Amiga.org
Amiga computer related discussion => General chat about Amiga topics => Topic started by: ElPolloDiabl on January 13, 2011, 01:53:56 PM
-
DARPA says that they have a way to make custom ICs on small production runs using electron beams to etch the silicon.
Article here, don't forget to read some of the comments.
http://www.theregister.co.uk/2011/01/12/darpa_asic_announcement/ (http://www.theregister.co.uk/2011/01/12/darpa_asic_announcement/)
-
No seriously I'm gonna make Amigas with a blowtorch and gelatin.
-
I wasn't aware that Amiga Inc. or CUSA had any plans to get into chip design, much less manufacture. Can you connect the dots for those of us who are less imaginative?
-
That is actually pretty neat! Although, I think their idea of "small scale" is still pretty big.
-
No seriously I'm gonna make Amigas with a blowtorch and gelatin.
Cool! This is a breakthrough! Where's your waiting list?
(The beam etching does actually sound very cool... But, seriously... I have a feeling the price will be WAY out of hobbiest reach for quite some time.)
-
I wasn't aware that Amiga Inc. or CUSA had any plans to get into chip design, much less manufacture. Can you connect the dots for those of us who are less imaginative?
The Natami people, on the other hand, might find this very interesting if they're looking to do new, faster models a few years down the line and their 68050 (or successor) and SuperAGA cores are well tested. It could allow them to reach far faster speeds very cheaply in comparison to keeping everything on FPGAs.
I for one couldn't care less if it bears the actual name Amiga or not - I care more about what software it'd run.
For that matter, if it gets cheap enough, someone might step up to the plate and offer replacement chips for actual classics for repairs etc.
Of course none of this is much better than free fantasy right now, given that the technology is likely years away from commercialization and even further away from being affordable *enough* for "our" use.
-
Of course none of this is much better than free fantasy right now, given that the technology is likely years away from commercialization and even further away from being affordable *enough* for "our" use.
Agreed. In fact I don't get the impression from the article that this is about making chip manufacture possible for your typical man-on-the-street, or even small businesses. The title even says: "Custom ICs in small numbers to be cheap as (normal) chips". I read this as, this will reduce cost overhead for small batches, but that's about it.
-
I think minimig shows the future of custom Amiga hardware. I don't think we are cutting edge enough for experimental manufacturing techniques.
Interesting stuff though.
-
This technology would have to make it into mainstream production first - that could still be years away. However once it is done, I'm sure that major foundries will pick it up.
However how much would making 100 Natami cores on this process cost? It's probably still more than 100 FPGAs. And in two years time the FPGAs will be cheaper for the same features, or faster and larger for the same cost. If Natami has something performing like a >200MHz 68060 later this year already, I think we'll be okay for classic Amiga enhancements for a while :-)
-
Cool! This is a breakthrough! Where's your waiting list?
(The beam etching does actually sound very cool... But, seriously... I have a feeling the price will be WAY out of hobbiest reach for quite some time.)
Two more weeks.
-
Two more weeks.
hahahaha! :D
-
Neat technology. Sounds really complicated though.
Shouldn't you have titled this Another possibility for the future of the Amiga?
Since the Amiga doesn't have just one future, its a multiplicity of futures.
-
Cool! This is a breakthrough! Where's your waiting list?
(The beam etching does actually sound very cool... But, seriously... I have a feeling the price will be WAY out of hobbiest reach for quite some time.)
Ahem... www.amiga.org/forums/group.php?groupid=50 (http://www.amiga.org/forums/group.php?groupid=50)
-
Screw you, Kelsey.
-
This technology would have to make it into mainstream production first - that could still be years away. However once it is done, I'm sure that major foundries will pick it up.
However how much would making 100 Natami cores on this process cost? It's probably still more than 100 FPGAs.
Who knows? But the Reg claims their aim is for it to bring the unit costs for single wafer runs (anything from a couple of hundred to a couple of thousand units depending on size of the chips and wafers) down to the same as you get for large scale runs, so presumably that'd mean prices comparable to off the shelf Intel or AMD CPU's *if* the Register hasn't massively misreported it and *if* they meet their goals... Those are of course two very big if's :)
And in two years time the FPGAs will be cheaper for the same features, or faster and larger for the same cost. If Natami has something performing like a >200MHz 68060 later this year already, I think we'll be okay for classic Amiga enhancements for a while :-)
Keep in mind they're looking at producing sub 65nm chips. Even at 65nm a good design could let you get to at least 2GHz+ (AMD's Phenom series for example reached at least 2.6GHz on 65nm). If this tech pans out and becomes as cheap as they are aiming for, it'll wipe the floor with FPGA's for anything where speed matters and people have the skills and time to do the designs. Though there's still the increased complexity of designing/verifying a design that can handle those speeds reliably, so whether it'd actually make a difference for projects like Natami really boils down to how good they are at chip design.
I for one would love to get a 2GHz+ m68k compatible CPU, but I'm not holding my breath quite yet :)
For now it's all speculation, but it'll be fun to see where it goes... If they succeed it'll be a *major* upset of the chip business, since so much of it now is centered around a dependency on this small number of multi-billion dollar fabs and a massive barrier of entry for small chip design companies. If that changes you're likely to see tons of new startups try their hand at designing everything from new CPU's to very specialized ASICs for specific purposes.
-
Keep in mind they're looking at producing sub 65nm chips. Even at 65nm a good design could let you get to at least 2GHz+ (AMD's Phenom series for example reached at least 2.6GHz on 65nm).
You can get to 2Ghz + at 130nm, which is what the first gen Athlon64's were, later dropping down to 90nm. The problem is not getting to the speed, but in how much power it takes to get you there. Which is one of the reasons why Intel using 45nm (and later 32nm) for their Core iX series have significantly lower power requirements than current AMD parts as well as better performance.
For now it's all speculation, but it'll be fun to see where it goes... If they succeed it'll be a *major* upset of the chip business, since so much of it now is centered around a dependency on this small number of multi-billion dollar fabs and a massive barrier of entry for small chip design companies. If that changes you're likely to see tons of new startups try their hand at designing everything from new CPU's to very specialized ASICs for specific purposes.
Could certainly set the cat among the pigeons. :D