Wanted: 68040!

[zipper]

alexh wrote:

Kin-Hell wrote:
...but a CSPPC/CSMKIII card has two Capacitor settings within the socket.

They are not capacitor settings are they? I do not know what they are for sure. They perhaps somehow alter the output of the voltage regulator? I honestly do not know 100% how power electronics work (in general) let alone how the CSMKIII works.

A plain jumper, under the 68k chip http://members.iinet.net.au/~davem2/overclock/csppc.html

Kin-Hell wrote:
An 060 using 60ns ram on say a CSPPC or CSMKIII card with SCSI III drive is not as snappy on the desktop as a Warp Engine using 60ns Ram on a SCSI II drive.

How very strange. I suggest that some of the OS/desktop software you are using might use a feature of the 040 that is perhaps emulated in the 060? That would explain it. Or perhaps a bug in said software?

Interesting non the less.[/quote]
What about the Chip RAM speed of said cards?

[adolescent]

delshay wrote:
MC68040V should work on most 040 board you just have to add a regulator ( if needed ).

No.  The 68040V pin-out is different too.

[alexh]

zipper wrote:
What about the Chip RAM speed of said cards?

Can the speed at which accelerators access chip RAM be different? I would not have thought so.

If you were right, and KinHell is describing a card which doesn't have a gfx card, just AGA then that might explain it.

[Argus]

As Dave Haynie could probably explain better, the chip ram and fast ram on the 32-bit Amigas was designed to interface with the 68030 chip cycles (which was the designated cpu at the time Dave was designing the A3000).  Adding a 68040 or a 68060 in asynchronous mode via the fast slot changes this relationship.  The A3640 for instance was notoriously slower at accessing the motherboard fastram on the a4000 (and A3000) due to this difference in cpu cycles and lack of burst mode capability.  Phase5 iirc added special hardware onto all of their 040/060 boards to replicate in hardware the interface of the accelerator card with a 68030, at least for chip ram access.  Thus, if you run an AIBB or similar check you'll find the Phase5 boards outperform other cards with respect to chip ram access speeds, which is key when talking about how fast the AGA chips are going to go.  Other manufacturers (GVP, Macrosystem, etc.) didn't fully implement this or ignored it altogether, opting instead for the fastest accelerator ram-to-cpu speeds.  Thus, they generally outperform the Phase5 boards on fastram (that is RAM physically on the accelerator) speeds, which helps more if you are talking about graphics cards or zorro card speeds rather than the AGA chips.  Also, I believe GVP is the only company that really worked on cpu burst cycles to the motherboard. There is a jumper on the GVP4060 board which purports to allow burst cycles to the motherboard, but I'm not sure if this actually works.  Dave or Greg Berlin (who designed the A3640) could explain this better, because the A3640 was compromised as a quick design to get a cheap 040 board ready for the A3000 and left out burst, 68030 async logic, etc.

[Kin-Hell]

All very interesting stuff.  :-)

[alexh]

@Argus

Interesting, but benchmarks on Amiga Hardware Database show that the CSMKIII (& CSPPC) has better fast RAM performance than any other card.

The WarpEngine was not included in the test but it would have to be something special!

[delshay]

adolescent wrote:

delshay wrote:
MC68040V should work on most 040 board you just have to add a regulator ( if needed ).

No.  The 68040V pin-out is different too.

MC68040V was fitted to a Blizzard PPC this is a test processor only & will remain so but will play no part in my final projects.

it was used only when the Blizzard PPC was converted from 040 to 060 where socket was fitted & voltage changed was done & a MC68040V was fitted to check everything was ok.
 
it has a different pin count but as far as im am aware i did not find a problems & ran ok at 3.3v settings on the Blizzard PPC board.


a full test will now be done to see if its compatible with the Blizzard PPC but will check docs first.


will post here with results in a few days.

[Kin-Hell]

This just got even more interesting!  :-D

[delshay]

@adolescent

being look at the docs on the MC68040V pinout.

it seems you are right i bought the card off EBAY with the above processor did not pay attension to v marking on the processor at first,just plug it into my A1200 when i recive it.

even when i changed the processor for another 040 processor the card still did not work and it suppose to be working.

ok im not going to have a go at the person who sold me the card because i found the fault & the card is working fine.

i wiil be going over docs in fine detail,did not think to go over the docs.


thanks Del

[Damion]

As far as the popular A1200 040/060 cards, the Apollo had faster chipset access (approaching Blizzard 1230 performance under certain circumstances), but not by a large margin. Fastram performance may also be slightly faster than Phase5/DCE cards at the same memory clock. However, the Blizzard can generally handle a higher memory clock (1/1 cpu/memory at 66MHz+), potentially giving much better fastram performance, whereas the Apollo can barely handle a 50MHz memclock, even with excellent RAM. In my experience, anything above 50MHz and the memory speed must be halved.

