconn SCSI cable with 6 plugs on it, and two Foxconn matching terminators.
My chain goes:
TERM > Cyberstorm > NOTHING > NOTHING > Hard Disk > TERM
At the moment I have enabled Terminator Power and Force Single Ended mode on the hard disk. The problems started before I set these options though, and I think it's still a bit flakey.
When I first set out to partition the hard drive, HDToolBox read the drive geometry and came up with 16.9GB every time. I was curious about it not being 18.2GB, but ignored it and continued to partition the drive. I put SFS in the RDB and set my partitions accordingly.
Having copied everything across, I switched off and booted without the old IDE drive connected. No problems there, it booted up to my desktop fine.
After using it for a few hours, I started getting "Parity error" and "General error" regarding the drive - I'd then to end up switching off and attempting to restart. At times the hard disk wouldn't be recognised and I'd get the purple disk insert screen, no matter how many times I switched on and off. After waiting half an hour it booted up again. :-?
But now I have an even more worrying problem. The last partition on my drive is now "Uninitialzed" and not readable. And now when I read the drive geometry using HDToolbox, I get 16.0GB!! So that's another 900MB gone out of the blue.
I've tried different cable arrangements (shortening the distance between controller and drive) but no difference.
Other than the problems, the drive loads things very quickly and it makes short work of a window with hundreds of icons.
But can anyone help? What the hell is going on with the geometry? Can I do some sort of low level format and restore it as it should be or...?
What can I do? :cry:
Thanks!
-
Harddrive manufacturers count GBs of capacity as 1GB = 1*10^9Bytes (1000* 1000 * 1000 Bytes), whilst computers count GBs of capacity as 1GB = 2^30Bytes (1024 * 1024 * 1024 Bytes).
So: (18.2 * 1000 * 1000 * 1000) / (1024 * 1024 * 1024) = 16.95GB.
I cannot explain the 16.0GB though, except that HDToolBox can be a bit wanky sometimes. More important though - did you loose the last partition?
With 20MHz SE SCSI as the CSPPC is capable of, the maximum bus-length is 1.5M and it is from end to end which counts. In my experience this 20MHz SE SCSI is rather sensitive and I have had troubles with my previous cabling which was somewhere between 1 - 1.5M, even if the harddrive was very close to the controller (the terminators at the ends were still far from the controller and harddrive), so try a shorter cable and see if it makes any difference.
You should not need to set the drive to force SE mode, as only SCSI equipment manufactured wrong wont autosense.
Also, do you have the most recent official CSPPC firmware (http://www.aminet.net/package.php?package=biz/p5/Flash-300399.lha)? Else the older drivers can be a source of issues with the SCSI bus.
/Patrik
-
When I first set out to partition the hard drive, HDToolBox read the drive geometry and came up with 16.9GB every time. I was curious about it not being 18.2GB
Welcome to marketing gigabytes:
18.2*1000*1000*1000/1024/1024/1024 = 16.95
Can I do some sort of low level format and restore it as it should be or...?
Whatever you do, don't try to low level format. Modern drives are low level formatted at the factory and can't be re-low-level-formatted.
What can I do?
I'd recommend you use SCSIConfig instead of HDToolBox. It is the official tool provided by the manufacturer.
However, some of the problems you have sound more like some hardware problem. If you have some way of making sure the HDD is ok (SCSI controller on a pc?), that would help.
-
Hi Patrik + Piru - thanks for the fast replies.
So 16.9 was right - that's good news then. Thanks for explaining about how size is calculated! :-)
The partition is still there but is not readable. It shows up in HDToolbox still. I had a quick peek at SFSSalv and it showed my files there but I didn't go any further. It's not important if I lose the files on the last partition as it was just a small storage space.
My cable length is 80CM, so while there's plenty connectors on it, they are reasonably close together.
Can I terminate in the middle of the cable? (leaving the spare plugs at the end)
Or, is it a good idea to chop the cable down and have only the connectors I need?
I have the firmware which has a "PCI" button - I think it was for GREX (though I don't have one). I'll try flashing the firmware you linked to. I'll also remove the force jumper.
I'll give SCSIConfig a go... and NOT touch anything related to lowlevel formatting. :-)
Unfortunately I don't have a way of testing the drive on anything else. However, I got the drive from eBay and the seller was very descriptive. He showed screen shots of numerous tests he did. When it arrived it was very well packed and I don't think it could have been knocked. It looks brand new, too.
Here's the listing if you want to read the tests:
http://cgi.ebay.co.uk/ws/eBayISAPI.dll?ViewItem&ih=019&item=290007318815
Thanks once again!
