View Single Post
Old 13 March 2017, 00:37   #25
Join Date: Jan 2010
Location: Kansas
Posts: 1,284
Originally Posted by Juz400 View Post
Looking at a GTX1070 benchmark via one of these devices shows that at x1 speeds you can still push an awful lot of data through PCI-E a x16 wont be fully maxed out for years yet.
The slowest PCIe 1x is 250MB/s where the fastest common 32 bit PCI@66MHz is 266MB/s (There is 64 bit PCI but it is rare). Each additional PCIe lane doubles the throughput and newer spec versions are significantly faster as well.

Originally Posted by Juz400 View Post
Would a SerDes have access to the memory onboard the Vampire card?
So you could say have an interrupt signal in your code that instructs the cards CP to use pull mode and grab a list of instructions from memory then wait until your next interrupt
If this is the case, then it is not impossible to get the card to do something.
An FPGA is how it is wired (programmed) internally. Lower end FPGAs do not have a memory controller so they must be created in the FPGA. The Altera Cyclone V (as I recall) has a memory controller so I expect most mid to high end FPGAs do. SerDes are only in mid-high end FPGAs so I expect these FPGAs have a memory controller also. I expect a memory controller would simplify transfers to/from the SerDes or at least save complex logic in the FPGA which would do it. The SerDes would be severely handicapped without access to memory although most FPGAs have SRAM (perhaps used as a buffer for the SerDes?). I read a little about the SerDes in an FPGA once but this in not my area as you can see. If I understand what you want to do correctly then I expect it is possible with proper FPGA programming and an FPGA with SerDes probably has tools to make it easier.

Originally Posted by Juz400 View Post
The programming of the shaders and vertex units looks very complicated and getting a `graphics card driver` up and running would be quite the labour of love.

But if you look at the demo scene, they get a picture and sound out of almost anything you can solder together, program and wire up to an AV system. Hardware banging the card just to see what happens could be quite fun for the more adventurous programmers out there. It could be 1987 all over again with a new copperlist and blitter registers to figure out!
I modified the Voodoo 3-4 Warp3D drivers for my Mediator (mainly to fix bugs and horrible optimization at first). This card has the best gfx hardware documentation available (still not for beginner programmers) and is fairly easy to program. More modern cards are generally much more difficult to program. Yes, programmers could program various gfx cards at the bare metal (educational) but I think this is the wrong philosophy. IMO, the Amiga would benefit from standardization with one 3D hardware (upgradeable but similar design) and programmers should choose to use the OS and drivers provided which would better utilize multitasking, SMP and GPU parallelism. This would also allow the gfx to be integrated on the motherboard instead of transferring data across relatively slow buses. This would make the Amiga more like a console with efficient use of hardware to make up for affordable hardware with less performance. Modern gfx cards are a big hindrance to small designs because of the heat they produce so it is important to do more with less like the Amiga used to be good at. Sure, it would be nice to have optimal and bug free drivers for hundreds of gfx cards but I believe this is a less feasible and less practical solution for the Amiga.
matthey is offline  
Page generated in 0.07333 seconds with 9 queries