So what is XMOS good for?

[Hans_]

Drives up price.

Makes it look like they are making unique hardware that would be worth the cost.

Makes them look busy.

Uh, those XCore chips cost less than $10 US. I know that it's fun to be negative and piss on someone else's work, but do try to keep the criticism sane.

XMOS call their chips "software defined silicon." What I think that they mean with this** is that there are no built in peripherals. Instead, its low latency event-driven multi-threaded architecture means that you can create any peripheral I/O that you like using software. This is something that "normal" micro-controllers can't do (or would struggle to do reliably).

What would it be used for on the X1000? No idea exactly. It's greatest use is likely to be with the Xena slot. Forget about using it for graphics or emulation, because it's no processing powerhouse, unless you string a whole set of these cores together (via the Xena slot) and figure out how to use them as stream processors. Given its strength with low latency event driven processing, this would be useful for hardware hackers, people doing robotics stuff, etc.

"But that's not desktop," I hear some people say. Who cares? People call the Amiga a "hobby" computer, so why not put stuff in it that caters to computer/electronics hobbyists?

Why not just put it on a PCI card or on a USB dongle? Well, what's the point of having a low latency device if you then put it at the end of a high latency bus? Sure, you could still develop XCore stuff via a USB connected Xcore chip, but you would immediately take anything that would work better being tightly integrated off the table.

Why put the XCore chip on all machines instead of just the hobbyists who want to use it? Well, let us say that one of these hobbyists makes some cool gizmo that you like, and want to have. With standardized hardware you could ask the hobbyist to make you one, and plug it straight in.

All of this is completely hypothetical. We're not going to know what can be done with it until the machine is available, and in the hands of those who can develop stuff. If you don't see the point in the XCore chip, then maybe you should focus on the rest of the machine; multi-core >1.6 GHz, PCI-Express, etc. Surely that's interesting enough in its own right?

Hans

** I'm not familiar with their architecture, so I could be completely wrong

[Hans_]

8 XMOS chips isn't even going to scratch the surface of what a modern GPU can do in terms of parallel processing, even using basic GLSlang (let alone CUDA, Stream or OpenCL).

Agreed. However, GPUs can't read back from the destination buffer, so they're not good for Infinite Impulse Response (IIR) filters. It's possible to do with a GPU via tricks (with sizable extra overhead), but this is something that a set of stream processors might have an edge. Of course, you're still contending with a brute of a processor, so 8 XMOS chips still won't do.

IIR filters are used in audio processing, and Gaussian blurs of any size can be done efficiently using IIR filters, whereas a Finite Impulse Response (FIR) based Gaussian blur slows down exponentially with blur size.

Hans