Individual Computers - New accelerator cards for A600 & A1200

[Schoenfeld]

Originally Posted by melbourneben!

Jens, I run WB 3.0, will the new version of ACAtune(V15e) resolve the issues with assigning the memory in WB 3.0?

That issue was already fixed in the previous version. We haven't brought it back in this one :-)

Jens

[Seaside]

Some questions.

Is it safe after all to use 80MHz OSC on the ACA630 ?

Can i remove the current CPU and replace it with Ceramic PGA ?

Also can the memory raised to 64 or 128 using same type chips ? Or this has to do with the initial programming ?

When the ACA will become available again ?

Thanks.

[Schoenfeld]

Originally Posted by Seaside

Is it safe after all to use 80MHz OSC on the ACA630 ?

No, it isn't. See further back in this thread for detailed technical explanation.

Originally Posted by Seaside

Can i remove the current CPU and replace it with Ceramic PGA ?

Yes, you can. Doesn't make sense, but you can.

Originally Posted by Seaside

Also can the memory raised to 64 or 128 using same type chips ? Or this has to do with the initial programming ?

Memory size is hard-coded in the CPLD. Neither the CPLD, nor memory are user-upgradable.

Originally Posted by Seaside

When the ACA will become available again ?

Not in the form that the ACA630 came in, as that was too expensive. I'm thinking of a lower-cost, lower-performance version for the A600.

Jens

[Photon]

Suggestion for a future accel:

- A1200 card with no CPU or memory
- Just a 68060 socket and 1-2 sockets supporting up to 512MB of the abundant PC DDRx memory

This will ensure the prolonged life of high end Amigas while you get rid of the expenditure for CPU and memory.

And you could give a modest rebate if buyers send in working or non-working Apollo cards, then you could get rid of those.

[Mr B]

And the card would of course have a switch / jumper cluster, as well as a flexible clock frequency generator so it could run on both 68040 & 060 CPU's.

However, i see issues here. A buyer with a flaky CPU will inevitably blame the card, the CPU was bought as "working pull" from some asian guy... Makes it a nightmare to support. And the sort of optimization Jens is doing to push the performance from the CPU's would be impossible with a wide variety of CPU & memory combinations.
B!

[vroom6sri]

Jens - So, just coz I can't stand suprises, when you stated earlier that they " heat transfer pads will be fixed for shipping with an Amiga-related item..." does that mean we'll be getting a bonus item of some sort??!?!

Don't get me wrong, the fact that you're making such great hardware is good enough but I was just intrigued by your statement.

[Photon]

Originally Posted by Mr B

And the card would of course have a switch / jumper cluster, as well as a flexible clock frequency generator so it could run on both 68040 & 060 CPU's.

No I meant this for the high end. Beyond which there is nothing faster. You get the oscillator for your CPU's speed, and wait states adapt the memory to this speed.

Originally Posted by Mr B

However, i see issues here. A buyer with a flaky CPU will inevitably blame the card,... the CPU was bought as "working pull" from some asian guy... Makes it a nightmare to support.

I have a feeling the only way a card like this will see the light of day is if buyers accept "Individual Computers accept no responsibility for inserted components, card has been testing with this CPU and memory specs". It would save test-time, handling/ordering and outlay for him and reduce his commitment, which might make it happen (I hope! ) At the same time it's not very painful for the user, the card could be made cheap and therefore in volume, and PC RAM costs nothing and it's not that hard to get another used 68060 if one doesn't perform.

That's why he's been building cards only with chips (CPUs and RAMs) that are NOS, and that's why we've gotten 68030 cards and not any 68060 cards.

If we don't want to downgrade to 68030s when all the 68060 cards eventually die, we must support this "I only guarantee the board". (Unless someone else makes such a board or full 68060 accels that come with test CPU and memory, of course.)

Originally Posted by Mr B

And the sort of optimization Jens is doing to push the performance from the CPU's would be impossible with a wide variety of CPU & memory combinations.
B!

While it pays to push slow components to perform at max, there is nothing faster than 68060+modern PC DDR RAM for Amiga, so he'd only need to be as fast as Blizzard/Apollo, and I'm sure he'd easily do that with 10 years newer components without optimizing it to the edge.

I don't see a combination problem that could crop up. Blizzard/Apollo already supports PC memory. CPU timings are history, and there are cheap off-the-shelf chips to talk to a generic PC RAM of the type that he chooses. Even within the type he could support a quite narrow range of RAMs and there would still be thousands of stores that would have it in stock.

[Mr B]

Originally Posted by Photon

I don't see a combination problem that could crop up. Blizzard/Apollo already supports PC memory.

For which there are memories that simply wont work, or "works" with intermittent issues. Still don't see the problem?

