scsi "killer" ?

[whaka]

Hi,

i have a 3000, and i don't specially use the internal scsi, i just plug a cd-rom drive occasionnaly.
so, like for the ide port in the 4000, when no devices are connected, boot time is more longer.

do you think it's possible to realize a simple scsi killer, like the ide killer for the 4000 ?

i have the pinout, and a connector for experiment, but i don't have knowledge on scsi protocol... so, i need a bit help...

thanks

[Pat the Cat]

Put a passive terminator on the end of the SCSI chain (last device on the connection) should shorten bootup times, if you're not doing that already.

Typically it's an external plug that goes into one of the two SCSI connectors on an external drive (which sounds like what you are doing).

If it's an internal CD-ROM, put the passive terminator on the last connector on the chain, use the 2nd last connection for the SCSI CD-ROM drive.

The SCSI interface on an A3000 also has a terminator fitted (3 banks of resistor packs), sometimes these are damaged or have been removed. BOTH ends need the passive terminators, which make the response time of the interface a lot "snappier", and prevent cross talk that causes spurious spikes on the data connections of a long SCSI cable. There is a maximum length of cable, usually 3 meters for the whole length (internal and external cables combined, although sticking to 1.5 meters total length is a good plan).

There are lots of other things that can issues on boot times of an A3000, but I'll let others chip in, as it has been many years since I used one.

[Amiga1992]

He asked about when there's NO SCSI device in the chain.

[Daedalus]

Hmmm, some irrelevant and only vaguely useful/correct information there. The delay he's wondering about is nothing to do with termination. It's the delay resulting from the system querying each device unit ID in turn to determine which IDs are present.

When devices are present on the bus it can be possible to set the last unit flag on the device. This stops the system searching for any IDs beyond that one, thus speeding up the boot process. But with no devices attached this isn't possible, and the entire bus is scanned.

The A3000 has a setting in its NVRAM for slowing down the detection process. This is to allow slow drives to fully spin up before being queried, but adds the time to the boot process. It can be disabled using tools from Aminet, and that should help speed it up.

In theory it would be possible to create a device that responds to queries with a last unit flag, but such a device wouldn't be a passive item like the IDE or floppy killers - it is more likely to require some sort of intelligent controller to deal with the queries correctly. A far simpler idea however would be to get hold of an old SCSI hard drive, or if power/heat/noise/space are a priority, a SCSI2SD adaptor, set that to unit 0 and set the last unit flag on it. That will give the minimum boot time possible without going to drastic lengths, however will need reconfiguring to a higher ID when connecting the CD-ROM drive.

[mark_k]

Check SetBatt on Aminet. If your A3000 is currently using the long SCSI selection timeout (2s) you can use that to change it to 128ms.

[idrougge]

But Daedalus, the CD-ROM will never have the last unit flag set either, so that can't be the answer in this case. I would not exclude the possibility that it's a termination issue.

[Daedalus]

Yeah, the CD-ROM drive won't have the flag but it will normally respond almost immediately, which will cut down the query time by eliminating that unit's delay. I can't see how termination can be the issue. If there were other devices attached, maybe. But with no devices attached it won't receive any sort of valid response, terminated or not.

Edit: Ah, I wasn't suggesting that setting last device flag on the CD-ROM would solve the issue, sorry if it sounded like that. That's why I suggested an old HD or SCSI2SD attached to the bus.

[idrougge]

But with seven units, the CD-ROM should only cut one seventh of the bus scanning time.

[Daedalus]

I see what you're saying. So what sort of termination issue would cause long delays, or rather, would cause the delays to go away when a CD drive is connected? Especially since we should assume that the CD drive is working fine when attached.

[idrougge]

The difference as far as I can tell, at least going by your explanation that scanning the bus is what takes time, is that the CD-ROM terminates and the bus is otherwise left unterminated. So either the OP does always terminate his bus and the scan process doesn't work quite as you described it (exiting as soon as it finds one unit on the chain) or the OP doesn't plug in an external terminator when the CD isn't plugged in, and this causes the controller to "try and try again" before giving up on getting anything worthwhile out of the bus.

[fgh]

Only option to completely get rid of any scsi delay is to burn a custom rom without scsi.device. Pulling the scsi chip does not help

Otherwise: Two nvram settings are relevant - Slow scan and lun scanning. (worst case = it might be scanning all luns for each id)

Idrougge: You're misunderstsnding, read up on the rdb 'last unit/drive' flag..

[Daedalus]

No, I think he gets it alright, though a controller won't stop scanning once it finds a device *unless* the last device flag is set (and that flag can't be set on CD-ROMs or other removable media drives), since then it would never see more than one device. It's a good point though: I don't think the A3000 was ever really intended to run without any devices attached, so it might rescan the bus a couple of times if it doesn't find anything. I'm not that familiar with its intricacies so I can't say for sure, but that would explain the much longer wait time.

[idrougge]

Mind you, I don't remember having any overtly long startup times on my A3000 when I ran it on a Buddha without SCSI units.

[Pat the Cat]

Originally Posted by Akira

He asked about when there's NO SCSI device in the chain.

Interesting. The A3000 was released with Kickstart on Hard disk - so without a SCSI device in the chain, how could such a machine start up?

Sure, there were kits later which had real Kickstarts on ROM chips, but that wasn't how the machine shipped, and it is very stupid of you to pretend otherwise.

Of course you will ignore my question, being a trashtalking no nothing zit in need of squeezing.

And yes, if there's no devices in the chain, terminators do indeed boost the time it takes to scan all possible addresses on the bus (and discover no devices).

Try it. Not Akira obviously, last time he had an original thought, Abraham Lincoln was still in the White House.

