View Single Post
Old 10 March 2017, 18:55   #11
matthey
Banned
 
Join Date: Jan 2010
Location: Kansas
Posts: 1,284
Quote:
Originally Posted by Juz400 View Post
The ATI Radeon 9200 PCI appears to used in some Amiga setups, that was the RV280 chip
ATI X600 was the first PCIE card, that was the RV380 chip
Are these going to be completely different cookies? Nothing similar between them?
There is only a 68k 2D Radeon driver. A 2D driver is fairly straight forward but offers little benefit over SAGA. The Radeon would likely give faster 2D gfx operations but slower bitmap accesses. The 2D operations quickly reach a point of diminishing returns anyway. The 2D operations in an ASIC could be faster than the Radeon (less latency anyway). For 3D, the Radeon is going to be faster than SAGA in every case as SAGA does not have any 3D support currently. A 3D driver is much more difficult to write, the Radeon does not look like an easy chip to program and the documentation I have seen does not appear complete or easy to understand. I would expect a 3D driver to take 6 months for someone experienced in writing GPU drivers (perhaps 3 months for someone experienced in the GPU architecture) and would be expensive by a professional programmer.

It is possible to add 3D support in the FPGA. The synthesizable HDL code for a 3D GPU core can be bought (PowerVR, VideoCore IV, etc.) or a new home grown core created. An SIMD unit in the CPU can do some 3D operations providing a 50%-150% speedup while being low latency and very energy efficient. Some of the old GPUs (and probably very high efficiency GPUs) are based on SIMD units. The Amiga Hombre chipset using PA-RISC was likely primarily SIMD driven as we examined in a thread here on eab.

http://eab.abime.net/showpost.php?p=...8&postcount=22

Yes, it was handicapped by RISC fallacies and wouldn't have been much faster than a 68060 but there wasn't much competition in 3D then. It could have run in parallel to the 68060 freeing it to do other work. The Amiga blitter could be used in parallel until the 68040 or 68060 where it became a decelerator if used. The same blitter (logic) in an FPGA today will outperform many CPUs and provide a huge speedup for 2D operations today.

Quote:
Originally Posted by Juz400 View Post
You can pick up those cards for the price of a Whopper and fries.

...

I guess commodore made a mistake in fitting the A500/A600/A1200 with a bus connector underneath because nobody would want more than 512k/2mb
I understand the advantages of cheap commodity hardware and expandability but tight integration and standardization may be more important as chip technology slows down and Moore's law comes to an end. Chip designers can no longer add all the bloat (transistors) they want and count on die shrinks to make it possible (efficiency matters again). Create a 64 bit CPU with 20% worse code density and you will likely end up with a couple of CPU cores (for SMP) less than a 32 bit CPU which is able to improve code density by 20%. Could the 68k improve code density by 20%? I believe ISA improvements could improve it by 10%. If loops did not need unrolling and functions did not need inlining for performance then that may improve it by 10% more (the cost may not be zero but it only needs to be very close to zero). Most of the computing world does not need 64 bit addressing and pointers (if a CPU has good code density) which is going to be some 10% slower to begin with (everything else equal). Energy efficiency is also more important now as smaller closer together transistors leak current between them. The Amiga could use this opportunity to catch up if there were any tech savvy leaders with a vision.

Quote:
Originally Posted by Juz400 View Post
We prob met before on the Natami forum a while back, everyone screaming we need this and this and this until it broke.
Yep. I was on the Natami forum. Good times and very educational. I can't believe how much interest the Natami generated only to be ignored. I posted this on a thread here on EAB.

Quote:
Originally Posted by matthey View Post
There are many low profile and lurker Amiga users. The Natami forum shows how they can come out of the wood work even without advertising. The Natami "MX Bringup Thread" with 761487 views still amazes me and is evidence of how incompetent the current Amiga owners are to ignore and avoid the much larger Amiga market and the economies of scale which could be used to make it affordable.

http://www.natami.net/knowledge.php?b=1&note=33366
Of course the Natami forum is finally gone but that was the link with 761487 views at that time.

Quote:
Originally Posted by Juz400 View Post
I was thinking `Future revisions` with this whole Idea, you mention cheap ethernet which is on the money too, how many little WiFi cards are there floating about for peanuts that are tiny and would disapear like a CF card inside an Amiga with a minipcie slot.
We need to get more Amiga users online one way or another (preferably without emulation) to invigorate the community. Another option is adding USB and using those cheap little wireless USB adapters. USB would be handy for other things too but the fastest way to add it would be with a USB chip which adds cost (not just the chip but larger board and layout). Oh, I suggested the cheap mini PCIe WiFi laptop cards to Gunnar at one point also. Yep, they are practically free.

Quote:
Originally Posted by Juz400 View Post
Its absolutely something to plan ahead with now even if not implemented in hardware yet due to current BOM costs.
Im an oldie like most people here, I understand the economics of running a business with enough profit to put back into the R&D and production costs without making your product too expensive for anyone to enjoy
If Nobody can afford to enjoy it, everyone looses
There isn't really plan ahead with Gunnar though. He likes to hyper-optimize the core for FPGA and today. The so called "team" has competent CPU designers but Gunnar doesn't always know best. I left when the research stopped and the "team" became a joke.

Last edited by matthey; 10 March 2017 at 20:06.
matthey is offline  
 
Page generated in 0.11791 seconds with 9 queries