Linked by Thom Holwerda on Tue 5th Jan 2010 13:44 UTC
Amiga & AROS After days of wild speculation and ridiculously fast-growing threads on AmigaWorld.net, we finally know most of what we need to know about the new Amiga. This is not just a random PowerPC evaluation board that you can stuff in a generic case - no, this is an all-new system with a custom motherboard, and some very, very interesting innovations - like a fully customisable co-processor. Twenty-five years after the introduction of the first Amiga, this is one heck of a machine.
Thread beginning with comment 402853
To view parent comment, click here.
To read all comments associated with this story, please click here.
RE[2]: Innovation and all that...
by PrimalDK on Wed 6th Jan 2010 18:38 UTC in reply to "RE: Innovation and all that..."
PrimalDK
Member since:
2005-07-12

Skimmed the spec docs and you're right -- my bad. The merits of "go to the source" instead of just reading the sum-up...

It seems it divides its clock cycles between each core as well, which somewhat lessens the usefulness of a multi-core design.

In other words, a general purpose CPU at a nominal 2.5 GHz clock frequency will have 1x nominal freq per core but the XCore only 1/n per core, where n is the number of cores.

By real-time tasks I suppose you refer to hard real-time. All general-purpose processors do soft real-time well enough -- well, actually, the operating systems do. You certainly are able to watch your videos and do your software-mixing of 128 audio tracks without hickups -- all inheritly soft real-time tasks (dead-line with escape strategy).

One could argue that all dead-lines are merely a matter of resource availability/use (excepting execution strategies that include prediction)...

An interesting property is the ability to quickly scale down the frequency automatically or as explicitly directed to 1/100th the nominal power, making 256 processors use 2.56x the power of 1 running full steam.

Try that with a modern C/GPU.

Reply Parent Score: 1

SamuraiCrow Member since:
2005-11-19

They're probably using an 8-stage pipeline in the processor. In this case switching between the threads on each clock cycle is used to avoid collisions between the data of the independent threads. It's a cheap trick but it gets the job done.

Adding cores in parallel only increases the number of I/O lines. To get added speed you have to add processor chips in parallel rather than using more cores per chip with this one. It's probably because the multicore chips have only one local-store bus to fetch from. Maybe it's a good thing that A-Eon is using the cheap chips.

Reply Parent Score: 2

Dasher42 Member since:
2007-04-05

It sounded like you were thinking that the XCore is an FPGA. I agree, an FPGA as a standard part of a desktop computer would be brilliant. I'd love to see hardware acceleration on-the-fly for more tasks. I would think that an FPGA would have been a natural choice, especially given the success of the MiniMig in emulating a classic Amiga.

Reply Parent Score: 1

rhyder Member since:
2005-09-28

I would have prefered to see the OS come supplied with a standard API to use the GPU as a DSP. This is already proven technology. If you need more power, you could always add a second PCI express graphics card.

Reply Parent Score: 2