A3640 V3.0 refuses to run speed test on SysInfo V3.24 ?!

[Effy]

If I knew it was a V3.2 then I still would have sold it for the same price  :lol: Sorry for all the confusion but I really didn´t want to cause that much commotion  :-) Happy to hear you have now got a full 040  :lol: But be careful with overclocking faster than a 66 Mhz crystal as I have read the problem may not be the processor but the A3640 itself ...

[Kin-Hell]

by alexh on 2006/3/13 21:22:32

Quote:
A3640's V3.0 with the cut 'n' jump are almost good enough for most systems.
Unqote:

Not if you want to run SCSI without DMA issues.
I'd also like to add, 3.2 cards are not as common as 3.0 and 3.1's. Id also like to add that after cleaning the card up, Pal Chip U209 (ending -03 denoting the card to be a v3.2 despite the v3.1 on the PCB edge) is a re-worked chip.
I spect that was a "Wang" job back in 1992 when they had the repair & upgrade contract for Commodore. When I aquired my replacement 040 card from Wang, the Wang Engineer told me of DMA issues on SCSI access (not bs IDE-SCSI) when using a 3.0 card in an A3000/A4000, the 3.1 fixed the DMA SCSI issues for the A4000 but the 3.2 was needed to fix DMA issues with the A3000.
I was never aware of the incompatabilities you mentioned, so all these years on, it still goes to show you never stop learning!

Anyways, it kinda makes the v3.2 worth that little more than the 3.1 & if you are serious about your Amiga, you wont have ANYTHING else but SCSI. The IDE interface on the release of the A4000 series was a joke to any Serious Amigan. I remember coming real close to wang-ing my 9" angle grinder on the IDE header the day I fitted my 8Mb Octagon SCSI card. It was still under warranty so I shyed in face of.:mickeymouse:

No problems this end Effy. Thanks for the original card, it looks like Brand New since I cleaned it up & its now capable of running the Destop Toaster.
Hope you're well, have missed seeing your Amiga Photos on eBay of late. We need more Cleavage in the Amiga Section. ;-)
Thanks!  :-D

*Edit* yeah Effy, I know about the cards issues beyond 66Mhz. This XC chip sould be real sweet & nice'n'cool at 60.
Gotta get a socket first. :-D

[alexh]

Not if you want to run SCSI without DMA issues.

It's not SCSI related at all. It's local bus mastering related. I've got several SCSI cards that work fine on 3.0 cards with the cut n jump.

All 3.2 does over 3.1 is it saves upto 10 microseconds per DMA cycle.

[Kin-Hell]

alexh wrote:

Not if you want to run SCSI without DMA issues.

It's not SCSI related at all. It's local bus mastering related. I've got several SCSI cards that work fine on 3.0 cards with the cut n jump.

All 3.2 does over 3.1 is it saves upto 10 microseconds per DMA cycle.

With all respect m8, wtf do you think DMA is doing whilst trying to Master the Bus? The SCSI protocol relies on DMA calls being correct whilst acessing its bus in line with all other Amiga calls. Your Syncronous & Asyncronous SCSI calls will affect the performances across the SCSI bus if it can't utilise DMA correctly.
Modern PC boards Still have an Option for Bus MAstering in the Bios. This Call affects & is affected by DMA (Direct Memory Access) Watch your Hard drive speed at DOS levels slow up for Bus Mastering NOT being enabled.
Cut & jump is obviously a "work-around". Do you have a link with reference to the cut & jump procedure?
Thanks.  :-)

[alexh]

The SCSI protocol relies on DMA calls being correct

:rtfm: The SCSI protocol doesnt require DMA. Some SCSI cards use DMA local bus mastering, some dont. You can even switch bus mastering off on cards that dont work with a certain revision A3640 to make them work!

One of the problems with the 3.0 cards is a DMA local bus mastering problem it is NOT (as previously stated) a SCSI problem.

If your SCSI requires DMA (Like the A3000 internal SCSI) then you are going to require a -02 U209 PAL or higher or switch off DMA. If your card doesnt use DMA, you'll be alright. Simple as that.

As you say a card designed to use DMA will probably be painfully slow without it (depending on CPU power) but hey.

Cut & jump is obviously a "work-around". Do you have a link with reference to the cut & jump procedure?

The Cut n jump fixes a problem with sampling the bus signals (STERM) so your Cybervision will work. The 3.1 -02 PAL fixes the bus mastering issue. The 3.2 -03 PAL fixes the the bus mastering issue in a different way without delays of upto 10 us per DMA access (Fixes GVP PhonePak and increases throughput).

http://wonkity.com/~wblock/a4000hard/a3640ref.html

There is a better page somewhere with pictures, I'll try and find it.