-
For me it sounds like the disk is getting worned out... When you mention, works - then dies - wait two hours - working again, I have seen disks behaving like that (before they die a silent death..)
You cant be sure before you try with another scsi drive though :(
-
@d0pefish
sorry to state the obviuos but... are yours active terminators?
if you want to take a look at the BadBlocks list, download Q-Device from Aminet
http://it.aminet.net/disk/misc/q-device.lha
-
Ohhhhh damn. :cry:
I just had a look with Q-Device, and on the Diagnostic tab, I used Show defects.
512 defects, and it lists all the cylinders etc.
Is this drive faulty?
-
It could be, the symptons you described are very common for disk going dead :(
The best would be to try another drive (as i allready said :))
-
AFAIK 512 badblocks are ... a lot :-(
Even if i have 2 SCSI HDs on my CSPPC (same FlashUpdate of yours) and the "less knackered" (less bad blocks), is the one giving problems. (and in the last partition here too)
Unfortunately this happens when we purchase used HDs
Here i have de-activated the last partition waiting for the final BANG! ;-)
-
Well, looks like I'll have to find another drive and look into returning this one. :-( Such a shame as I was starting to really like the high speeds until all the errors began. Obviously I can't rely on this drive.
I even gritted my teeth, got SCSIFormat from Aminet and did a complete wipe including the option to clear the defect table... no difference, still showing up as 16.0GB.
Anyone got any recommendations for other hard drives to use with the CyberStorm? Something equal or better than the Atlas 10K II preferably... 18gb upwards? What are YOU using with your CyberStorm SCSI? :-)
Noise isn't an issue, I just want something quick. Are there any drives that spin up faster than the Atlas 10K II? One thing I notice with this drive is that it takes a little while to spin up.
Thanks for the help guys, I appreciate it. :-D
-
i have:
1- 9GB 7200rpm Fujitsu MAB3091SP (68pins)
1- 18GB 10000rpm COMPAQ BD01864552 (80pins SCA with adapter)
The faster is clearly the COMPAQ (PFS3 and SFS like filesystems)
The problem still to find a reliable seller (possibly not used or at least NOS)
I personally have purchased them by a friend here in Italy about 2 years ago. They were both used but in "good" condition.
Good luck :-)
-
I even gritted my teeth, got SCSIFormat from Aminet and did a complete wipe including the option to clear the defect table... no difference, still showing up as 16.0GB.
This may be a long shot, but being a reasonably modern HD, it may have the so-called 'SMART' system, automatically mapping bad blocks and preventing them from being used. Maybe the missing space is due to the large amount of bad blocks?
Typically, used SCSI drives are often scavenged from old servers, which means they've had a busy life, and might not live for too much longer.
-Paul
-
Things are looking slightly better now.
I emailed the seller of the drive and he gave me a nice response. He thanked me for not dumping negative feedback on him. :-) He says he's got a stressful day ahead of him tomorrow so he's going to get back to me on Sunday about what we can do about the situation.
He even said:
BTW, respect for still using an Amiga! Fine pieces of kit.
I have a feeling it might work out well. :-D
Meanwhile I have purchased ANOTHER Atlas... this time a newer Maxtor version with 36GB storage. I picked it specifically because it has a good warranty.
I just hope this one works alright... :crazy:
-
Just some quick thoughts
1. Never, never, NEVER terminate anywhere except the end(s) of the chain.
2. Can't remember from post, but is the chain terminated at *both* ends - the CS does not provide termination on itself, you must use a terminator at both ends. (er, yes it is)
3. You *are* using active terminators, aren't you? Ever since Fast-SCSI2, the standard (strictly speaking) requires active terminators. Check that yours *are* active.
4. The 1.5m max SCSI bus includes the device interfaces... a 1.5m cable plus devices exceeds this length, not by much, but it can be critical.
5. A low-level format, if it is offered in the SCSI controller BIOS, can map out bad sectors. Most drives (ie manufactured in the last 10 years) carry "spare" tracks, and remap bad sectors to the spare areas. (I remember someone "discovering" this, and created a utillity which made the spare area available.. 'double your drive capacity for free', I seem to remember. Not a good idea!)
6. Remove the Term Power jumper, and test. Check that the drive is not terminated (it shouldn't be, but check). If this still causes issues, remove the SE jumper. Also, many drivs carry two sets of jumpers for setting ID, term, etc. Check for the existence of a second set (underside and/or opposite end), and ensure term, term power, and ID match the other set.