Originally Posted by Photon

CPU timings are history, and there are cheap off-the-shelf chips to talk to a generic PC RAM of the 1 type that he chooses.

I like the attention to detail Jens brings, where timings aren't a "best effort" generic works with anything setting, but something he slaves over to bring us the best performance he can.

Originally Posted by Photon

No I meant this for the high end. Beyond which there is nothing faster.

I saved the first, for last. Your barking up the wrong tree. We'll get emulated 680x0 that will be a lot faster, and compatible with a wider array of software then a top of the line 68060, most likely long before the adaptation of "empty" accelerator cards. I'm not going to be surpriced when someone pulls a chip out of their rear end, that emulates a 68030 couple of 100 MHz. Right now the chips to run it of is to expencive, but they'll end up as NOS soon enough them to.
B!

[mfilos]

Originally Posted by vroom6sri

Jens - So, just coz I can't stand suprises, when you stated earlier that they " heat transfer pads will be fixed for shipping with an Amiga-related item..." does that mean we'll be getting a bonus item of some sort??!?!

Don't get me wrong, the fact that you're making such great hardware is good enough but I was just intrigued by your statement.

Heatsink packaged arrived today
Very well designed and with very good information even for people that never used a heatsink before in their life.



The Amiga-related item is an original Lisa chip that holds the transfer pad in place as you can see from the picture bellow.



Thanks for this Jens

[phl22]

Cooler added?

[Retrofan]

Originally Posted by phl22

Cooler added?

Yes, and it was for free and with another gift... I hate them

[phl22]

so bad?

[Photon]

Originally Posted by Mr B

For which there are memories that simply wont work, or "works" with intermittent issues. Still don't see the problem?

Of course there are RAMs that don't work. You could insert an Apple simm or a way too slow SIMM. The point that you missed is that I had no trouble buying two 32MB SIMMs for 8 EUR for my Apollo. That's it.

Originally Posted by Mr B

I like the attention to detail Jens brings, where timings aren't a "best effort" generic works with anything setting, but something he slaves over to bring us the best performance he can.

I saved the first, for last. Your barking up the wrong tree. We'll get emulated 680x0 that will be a lot faster, and compatible with a wider array of software then a top of the line 68060, most likely long before the adaptation of "empty" accelerator cards. I'm not going to be surpriced when someone pulls a chip out of their rear end, that emulates a 68030 couple of 100 MHz. Right now the chips to run it of is to expencive, but they'll end up as NOS soon enough them to.
B!

You can already do that with a fast PC and WinUAE, or with a PPC Amiga and some software. That, however, is not a continuation of Amiga.

I hope you understand that the reason to buy any hardware from Jens at all is because you want a continuation of Amiga? Because WinUAE can already run software faster than his accels to date and extremely compatible, and with a lot of extras like no need to buy an Indivision for a PC screen.

Let's make it a high end continuation of Amiga, is my simple suggestion

[phl22]

any opinions? my rom 3.0

[Mr B]

Originally Posted by Photon

Of course there are RAMs that don't work. You could insert an Apple simm or a way too slow SIMM. The point that you missed is that I had no trouble buying two 32MB SIMMs for 8 EUR for my Apollo. That's it.

YOU had no trouble, so therefor there is no issues, ever. Ok, i get it. In the real world there is still the possability of something being within the right specs, and fully functional, that simply wont play ball with some pieces of hardware. It still happens to this day, that some RAMsticks doesn't work with some motherboards / CPU's.

Originally Posted by Photon

You can already do that with a fast PC and WinUAE, or with a PPC Amiga and some software. That, however, is not a continuation of Amiga.

I'm not sure i follow your reasoning. If you got a accelerator which has a Motorola 68030@50MHz, or if you have a FPGA running the equivalent, your not going to see, hear, feel or imagine anything being different. What actually is different tho is that you can use a faster FPGA, and instead of being limited to the speedgrades that the 68030 actually was produced with, you could run one at 200MHz. (Number picked at random) The only difference here being the FPGA CPU, which as of right now is to expensive. This is NOT emulation in the same way that UAE is. This is using modern hardware to replace hard, or impossible to get bits and pieces to make new hardware for existing Amigas. It's like using a modern piece of ram instead of scavenging old ramsticks, since there simply isn't supply for it. And guess who does all of this?

Originally Posted by Photon

I hope you understand that the reason to buy any hardware from Jens at all is because you want a continuation of Amiga?

Yeah. That guy. He uses modern RAM coz he cant find the old stuff for a good price, and he used a FPGA chip to make a complete system. Not using them for CPU's but for the custom chips, i believe. I have no doubt that once it makes financial sense to make the accelerators using FPGA's he'll make em with FPGA's. And when you buy a accelerator with a FPGA design, you wouldn't know, ever, unless someone told you.
B!

