The sales rate is approx 120 thousand units a year, BUT, thats if people who want it can afford it. Thats how mass production works, or suposed to if you dont want infinite debt and decadal bankrupts. If you can pay off so much of your RnD, design, implenmentation, licensing etc costs in teh first few hundred test run cases, then people who supplied you can back source parts more easily, but also hold you to ransom, depending how unethical they are. If your getting parts from the US, you could be up the creek.
Whats most sickening, is that the PA6T is a purely Virtual, paper based CPU. No company makes it as a singular item on a dedicated line, its just another data file on the TMSC megafabs. Same as ARM.
If the design was as modular as they say it is. .5000 sub domains for clocking and energy control? then dropping in upgrade modules as technology advances should be far easier?. .USB3, SATA 3, PCI-E 3 64 lanes, 16 cores, larger Caches, Flash, MEMS,...
How would you like the SoC with a Gig Flash chip for storage embedded?
Now that I found what the term is Ive been looking at for many years, Id love to see a homeomorphic encryption CPU variation. No main OS can run on it due to the thousand fold processing requirements, but that would still let AOS equivalents run very nicely indeed.
I like the Real Time Ray Tracing.. but.. wouldve been better if that was the default state of the Workbench, for demo purposes.
If Kefrens, Desert Dreams is used as benchmark, then amodern GPU should be able to point manipulate approx 60 billion points at 50 frames a second? or 300 billion points a second approx.
Lets see a demo where a 1000 cube voxel 3D workbench is manipulated in real time, using a GPU fully. I dont expect it for a couple years and by then we will have two more generations of GPU to have problems with, but its a nice place to start.
Latest use for XMOS.. switchable less than 10ns.. A new Trillion FPS camera, imaging system been anounced. It sees light as it moves. Slowly. So, given the retriggerable method used, I propose XMOS be considered for billion frame per second plus imaging. 8)
_________________
The older and more respected a scientist is, the longer it takes to prove him wrong. |
9 crawler(s) on-line.
46 guest(s) on-line.
1 member(s) on-line.
RodTerl 