Drive manual (if you dont have it) can be found here
http://www.maxtor.com/_files/maxtor/en_us/documentation/manuals/atlas_10k_ii_product_manual.pdf
7. Don't use HDToolbox, it's crap.... use anything but!
It could well be the drive, but all troubleshooting should be done from the "wire up". If you have access to a PC with SCSI controller (such as an Adaptec 2940UW or 2940U2/U2W, 29160), check the drive on that. There are also *much* better diag software available under PC OS, which might help narrow it down, such as the excellent Seagate Seatools, and some Maxtor stuff. These can also display SMART error logs, (dis)enable SMART, drive analysis, and so on. If you get really stuck - and are in UK - post it to me and I can check it.
Thinking about HDToolbox, check the maxtrans and DMA mask settings. something like 0x7777777 (? - more knowledgeable peeps can advise on that). Hmm, and something I've never played with, given the vast amount of time since I last played with an Amiga... the partiton size.... presumably you are booting from the IDE drive, not the SCSI? If it is the later, then is there not an issue with partition sizes under the standard Commodore filing systems?
-
Do these two things:
1.Replace the SCSI cable (Parity/geometry errors). Get the shortest one possible.
2.Disable TERM_PWR on the disk.
This will solve your problems.
-
@d0pefish - any joy? Whats the latest?
-
Hi there,
Well, nothing you guys have suggested has had any good effect. :-(
However, I did manage to get the drive connected to my PC. I got hold of an older SCSI card and tried some stuff using a very useful tool called MHDD. I did all the erasing stuff it offered and low level formatting etc - and nothing I can do can bring the 900MB back!
The connection between hard disk and PC is stable and sound. I am certain the problem lies with the disk itself.
I'm waiting for my new hard disk to arrive, probably tomorrow if I'm lucky, and I'm also just waiting on what the seller of the original hard disk wants to do. I'm hoping he will refund me. In essence, I was sold a hard disk that I cannot rely on.
I'll keep you posted with how I get on. I also have a new, shorter and round SCSI cable on its way to try too, mainly as it'd be more suited to the cramped space inside the A4K case, but also just in case this cable is a bit funny.
Cheers for the help! :-D
-
Update
New hard disk and cable arrived. So far working flawlessly; have all my partitions set up and I'm having fun filling it with games and demos. :-) The newer Atlas is much quieter than the original and hasn't had any errors so far (touch wood)! It's reporting the correct size. :-)
The seller of the original drive has agreed to take it back and I'm awaiting confirmation of his postal address so I can return it to him. :-D
Happy ending! :-D
-
I cannot BELIEVE what has just happened.
After using my Amiga for a few hours today, I rebooted to find that - once again, EXACTLY like before - my last partition was corrupt.
And also even more frightening is the fact that exactly like before, the hard drive is reporting a smaller size. Again, I'm missing a couple of gigabytes when reading the geometry from the drive.
I was totally convinced my problems were over as I have had no bother whatsoever with it up until now. So I had loaded up my drive with lots of demos and games, and I had everything *PERFECT*... it took me days to get everything sorted and tidied up how I liked it.
And now it's all gone; a lot of my stuff was on that partition.
I am so frustrated. I haven't even completed the arrangement with the seller of the previous drive yet, and now my new one is dying. It has a warranty, but I am going to have to go through all the hassle all over again.
I really don't know what to do now as I'm losing hope. :madashell: :madashell: :madashell: :madashell: :madashell:
-
I think it's the Atlas drive that is not very CSPPC friendly. I've had similar problems in the past with that same drive, and once I switched to a Seagate Barracuda problems dissapeared. I've had best success with IBM U160 or better SCSI drives and CSPPC. They always work fine with CSPPC and would give speeds close to the theoretical limit (30 to 35MB/s, so pretty close to the 40MB theoretical limit).
EDIT: Fujitsu U160 or better SCSI drives also work great with CSPPC, from my experience.
-
Thanks for the advice on other choices. I don't think I want to even heard the word "Atlas" ever mentioned again. :roll:
I forgot to mention - this time I haven't had a single "Parity error" or similar pop up. Maybe the new cable has helped in that respect...
-
dopefish
Why is all the data "gone" there are recovery tools you can use you know...
magnetic
-
I know there are recovery tools.
But I can only get a tiny amount of what I had back, because according to the drive, the 4GB or so that held most of my data on that partition no longer physically exists. SFSSalv can only access the tiny piece of partition that is still visible to the computer.
Which is extremely infuriating, as I created about a hundred or so directories by hand to organize a lot of demos by group name. The only directories SFSSalv can recover have lost their names.
Have I found a major, major problem with Atlas drives coupled with the CyberStorm, or have I just been unlucky as hell? Either way, I've been shafted by these drives twice, and won't be going with Quantum/Maxtor again... :-(
-
d0pefish wrote:
...