This is another good A3640 site. This repair of the capcitors can sometimes resurrect seemingly dead A3640's

http://joj.home.texas.net/amiga.html

[Kin-Hell]

alexh;

It's really a matter of choice. Either you use Hardware that supports DMA or you don't.
Having used all sorts of Computer Platforms since the early 80's, if the Bus Device didn't utilise DMA, I wouldn't touch it with a barge pole.
Sure, some Bus devices dont actually require DMA to function, but believe me, DMA makes a difference.
DMA was an Industry standard, setup for all types of buses to use & SCSI controllers with their own onboard RAM usually performed better than those without.

With all respect, the cut & jump is a work around but far from being correct & the performance will suffer accordingly.
I'm sure DCE used DMA protocols on their PPC & Motorola accelerators with SCSI because they wanted Maximum performance possible.
Thanks for the info on A3640 cards, much appreciated, but you would be better off embracing as much DMA as you can.
Get and upgrade when you can.

I'd also like to add, some post'ees in this thread just Blamed SysInfo for being Buggy. Not the case....their Amigas just arn't set up correctly, or, they are using bs LC chips which obviously dont utilise the Library code correctly as the Full 040 does. Interesingly enough, EC CPU's dont crash out on SysInfo either so that would conclude the arcitecture of the LC chips not conforming to the Library drivers correctly. Don't also forget, this was tested on the SAME install of the OS.
To conclude this thread Effy started on my behalf, the card Effy sold me was fine. It was the cr^p LC chip that borked SysInfo up. It also amazes me why anyone serious about their Amiga would not want an FPU in the first place. :-?

[motrucker]

Why in the world is everyone beating their heads against the wall trying to get a proven crap program to work?!
Everyone in this neck of the woods calls it MisInfo, to make the point.....
SysInfo _may_ have worked OK back in the early days, with 68K up to 68030 CPUs, but some '030 systems would throw it for a loop.
As some one else said - use SysSpeed. It does work!, and works well with '040 & '060 systems.

[Kin-Hell]

motrucker wrote:
Why in the world is everyone beating their heads against the wall trying to get a proven crap program to work?!
Everyone in this neck of the woods calls it MisInfo, to make the point.....
SysInfo _may_ have worked OK back in the early days, with 68K up to 68030 CPUs, but some '030 systems would throw it for a loop.
As some one else said - use SysSpeed. It does work!, and works well with '040 & '060 systems.

SysInfo also works u muppet; Works fine on ALL Amigas I have ever owned & setup, & get Real, its only an "Info" program...not a friggin benchmark.
SysInfo doesn't Bomb out on my A4000T with PPC 604 @ 233 + 060 & 128Mb Ram with SCSI III.

The b*tch about the whole Scenario turned out to be a Bull Sh*t issue with a Cr^p LC version of an 040 CPU. I thought I was buying a Full 040 on the A3640. Infact, the Dork that ever Pulled the Original 040 CPU left Butchered marks on the Socket, but then, I might be imagining that as well eh!

FFS!....>Sometimes there is just "NO" telling anyone FACT is there! Almost as bad as alexh going on about DMA. :roll:

You will have to excuse the flaming I throw at you, but for crying out loud, at least try & show some respect for an Old Amiga User since the A500 was first released. I have owned & used every Amiga Since. My Beginner Status here on the Forums does not mean I am a Noob. I don't know everything there is to know about Amigas, no one ever could, but I dont talk about stuff I dont know anything about. Moreover, I never ever saw cr^p like this ever, till I ended up with an LC (low-cost) CPU.
Like I said earlier.....Key-Ring anyone!?!?!?!?

[alexh]

FFS!....>Sometimes there is just "NO" telling anyone FACT is there! Almost as bad as alexh going on about DMA.

Oops, I just wanted to say to all those out there reading that it was a local DMA issue, not necessarily a SCSI issue.

Anyhow everything is sorted. Case closed

[motrucker]

Kin-Hell wrote:


SysInfo also works u muppet; Works fine on ALL Amigas I have ever owned & setup, & get Real, its only an "Info" program...not a friggin benchmark.
SysInfo doesn't Bomb out on my A4000T with PPC 604 @ 233 + 060 & 128Mb Ram with SCSI III.

The b*tch about the whole Scenario turned out to be a Bull Sh*t issue with a Cr^p LC version of an 040 CPU. I thought I was buying a Full 040 on the A3640. Infact, the Dork that ever Pulled the Original 040 CPU left Butchered marks on the Socket, but then, I might be imagining that as well eh!

FFS!....>Sometimes there is just "NO" telling anyone FACT is there! Almost as bad as alexh going on about DMA. :roll:

 

My god, another .............Sorry about this "post". This is what I get for leaving my computer logged in here when three drunken "friends" come over.

[Piru]

@Kin-Hell

I'd also like to add, some post'ees in this thread just Blamed SysInfo for being Buggy. Not the case...

