FPGA based accelerators?

I think there may have been a similar thread somewhere here in the past, but I am wondering, as there are currently some Amiga accelerators in production, are we getting close to the point where someone will make an accelerator using a FPGA running faster than an actual physical Motorola processor can run?

An accelerator running at 200 to 300 mhz in 68k, with a quarter of a gig of memory and perhaps some built in networking and RTG graphics capabilities would be extremely appealing to me. I want to have a machine that I can not do any business work on but still use some entertment software and do a little browsing, and maybe write a little code for.

I nice dream my friend....

But I think the 68k-CPU's world is not very interesting for hardware manufaturers.

Look at Minimig, a very good & nice solution. But very expensive and only emulates A500...

But if a whole A500 can fit onto an FPGA, surely a CPU could fit and run faster? I do not know much about electronics, but maybe it could run on existing boards? Say on the ACA1230 swapping out the 68030 and replacing it with the FPGA and using a faster crystal?

I'm not an electronic master either, but I think that if it would be big market for such expansion cards, we will have it already.
But as Amiga retro-market is relative small, none is interested to loose money by developing such board.

In short:
ROI will be negative.

Sure it's possible m8.

But look at prices for Amiga expansions, accelerators, ......

The new ACA's are not cheap. The solution for A600 is great, but for A1200 is a very poor solution, any B1230 IV is more powerful and can fit 50Mhz clock, 128MB RAM, FPU & SCSI (this one can adds 128MB RAm more).

I think one PGA board will be very very expensive.......

If the only difference between the ACA1230 running on a genuine Motorola 030 and a FPGA is the cost of the FPGA and the faster crystal, I do not see the issue. I do not think price should obscure the practicality. If the only way to have a 200mhz 68k machine is to run it on a FPGA, and the development costs come down to writing the 68k emulation for the FPGA (which I would suspect already exists) and testing, and it would otherwise run on existing hardware, I would suspect it would be more than viable as a business proposition.

But even if it is not, this is still something it has to be worth exploring if hobbyists could do it themselves.

Quote:

---

Originally Posted by Jumping Anaconda

But if a whole A500 can fit onto an FPGA, surely a CPU could fit and run faster? I do not know much about electronics, but maybe it could run on existing boards? Say on the ACA1230 swapping out the 68030 and replacing it with the FPGA and using a faster crystal?

---

Sorry to correct you but the cpu of the minimig does not run in the fpcga there is one on board that is overclockable.

---------- Post added at 13:33 ---------- Previous post was at 13:27 ----------

• FPGA: Xilix Spartan-3 XC3S400
• Loader: PIC18F252 using standard MMC/SD card
• Memory: 2 MByte (1M x 16) 10ns SRAM
• CPU: Freescale MC68SEC000AA16 (16 MHz low-voltage MC68k-clone)
• Power: 5.0 V only (on-board LDO regulators creates the needed voltages for FPGA and logic)
• Video: 12 bit via simple resistor D/A (4 bits/color)
• Audio: LM4834 headphone/line amplifier with DC-controlled volume
• Serial: MAX232 compatible level converter

Quote:

---

Originally Posted by Jumping Anaconda

If the only difference between the ACA1230 running on a genuine Motorola 030 and a FPGA is the cost of the FPGA and the faster crystal

---

It is not.
Plese add cost of FPGA software development/testing/debugging/after sale support/future upgrade:
2000hours x 60EUR/hour = 120.000EUR
Now divide this number by unit sold; let's say 1000 and it will give you 120EUR price increase.
Plus 20/30EUR cost for such fast FPGA (fast FPGA is more expensive than normal 030 CPU).
In total: 150EUR price difference per piece.
And distributor price for 030@200HMz FPGA card will be 300EUR
Very close to 060 second hand boards.

Second:
For what to you need this extra speed?
For WHDLoad?

Third:
What if I'm wrong and it will require not 2000 but 5000hours (or more) of development?

Fourth:
It's always good to say something and to state that you are not specialist and don't know what you are talking about. ;)

Like I said, does it really need to be that complicated? As far as I was aware, FPGAs are designed to be used as drop in replacements for old or out of date ICs. There is off the shelf code for 68000s already that runs on FGPAs such as the Spartan III. What is to stop a hobbyist dropping that into a real A500 right now as a straight replacement for the standard CPU, as long as they can find the right package?

And if the 68030 code exists, and it it can run it fast enough, why not drop one into an accelerator and clock it up to as fast as it will go?

I expect it is more complicated than this, but I would like to know what the practicalities are. I mean, do you have to flash the cpu to the FPGA at every cold boot? Or can it run entirely out of its own package?

Quote:

---

Originally Posted by Jumping Anaconda

Like I said, does it really need to be that complicated?

---

Yes, it is.