Anyone got any recommendations for other hard drives to use with the CyberStorm? Something equal or better than the Atlas 10K II preferably... 18gb upwards? What are YOU using with your CyberStorm SCSI? :-)
...
I don't know the Atlas 10K II? - my SCSI-chain is as follows:
act. term - CSPPC - 4.3gB UW-SCSI-HD (IBM) - 160gB IDE-HD (Maxtor) via ACARD UW->IDE converter (faaaaast!) - act. term. (upper 8bit) adaptor UW->SCSI II - SCSI-CD-ROM - SCSI-CD-RW - IDE-DVD-RW via ACARD UW->IDE converter - SCSI-tape-streamer - ext. SCSI scammer - act. term (lower 8bit)
All devices have the internal termination disabled.
Regarding your question if you can terminate the chain in the middle of the cable:
No.
IIRC the signals can get "reflected" and the *ENDS* of the cable if those are not terminated actively.
IMO it would make more sense to cut the cable to the length you need and then terminate the ends.
-
Thanks.
While I did wonder about the *idea* of terminating mid-cable, I never actually did it. I was just wondering if that were possible just out of curiosity.
My SCSI chain is fine, the cabling is fine, and I'm terminated properly; at the ends. I have correct cabling and correct terminators - and my chain is nowhere near the complexity of yours. I don't need to cut the cable now that I have a new one.
The problem lies with the Atlas drives I've been using. I am going to try a Fujitsu model next, so the only thing I need now is luck, which I seem to be having a shortage of at the moment. :-(
-
its a really weird problem (close the one i'm experiencing currently here even if with different behaviour.... probably a dying HD here)
The "parity" problem you got (at the beginning) should be related to the device itself and/or cabling/termination (gone now with the new cable)
The fact that HDToolbox reads the geometry wrongly the first time, it could happen.
Have you tryed to search on the net the "real" geometry of your Atlas? and type in HDToolbox manually?
I could suggest you to try with the Phase5 HD SCSIConfig but then you have to use it always.
Another suggestion.... when you install a new HD, save the RDB somewhere. If something will go wrong, you can load the RDB again and see your HD like it was prepped the first time.
Which RomUpdate do you have installed? type in a shell:
version cybppc.device FULL
Edit- ah... just for curiosity, how many "bad blocks" are reported by Q-Device with this new HD?
-
Install MorphOS in other hard drive and run SFSdoctor. It works great.
And don't forget to use good connections and cables.
-
The thing is, all the prep tools I tried were reading the drive as 33.9GB at the beginning. SCSIConfig, HDToolbox, HDinst even Q-Device... all reported 33.9, which is correct according to the formula used to work out "real" capacity versus "manufacturer's" capacity (36.4).
Now all the tools are consistently reporting the drive as 31.9GB. Q-Device even says that the "manufacturer's" capacity is 34.3... so where have those couple gigabytes gone?
The RDB is still set as it was when it read the drive at the right capacity, but if I edit it and re-read it, all the values drop down to make it 31.9. The RDB isn't corrupt or anything, as the computer still boots up from the first and second partitions.
I have MorphOS installed so I will try SFSDoctor to see if I can get any data back.
cybppc.device is phase5 version 44.67 (11/03/1999) (Ralph Schmidt). I'm certain this is the latest official version - the behaviour doesn't change at all with the newer unofficial G-Rex version. (which I was using when the first hard drive screwed up)
The hard drive shows itself at the lower capacity when connected to my PC.
I was reading up on defects on SCSI drives - apparently it's only the GROWN defects and not the PRIMARY ones to be concerned about, as the primary ones are marked by the manufacturer when the drive was made. In that case, both drives have no grown defects which makes things even more odd. The 512 defects on the other drive were primary.
This drive has 451 primary defects, and the number stayed the same before and after the sudden unexplainable shrinkage.
I just don't get it. :-?
EDIT:
Just ran SFSDoctor. THANKYOU!! It is listing all my directories. It doesn't matter so much if I cannot recover the files inside them, but at least if I can get all the named drawers back, that saves me a LOT of pain.
Thanks for the pointer to that tool, it's very nice! :-D
-
x56h34 wrote:
I think it's the Atlas drive that is not very CSPPC friendly. I've had similar problems in the past with that same drive, and once I switched to a Seagate Barracuda problems dissapeared. I've had best success with IBM U160 or better SCSI drives and CSPPC. They always work fine with CSPPC and would give speeds close to the theoretical limit (30 to 35MB/s, so pretty close to the 40MB theoretical limit).
EDIT: Fujitsu U160 or better SCSI drives also work great with CSPPC, from my experience.
This is consistent with my experiences using the cstorm.