Code: [Select]

; hunk 0, offset $33DE

        lea     $7FFFC,a1
        lea     $2000.w,a0
.iloop  move.l  (a0)+,d0
        cmp.l   a1,a0
        blt.b   .iloop

From SysInfo 3.24


The code is assuming that address $2000 is accessable. In certain cases the mapped memory only begins at $4000, and SysInfo will nuke (the test is executed in a way that any exception will kill the system).

Correct chipmem start address would easily be available from the chip memory MemHeader.

[motrucker]

Piru wrote:

I'd also like to add, some post'ees in this thread just Blamed SysInfo for being Buggy. Not the case...

Code: [Select]

; hunk 0, offset $33DE

        lea     $7FFFC,a1
        lea     $2000.w,a0
.iloop  move.l  (a0)+,d0
        cmp.l   a1,a0
        blt.b   .iloop

From SysInfo 3.24


The code is assuming that address $2000 is accessable. In certain cases the mapped memory only begins at $4000, and SysInfo will nuke (the test is executed in a way that any exception will kill the system).

Correct chipmem start address would easily be available from the chip memory MemHeader.

Again, if YOU ................Again, sorry for the outburst (see last "post")

[Piru]

Did my quoting confuse someone?

[Kin-Hell]

Piru wrote:
Did my quoting confuse someone?

No Piru, this thread is Just now getting out of hand. Everybody loves to have their 2 pennyworth & I admire fellas like you who can Pull Code apart to proove such events.
I must be honest, the 128Mb Ram on my PPC does not Show in the fast Ram Section,...take a 32Mb Stick out & it all shows, so sure, there are bugs. Even the SCSI LED stays on after a SCSI test but NEVER have I ever seen Sysinfo Crash (reset) to a GURU MEDITATION for doing a speed test.

My Point is that from my experience of Amigas & what I know, SysInfo as JUST an Info Tool is MORE than capable of standing up to 68000/020/030/040/060/PPC systems. My concern & why Effy started this thread was as to Why it would crash. Irrespective of SysInfo being full of bugs, it should not fall over on an 040 CPU. I was NOT aware it was an LC version of the 040 until I removed the Heatsink. Nor was I entirely sure that fitting a FULL 040 would stop it happening, as this was a Version 3.2 card despite the Silk Screening v3.1 & I have had no previous experience with a v3.2.

Forums are a hive of Information, but they are also one of the worst environments for Bad information when people talk about stuff they know sweet FA about.

Indeed, hats off to alexh for slating the Cretin in here earlier who was misleading innnocent people about the different CPU's. I read the thread after he had posted & was poised to state the same he did.

We appear to have gone thro DMA, Sysinfo...christ...even Code pulled apart & fair play to all remarks.

However, in my book as a Hardware Technician, DMA not being utilised properly means its a Fudge, resulting in lack of performance & thats the end of the matter for me.
DMA is a standard accross every computer platform known to man & please no comments about games consoles.

I have a result about my main concern after recieving the card from Effy in Belgium. I extend my thanks to him for starting the thread & being cool with me about resolving the issue.

I am now un-subscribing from this thread, my mailbox is a Junk shop for it.

Thanks to everyone who has had some valid input & those of you who know nothing about what you are talking about should go & study real hard! Forums lack facial expressions & words can appear harsh when read without the facial expressions that would normaly accompany conversation. Then it ends up a flaming match which I will not even waste my time retorting to. So far as respect goes to one "over-the-pond" postee, flaming in such manner scores you Zilch!
Maybe its just they want to get their post score upped!  :roll:

I expect some prat to retort to that as well!
Thats forums for ya!  :lol:

Extra special Thanks to Effy, alexh & Piru.
Oh yeah, & alexh, I use to have that old Sysinfo back in the 80's. Unfortunately, that went with all my Amiga gear back in 1996 when I sold out. If I ever come accross it, I'll post you a copy with pleasure.  :-)

[motrucker]

Allow me to apologize to you, Effy, Piru, and everyone else in this thread.
I didn't write the last two "posts" under my name, but they are still my fault, because I walked away from this computer while it was still logged into this thread.
The guys who wrote that are usually very nice people, when not in a drunken stupor.
My sincere apologies.
                          Eric
PS rest assured I will never again walk away from a computer that is still on line.

[Kin-Hell]

motrucker wrote:
Allow me to apologize to you, Effy, Piru, and everyone else in this thread.
I didn't write the last two "posts" under my name, but they are still my fault, because I walked away from this computer while it was still logged into this thread.
The guys who wrote that are usually very nice people, when not in a drunken stupor.
My sincere apologies.
                          Eric
PS rest assured I will never again walk away from a computer that is still on line.

Eric;
No problems this end, but at the same time, OMG eh!  :-o
All the best.
Charlie