killermike wrote>"A 1ghz processor will seem very slow by 2010. Particularly on an OS that can't fully take advantage of the multiple cores. By that time, the emulated solutions will be even further ahead"
Part of the motivation of putting multiple cores on the same chip (or at least in the one package) is that *each* does *not* need to be clocked at multi-GHz speeds to acchieve overall reasonable performance (many PC laptops today using CoreDuo's are clocked at 1.6GHz - in the past single core PC's were being clocked in the 3.5->4GHz region). Emulated solutions (in speed terms) don't have much of an advantage and that's not likely to change. If faster single or dual core Intel xScales (c. 1.2GHz) or these ARM9 based multi-cores (c. 1.0GHz) appear they'll improve performance of the native systems by a lot more - I don't frankly see much change in a two year timeframe on x86 performance (other than adding 3 (AMD) or 4 cores (AMD & Intel)) and that *won't* necessarily translate to much emulator speed increase as that is largely clock rate bound which may not change much.
The key issue is we need a new computer based on faster ARM cores (be they quad - in 2 years or dual xScale now). The number of cores is probably less significant than the (a). Clock speed and (b). The Cache size. So a dual Core xScale *now* at 1.2GHz may be a *better* platform than a quad core 1.0GHz ARM9 in 2010.
In either event one version of RISC OS (ROOL's release of RO5.xx) offers an open option for upgrade, Castle/Iyonix have experience of manufacturing whole ARM based computer systems they'd be the logical ones to push this. ROL could do OS development instead - but have a number of mountains to climb (i). Their OS is largely intended to run on machines *identical* to RISC PC's, (ii). Emulated to look like a RISC PC or (iii). Have restricted expansion options [hence requiring less driver/hardware related OS development] e.g., A9Home. For them (ROL) to produce a multicore aware RISC OS would be a challenge for those technical issues thrown up and considering their other vendor and customer committments. On the one hand if they produce an OS that can make native RISC OS run faster than emulated that could hit one of their big licensees. If they abandon their RiscPC base to concentrate on a multicore system then they alienate their subscription customers.
I wouldn't absolutely rule out ROL doing something in the multicore area - but it would be yet one more speculative demand with so many others demanding attention (RO for A9, Select updates etc.,) and that just makes it difficult for them.
It would also mean them trying to persuading a hardware vendor to push into a full computer design with multicore capability. That would not be a typical embedded design (like A9) and might not appeal to Ad6 whose focus is embedded - so who would build such a mytical ROL inspired computer? Short of ROL *themselves* commissioning a new machine I can't see this happening, that having been said I'd be happy to be proven wrong.
If (and it's a big *IF*) RISC OS appears on new hardware it'll more likely to be a derivative of Iyonix running RO5.xx. In either event IMHO something needs to be done *soon* regarding an Iyonix replacement and multicore (even dual ones like Intel were touting - never mind quad quad A9's) may be the ticket.