[jbenam]

Originally Posted by Photon

I have a feeling the only way a card like this will see the light of day is if buyers accept "Individual Computers accept no responsibility for inserted components, card has been testing with this CPU and memory specs".

I remember reading that Jens said that he hates receiving support tickets for products that don't work because of other variables. That's why he hates the Apollo accelerators, because they gave lots of problems with his products, and people blamed them instead of thinking that their Apollo accelerator was the real cause behind it.

Knowing that, having users add RAM and CPU by themselves would mean a lot of support tickets like "your card sucks, it doesn't read my RAM", just because it has crappy timings or it's incompatible with the card for some strange reason - that said, I don't foresee Jens EVER going that route.

We have to hope in another Amiga hardware dev, or in a free open-sourced PCB (which should be four-layered to support 68060, iirc?) project and have people build them by themselves

[Photon]

Originally Posted by Mr B

YOU had no trouble, so therefor there is no issues, ever. Ok, i get it. In the real world there is still the possability of something being within the right specs, and fully functional, that simply wont play ball with some pieces of hardware. It still happens to this day, that some RAMsticks doesn't work with some motherboards / CPU's.

It was easy. That means it's easy to get memory for your accelerator. Try getting a replacement memory for current ACA products.

Are you daft for real?

Originally Posted by Mr B

I'm not sure i follow your reasoning. If you got a accelerator which has a Motorola 68030@50MHz, or if you have a FPGA running the equivalent, your not going to see, hear, feel or imagine anything being different. What actually is different tho is that you can use a faster FPGA, and instead of being limited to the speedgrades that the 68030 actually was produced with, you could run one at 200MHz. (Number picked at random) The only difference here being the FPGA CPU, which as of right now is to expensive. This is NOT emulation in the same way that UAE is. This is using modern hardware to replace hard, or impossible to get bits and pieces to make new hardware for existing Amigas. It's like using a modern piece of ram instead of scavenging old ramsticks, since there simply isn't supply for it. And guess who does all of this?

Yeah. That guy. He uses modern RAM coz he cant find the old stuff for a good price, and he used a FPGA chip to make a complete system. Not using them for CPU's but for the custom chips, i believe. I have no doubt that once it makes financial sense to make the accelerators using FPGA's he'll make em with FPGA's. And when you buy a accelerator with a FPGA design, you wouldn't know, ever, unless someone told you.
B!

Now we're talking! I'm all for remaking Amiga chips in modern components, as I've mentioned 1000 times here. But you mentioned emulation, which is more like the PPC way.

However, a 68030 with FPU in FPGA would be hopelessly slow compared to 68060 even at 600 MHz. On top of which this would create a condition where someone would spend years making such a one before a card could be made.

You're replacing the possible with things that aren't here yet, while I'm proposing that a card with sockets that could be made cheaper and in larger volume than a custom card. I would actually buy a 100% compatible humble 68000 CPU and Amiga chipset remade in FPGA the second it was put on the market. It seems you haven't read my passionate posts on this subject, but this is a dream vision for me. However, my suggestion is a different product altogether. It's a CPU board for non-FPGA CPUs.

jbenam: I have heard of Jens' dislike for Apollos and am proposing a replacement to get them off the 'market' and get him in control. Do you know of any better way to make Jens happy?

That said, the ACA series do have support tickets, all the more reason a generic card with narrow specs and no responsibility/limited warranty period should be more attractive for a developer.

The dev could of course sell the cards with CPU and RAM tested and ready, but accept responsibility only for the board as at least the CPU would be a used part. But I think it IS an important point to separate the board from the replacement parts since some time in the future every user would need to replace socketed parts.

[Mr B]

Originally Posted by Photon

It was easy. That means it's easy to get memory for your accelerator. Try getting a replacement memory for current ACA products.

Replacement memory is quite easy to come buy. A lot more so then the 72 pin variety your so fond of. The process of replacing them is a bit harder, but thats not down to availability, thats down to performance, and eliminating variables that could bring issues to costumers.
But then your not looking for replacement, your looking to change the memory, for some reason or another, trying to prove that 72 pin memories in some arbitrary way is "better".

Originally Posted by Photon

Now we're talking! I'm all for remaking Amiga chips in modern components, as I've mentioned 1000 times here. But you mentioned emulation, which is more like the PPC way.

You do realize that a softcore is a instruction-set running on a CPU other then a 680x0? It's emulating a 680x0.

Originally Posted by Photon

However, a 68030 with FPU in FPGA would be hopelessly slow compared to 68060 even at 600 MHz. On top of which this would create a condition where someone would spend years making such a one before a card could be made.

