Any reason an A3000 wouldn't enable ramsey burst mode with SC memory?

[patrik]

Hi,

I have recently aquired an A3000. It is a 25MHz rev 6.1 machine with kickstart 3.1. It came with 2MB chipram and 1MB fastram as DIPs, so I ordered 8MB ZIP RAM to make it much more usable.

To be more exact, the chips I ordered are Toshiba TC514402AZ-60, which according to the datasheet is static column RAM.

The ZIPs work when mounted, however the performance is the same as with the DIPs. Measured the same fastmem bandwidth with bustest and in SysInfo it scores the same 0.9 times performance of the reference A3000.

If I check the Ramsey settings with RamseyConfig or SetRamsey, they both say burst isn't enabled, which I understand it should be with static column RAM. It does however explain the lack of performance increase. Both of them also crashes the machine if I try to enable burst, albeat in slightly different ways.

What also is rather worrying is that I noticed that it looks like the labels on the ZIPs have been sanded away and reapplied, so perhaps they are just rebranded regular ones.

Having almost no prior A3000 experience, I have to ask if there is anything else that could cause this, or is it simply not static column RAM?

[grelbfarlk]

Originally Posted by patrik

Hi,

What also is rather worrying is that I noticed that it looks like the labels on the ZIPs have been sanded away and reapplied, so perhaps they are just rebranded regular ones.

Attachment 55633

Uh yeah usually Toshiba goes around the world finding static column zips and asks the owner to bring them out to their van in the driveway to sand and repaint them every 20 years, everyone knows this-there is no other explanation.

The good news is that the previous owner has completed the bidecennial maintenance on them so you shouldn't have to worry. Where did you buy them from?

[TjLaZer]

Originally Posted by grelbfarlk

Where did you buy them from?

Perhaps from here?

[ Show youtube player ]

[alexh]

Originally Posted by patrik

I ordered 8MB ZIP RAM TC514402AZ-60 which according to the datasheet is static column RAM

They are 1Mx4 static column ZIP RAM. So 8MiB would be 16 chips which you would fit in BANK0 and BANK1. Leaving 16 ZIP sockets AND the DIP fast sockets empty. Also you will need to set the jumper J852 to position 1-2 which is for 1M parts. (If you had 1MiB DIP Fast previous jumper J852 will be in position 2-3 for 256K parts)

Originally Posted by patrik

If I check the Ramsey settings with RamseyConfig or SetRamsey, they both say burst isn't enabled

Check all the parts in BANK 0. I have read that the A3000 tests only BANK 0 during "POST" too determine if the chips are FP or SC.

[patrik]

Originally Posted by grelbfarlk

Where did you buy them from?

Hong Kong based ebay seller.

Originally Posted by alexh

They are 1Mx4 static column ZIP RAM. So 8MiB would be 16 chips which you would fit in BANK0 and BANK1. Leaving 16 ZIP sockets AND the DIP fast sockets empty. Also you will need to set the jumper J852 to position 2-3 which is for 1M parts. (If you had 1MiB DIP Fast previous jumper J852 will be in position 1-2)

Might this be the other way around? J852 used to be at 2-3 with the DIPs and if kept at 2-3 with the ZIPs, I get 1/4 of the ZIP memory. Moving J852 to 1-2 gives the full amount.

Originally Posted by alexh

Check all the parts in BANK 0. I have read that the A3000 tests only BANK 0 during "POST" too determine if the chips are FP or SC.

To reduce potential issues while testing, I have only populated bank 0. All chips have the same TC514402AZ-60 label, but as said it might just be bogus. Also tried with the other 4MB in bank 0 with the same result.

The setup looks like this:

[alexh]

Originally Posted by patrik

Might this be the other way around?

Fat fingers. I was reading the manual 4-16 and just typed it in the wrong way around.

Looks like you've done everything you can and your chips must be FP rather than SC.

Totally unrelated but I noticed from your photo you've got a Rev 07 buster and a Rev 04 Ramsey.