Originally Posted by Daedalus

No, I think he gets it alright, though a controller won't stop scanning once it finds a device *unless* the last device flag is set (and that flag can't be set on CD-ROMs or other removable media drives), since then it would never see more than one device. It's a good point though: I don't think the A3000 was ever really intended to run without any devices attached, so it might rescan the bus a couple of times if it doesn't find anything. I'm not that familiar with its intricacies so I can't say for sure, but that would explain the much longer wait time.

Correct. The machine was designed to load Kickstart from Hard disk - an early version of soft firmware.

[Amiga1992]

Originally Posted by Pat the Cat

Interesting. The A3000 was released with Kickstart on Hard disk - so without a SCSI device in the chain, how could such a machine start up?

Can you not boot an A3000 with a Kickstart disk?

it is very stupid of you to pretend otherwise.

[...]
Of course you will ignore my question, being a trashtalking no nothing zit in need of squeezing.

[...]
Try it. Not Akira obviously, last time he had an original thought, Abraham Lincoln was still in the White House.

Thanks for the unwarranted insults. I did not insult you at all, I just clarified the OP's request because your text hinted at you missing the point.

[DamienD]

What is it with you guys???

Pat, you've only been on EAB for six or so months and been banned twice already.

One more post like the above towards anyone and it will be a permanent on; absolute last warning!!!

[Jope]

Originally Posted by Pat the Cat

Interesting. The A3000 was released with Kickstart on Hard disk - so without a SCSI device in the chain, how could such a machine start up?

You feed it a superkick disk when it asks for one. The later A3000s were shipped with 2.04 ROMs though.

And yes, if there's no devices in the chain, terminators do indeed boost the time it takes to scan all possible addresses on the bus (and discover no devices).

No, the terminator is there just to ensure signal integrity. The host adapter will get just as few answers from the non-existent devices as without a terminator and the timeout value is down to the driver.

The DF0 and IDE terminators that Amigakit sell are incorrectly named in this sense, they are not bus terminators, merely hacks to fool the drivers into thinking there is something connected.

Correct. The machine was designed to load Kickstart from Hard disk - an early version of soft firmware.

Well as it happens, I never installed newer ROMs to my 3k, because I love the idea of superkicking. And as it happens, I can reach the power button without moving from my seat.

clicky

1.3 can be kicked off the HD too, but I'll leave it as an exercise to the reader to figure out why it is ghosted on my system.

You remember many things from your time with the Amiga, but it has become rather obvious, that some of your memories are beginning to fade. Please relax a bit on this forum, and accept that maybe sometimes those memories of yours need refreshing.

[Pat the Cat]

Originally Posted by Akira

Can you not boot an A3000 with a Kickstart disk?

Nope. You need a Superkickstart disk.

A1000 = no hard disk, no MMU. 256KB maximum Kickstart size.

A3000 = hard disk, MMU always present. No maximum kickstart size ... well, actually you need to hack them like crazy to get anything over 512KB.

Originally Posted by Akira

Thanks for the unwarranted insults. I did not insult you at all, I just clarified the OP's request because your text hinted at you missing the point.

Bearing in mind the number of times you have trash talked me and a lot of other people, I thought it best to get in my retaliation first.

No, really, no need to thank me.

Originally Posted by Jope

No, the terminator is there just to ensure signal integrity. The host adapter will get just as few answers from the non-existent devices as without a terminator and the timeout value is down to the driver.

The DF0 and IDE terminators that Amigakit sell are incorrectly named in this sense, they are not bus terminators, merely hacks to fool the drivers into thinking there is something connected.

Correct. And Kudos for keeping your A3000 "original".

The terminators just give nice clean signals. Without them the controller has a fair bit of static to sort out as non-signals and no devices present.

But, to be fair to Amigakit, the "floppy terminators" do actually decrease startup times, so they are used for a similar purpose.

Originally Posted by whaka

Hi,

i have a 3000, and i don't specially use the internal scsi...

If you try removing the SCSI cable from the hard drive, and turn the machine on.... if you get an "insert floppy screen" then you have a later / post production run version with onboard real Kickstart ROM chips.

If not, you've got an original with the Kickstart loading from hard disk.

Turn off and plug the cable back in either way. But I think you got an original.

[Amiga1992]

Originally Posted by Pat the Cat

Nope. You need a Superkickstart disk.

Which is a disk. With a Kickstart in it. Like Jope mentioned before and proved with a picture. You don't need the SCSI HDD to be present like you insist.

Originally Posted by Pat the Cat

Bearing in mind the number of times you have trash talked me and a lot of other people, I thought it best to get in my retaliation first.

No, really, no need to thank me.

Seems like you purposefully ignored DamienD's comment. Thanks for yet another unwarranted attack.
It seems to me like your definition of "trash talking" is proving you said something wrong. But I wasn't even doing that, I was clarifying and I even asked you a question, which you again answered with snarky remarks.

[Pat the Cat]

No, a Kickstart disk isn't a Superkickstart disk.

A superKickstart disk will not work in an A1000 either, and the vast majority of Amigas are usable BECAUSE YOU DON'T HAVE TO WAIT FOR A FLOPPY TO LOAD THE FIRMWARE. Sheesh, that's an extra 40 seconds of boot time... the poster asked because they wanted to reduce boot times, not increase them as you advise.

Originally Posted by Akira

Seems like you purposefully ignored DamienD's comment. Thanks for yet another unwarranted attack.
It seems to me like your definition of "trash talking" is proving you said something wrong. But I wasn't even doing that, I was clarifying and I even asked you a question, which you again answered with snarky remarks.

I am sure DamienD is perfectly capable of making his own mind up without you telling him what to do.