[Oberon] Oberon board availability (was Moving oberon to RISCV?)
skulski at pas.rochester.edu
Sat Jul 28 15:55:35 CEST 2018
> 10. Pepino can run RISC5 at 25 MHz. The clock speed is limited by the ASRAM chip.
> Not true - Magnus has implemented a 37.5MHz version; and the asynchronous
> SRAM is not the limit, the design of the RISC5 plays a part. Magnus
> explains this well.
I stand corrected. The clock speed increase is valuable, though not dramatic. I ascribed the speed limit to ASRAM based on my reading of Pong P. Chu's book on FPGA prototyping with Spartan 3. This book elaborates on the same Spartan 3 kit as you used for Oberon. Professor Chu devoted an entire chapter to the ASRAM interface. The limitation is that the nominally 10 ns chip cannot handle transactions every clock cycle (a well known fact). The idle cycles which are needed to avoid contention are bringing the speed down. In conclusion, using ASRAM is limiting the memory speed to - more or less - what you and Magnus have achieved. There may be other bottlenecks in the system as well, but this particular one is fundamental and hard to improve upon. Improvements in this area need another kind of memory.
A slight improvement can be achieved using ASRAM chips rated at 8 ns rather than 10 ns. Such chips are pin compatible with the 10 ns ones. The substitution can be done during the assembly.
> I don't agree that the Pepino is "closed", it has a great set of
> directions in which it can expand, so I think your implied criticism is
> rather unfair - but maybe you didn't intend to sound so critical.
Maybe I should have phrased it differently. I meant to say that Pepino is a tightly packed board where adding features will be sort of hard to do. Being a board designer myself I know how easy or how hard it can be. It is always possible at the expense of ripping up the design, so "closed ended" is always relative. My design is "open ended" in the sense that I tried to ease this task somewhat. But there is price to pay for such a flexibility, which is higher cost due to a mother/daughter design. It is adding a couple hundred $$ to the assembly. There is no free lunch here, unfortunately.
> Anyway, with Magnus' latest offer, Pepino is a really great deal for a
> Project Oberon machine, thank you Magnus!
Full agreement from my side. It would be wonderful to see Pepino's derivatives. For example, a version of Pepino with a ZBT chip rather than ASRAM. A price increase would be quite moderate. But it is not clear whether this improvement would increase Pepino sales. Rerouting the board is quite a bit of work, and then there is the question of firmware.
More information about the Oberon