There is someone on AmiBay selling both Rev 11 Busters and Rev 07 Ramseys at the moment if you want them.

http://www.amibay.com/showthread.php?87134
http://www.amibay.com/showthread.php?87319

Remember to get the correct tool for removing PLCC chips from sockets so you don't break the socket.

You could also look for the AM33C93A-16PC which is the newer SCSI chip.

All three upgrades not necessary unless you want to use A3640 and Zorro III DMA cards (like Deneb or Fastlane)

[SpeedGeek]

Originally Posted by patrik

What also is rather worrying is that I noticed that it looks like the labels on the ZIPs have been sanded away and reapplied, so perhaps they are just rebranded regular ones.
Attachment 55633

Just remove "perhaps" from the above comment and you have already answered your own question.

Originally Posted by grelbfarlk

Uh yeah usually Toshiba goes around the world finding static column zips and asks the owner to bring them out to their van in the driveway to sand and repaint them every 20 years, everyone knows this-there is no other explanation.

The good news is that the previous owner has completed the bidecennial maintenance on them so you shouldn't have to worry. Where did you buy them from?

Toshiba doesn't do this but Chinese scammers do it all the time!

[patrik]

Originally Posted by alexh

Looks like you've done everything you can and your chips must be FP rather than SC.

Thanks for confirming!

Originally Posted by alexh

Totally unrelated but I noticed from your photo you've got a Rev 07 buster and a Rev 04 Ramsey.

There is someone on AmiBay selling both Rev 11 Busters and Rev 07 Ramseys at the moment if you want them.

http://www.amibay.com/showthread.php?87134
http://www.amibay.com/showthread.php?87319

Remember to get the correct tool for removing PLCC chips from sockets so you don't break the socket.

You could also look for the AM33C93A-16PC which is the newer SCSI chip.

All three upgrades not necessary unless you want to use A3640 and Zorro III DMA cards (like Deneb or Fastlane)

I have actually ordered Buster and Ramsey from the exact amibay threads =). Also a AM33C93A is on its way.

The reason for ordering a Buster-11 was because I was curious to see if it makes the X-Surf-100 I put in it behave normally. The card works perfectly in both another ZorroIII Buster-11 machine and a ZorroII machine, but in the A3000 it keeps doing some weird "stalling". For example if you telnet to a host and start typing text, the network traffic pauses about a second every five to ten seconds or so. Really weird. I have tried both the available 030 versions of x-surf-100.device, same behaviour for both.

I only have Super DMAC-02 which I have read isn't supposed to work well with Ramsey-07, but some claim it works for them, so I figured can always test it.

Btw, when reading the Ramsey datasheet, I noticed that there are some general improvements in rev 07 apart from bug fixes. I think the SKIP mode which removes one waitstate for memory accesses is quite well known. But it also mentions that -07 can generate word-aligned DMA-addresses for the Super DMAC, while -04 can only do longword-aligned, so less chance to have to do a PIO transfer with -07. Had never heard about that one.

Not really sure I am having issues with the old WD SCSI chip, but it is always fun to test and tinker . Did get one random checksum error after adding a second drive, but it has never appeared again. And yes the diode situation has not been taken care of, but there is termination power from another device on the bus and termination at both ends.

Originally Posted by SpeedGeek

Just remove "perhaps" from the above comment and you have already answered your own question.

You are right, it does look extremely suspicious, but I wanted to get some more opinions as I have zero experience with SC ZIPs in A3000.

[Jope]

Please report back on your mismatched Super DMAC <-> Ramsey experiment once you've had a chance to try it.

[Marmotta]

Same Ram and same problem.... "china ram" is always FP????

[ChristianS]

I got the same chips from the seller on eBay. They are fake and not Static Column. They mostly work (night have been desoldered too hot when they were likely pulled of some board), but one was completely dead and I got random lockups with the remaining. I’ll use them with something different than an A3000. Might work with the A2386SX Bridgeboard, the Oktagon, or the Harms Professional A2000 accelerator.

As stated in a prior post - the chips surfaces have been sanded and relabeled if you look closely.