By looking at the SysSpeed benchmarks for the big-box cards, it would seem Argus is 100% correct in regard to the Cyberstorms and chipset access (and the MKIII appears to have significantly improved fastram performance). BusTest shows marginally less favorable results, but I have no idea how accurate these benchmarks are at actually determining anything substantial. :shrug: Fun but meaningless, LOL.
 
I've read that the QuikPak/GVP 4060 take advantage of EDO RAM, and pwn the rest in terms of fastram performance.

[Argus]

alexh wrote:
@Argus

Interesting, but benchmarks on Amiga Hardware Database show that the CSMKIII (& CSPPC) has better fast RAM performance than any other card.

The WarpEngine was not included in the test but it would have to be something special!

That's probably true, I was thinking of the comparison between the original Cyberstorms (MkI&II) with the GVP and Macrosystem boards.  Phase5 introduced 64-bit interleaved Fastram with the CyberstormMkIII/PPC boards (that's why you need to install matched pairs of SIMMs on these two boards).  The 64-bit memory design has to do with the PPC604 chip used in the CyberstormPPC iirc and that most likely explains the speed difference over the earlier generation boards in pure Fastram access.  But for any of those who own the Phase5/DCE 2060, Cyberstorm MarkI/II, Blizzard 1260 or BlizzardPPC, you have a 32-bit memory interface with your 68060 and pure Fastram speeds are more than likely eclipsed by the GVP and Macrosystem designs...not that you'd really notice as the difference is nothing like the problem with the A3640 and m/b RAM.  But AGA chipram speeds will always be faster with a Phase5 board, if that's important to you.  I honestly don't know much about the Quickpak or Apollo boards; I've heard they have their own unique qualities, EDO RAM, etc. (which in the case of the Apollos can make them finicky about the SIMMs they'll work with).

One last thing, the GVP4060 definitely can't use EDO SIMMs, you need FPM modules...save yourself the heartache... :-)

[delshay]

this device allows MC68040V to be plug into a 040 socket along with a MC68060.

http://www.emulation.com/catalog/off-the-shelf_solutions/production-test_adapters/upgrade_motorola/

just to add the MC68040V did boot fine on a Blizzard PPC with no problems im still looking at the docs.

i will try and findout if the above device needs to be config between 040V & 060.

[stefcep2]

I benchmarked my Apollo 1240 40 MHZ with 32 mb edo ram using syspeed and AIBB years ago and got fast ram benchmarks that were faster than Cyberstorm 3 68060.  No-one at the user group believed me until we ran the benchmarks on my mchine and on an A4000 CS 3 68060.  My A1200 scored higher in the fast ram tests. I later read that Apollos had the fastest ram bus of any accelerator, but I can't remember where I read this.

[alexh]

I wonder what is going on?

The SYSPEED fast RAM scores of A1240 on Amiga hardware database are (at best) 1/2 that of CSMKIII

6.95/26.24 vs 26.97/46.56

[Damion]

@Argus

DOH! It's just the QuikPak that utiizes EDO, then? Seems like a nice card. Do you know if any of them actually had SCSI onboard?

@AlexH

Well, one thing to keep in mind is the Apollo 1240 has a slower memory clock, and I'm pretty sure the MKIII should be performing better anyway. Compare it to the MKII 040/40 though, and the scores are much closer. Also, note the difference between the two Apollo 1240 cards, obviously hw and sw setup makes a difference.

You can take a look at this thread for some bustest results from my Apollo card, an MKIII and a few others (again keeping in mind that my memclock is set to 1/2 CPU speed). There's an overclocked Apollo 4060 card in there (100MHz CPU, 50MHz memory) that returned some impressive numbers.

And as far as chipset performance with the A1200 cards, show me a Blizzard that can hit 5 mb/sec in the SysInfo drive benchmark, with an IDE-Fix Express and a memory clock of only 40 MHz. Won't happen. :-P Actually, it would be pretty interesting if we could get someone to run bustest on their Blizzard 1260 card and post it to the thread. Overclocked, the Blizzard should maintain an edge in fastram performance, while the Apollo should return better "chip" numbers.

[vk3heg]

@ Thread:

Can I have my thread back please!

I've received a couple of offers, just waiting to sort things out now.