Years? How do you figure? The softcore is more or less already made, people run everything from 68000 to 68050. The point is that with a flexible design you could simply press a few buttons and go from a compatibility set to another. You need a fast 68000?, load one softcore. Need 68030, load that one. 68060. Well, right now i think you need to develop it first, but yeah, once done, just load that one. The only thing needed is for the FPGA's to drop in price. A lot. It sure wouldn't hurt for faster FPGA's to be available to us either, but for making the equivalent of the current ACA accelerators there are more then enough "oumpf" available today. So the next gen may very well have what it takes for 68040, and so on. It will be a while yet for sure, but it's a lot more likely then barebone accelerators.
B!

[Schoenfeld]

Originally Posted by Photon

I have a feeling the only way a card like this will see the light of day is if buyers accept "Individual Computers accept no responsibility for inserted components, card has been testing with this CPU and memory specs". It would save test-time, handling/ordering and outlay for him and reduce his commitment, which might make it happen

Nope, it might not make it happen, it might even make it illegal. Customer rights are clearly outlined in Europe, and the basic idea behind these EU directives is that nothing at all may be sold with no warranties. You have to guarantee that a device is fit for a certain purpose, and plain renaming the purpose does not get you around the laws.
There has been a case in Germany where a "clever businessman" wanted to sell light bulbs as "heat bulbs". Essentially the same that you're trying here: He claims that the purpose is to produce heat, not light, so he wanted to get around the "energy-saving lamps only" directive for any light bulb over 75W that has been in place since last year. Guess what - he failed in court.

I would most probably fail the same way if I declare a card "for viewing purpose only".

Originally Posted by Photon

Even within the type he could support a quite narrow range of RAMs and there would still be thousands of stores that would have it in stock.

Starting from DDR, memories must run at a minimum frequency, and the number of pins to talk to these chips gets higher due to some signals being differential (some even bidirectional-differential, which is not available on any 5V-tolerant CPLD).
However, it does not make sense to use DDR, because their first-access-penalty is not faster than the first-access-penalty of 200MHz SD-Ram, which is what I'm using on my new accelerators. However, it took me ages to dig up that type of memory. Many SD-Ram vendors have specified 200MHz types, but they have never brought it to the market. Even though I'm not running the chips at 200MHz, they still save me a cycle here and there, which would be impossible with off-the-shelve 166MHz types.

Even though I have designed the ACA1230 and ACA1231 cards with as little user options as possible, they still cause an enormous amount of support work. Seeing that even the battery for the RTC causes support work, I would NEVER let the user exchange anything critical like a CPU or memory. Here's where EU laws come in again: If the user has the possibility to make a mistake and it's technically possibe to avoid damage, then the product designer must do it. For me, this would mean that I'd have to build a sensing circuit that automatically sets CPU core voltage if you're upgrading from 68040 to 68060. This would be assimung that I'd include an 800,- EUR tool with every card, because without a PGA-18 puller, you WILL scratch and dent the CPU socket, which essentially de-values the card.

Imagine someone buys a card without a CPU, inserts something that he just happens to have, and then finds out that the card doesn't do what he wants. EU laws permit him to send the card back within 14 days without even saying why - a short note like "please refund" is enough, and the reseller has to take back the card. Can you imagine the kind of hassle if the reseller has to take back a card with a scratched CPU socket? Can you imagine how many customers would accept a card with a scratched socket in the first place? None at all.

I would think that only one person that's reading here has access to a PGA-18 puller. None of the resellers I'm working with has such a tool, and I'm pretty sure that none of them will buy something like that. Is that a bad thing? Surely not. Those of you who are now thinking that EU directives are holding us back should reconsider their opinion. EU directives protect the customer, and it's good that way. You're getting a product that's guaranteed to be fit for the purpose you've bought it for. And if you later find out that you got the wrong idea about the product in the first place, you don't lose any money. To avoid that situation, vendors and resellers must be as honest and as precise as possible in their product description.

Jens

[roy bates]

i agree with what youve said.(the laws are very strict on this matter for a reason)

but surely theres a work around for peaple that have a "known working" cpu,by this i mean anyone who wants to do such a thing could send you the cpu at there cost.(with a list of revisions of mask sets tested by you as a comparison)
then if it dont work with your card,you can simply remove it with your tool and send it back at the cost of the person who sent it to you.(admittedly you have to have someone do the testing)
which is probably the main reason for such a card not being made in the first place.(by this i mean the cost of the cpu)
the testing could be done at the time the card is being tested itself,altough i beleive this could be time consuming so maybe it could be a separate thing done when the card has to have the cpu inserted(this may be counter productive i know as you probably have to contract out a production line for the card assembly,and i dont know wiether you test the cpu sepratley from this)

as for the memory,128-256mb would be great for most peaple wouldent the sdram interface on your aca's work with an 060?just a thought.