SoC from ArtiCa Semiconductors & Apple?
Does anyone have any info on this company, and their relationship
with Apple. On wat machine while they use it if they do?
http://www.articasemi.com/
chk this out . any comments would be appreciated
with Apple. On wat machine while they use it if they do?
http://www.articasemi.com/
chk this out . any comments would be appreciated
Comments
Several really interesting features of the diagram are:
Support for 2 front side buses ( dual cpu ). But each cpu block in the diagram is marked as "Dual PPC". Although this could be a mistake, an optimist would take it as foreshadowing Apples introduction of dual core systems.
Support for DDR2.
Support for SATA 2 ( 3.0 gbps ), with hardware RAID.
32 PCI express channels.
This is really promising. It shows that Apple is working hard to keep up with the Jones', and that they are in fact aiming higher. It is also a single chip solution, so it should be much cheaper to produce then the current two chip setup.
Originally posted by mmmpie
Well, it certainly looks like an Apple style chip.
Especially when the top of the picture says Apple Macintosh
This seems like the replacement for the bridge chips in the next generation PowerMac. The dual core chips were rumored before and I would assume that this is the chip that Apple will use. Now the big question, when will it be ready?
Edit: Just thought of this. Hasn't there been a lot of desire to put the memory controller on the CPU. It looks like the next generation of 970 doesn't have it.
Originally posted by Programmer
Or this is just somebody's wet dream (i.e. they are either making it up completely or they are hoping to get their project funded). I certainly wouldn't put much stake in it -- I could have drawn a better picture myself in half an hour.
I agree, i seriously doubt the technical possibility of a chip which includes all the traditional southbridge I/O. I doubt any chip would have the bandwidth and imagine there problems with synchronization between memory controller, PCI, and I/O controllers.
Sounds good but yea someone's nocturnal emission.
(1) Why else would that Gary McMillian guy be an engineer, a PR dude and a webmaster at the same time?
Originally posted by Targon
I doubt any chip would have the bandwidth and imagine there problems with synchronization between memory controller, PCI, and I/O controllers.
Actually I don't doubt that part at all -- but its going to be made by one of the known existing big players, not by somebody noone has heard of. Today's GPUs are pushing far more bandwidth, and synchronizing between those devices isn't the troublesome part. SoC development has been happening for a couple of years now so they (those who have been doing it -- IBM, Motorola, Apple, nVidia, ATI, Intel, AMD, etc) are getting better at it. Much better.
Originally posted by Programmer
Actually I don't doubt that part at all -- but its going to be made by one of the known existing big players, not by somebody noone has heard of. Today's GPUs are pushing far more bandwidth, and synchronizing between those devices isn't the troublesome part. SoC development has been happening for a couple of years now so they (those who have been doing it -- IBM, Motorola, Apple, nVidia, ATI, Intel, AMD, etc) are getting better at it. Much better.
A GPU is a whole lot different to a sys controller. In addition this is a whole lot different to the Nvidia nForce4. Dude sync is a problem, Nvidia still have major problems with ATA I/O on their controllers. The Nvidia is basically a southbridge chip since it is employed for traditional southbridge I/O. It just happens to be placed on the board in place of the northbridge because as we know the Memory controller is on the the AMD CPU, obviously the nForce is taking on most of the traditional northbridge I/O thus the southbridge becomes redundant.
This 'Modular System Architecture' is basically a chip with ALL north/southbridge I/O in addition to housing the memory controller. There would need to some pretty spectacular engineering going on to sync up the mem controller with all the I/O on one chip. I can't imagine this becoming reality for at least 2-3 years. There is nothing out there on the desktop like this now.
Seems far fetched esp when we are talkin about PCI bridge chips in the mix... this could lead to some nasty limitations inside audio/video workstations where these pci latency tolerances are easily exposed.
We shall see what the future holds. im betting on the current Mac architecture being used further in the near future with the addition of PCI-e off the northbridge, perhaps DDR2 and Dual Core CPU support in the northbridge. I doubt very much we will see an entire SOC in a Mac for at least 2 years.
Originally posted by Targon
This 'Modular System Architecture' is basically a chip with ALL north/southbridge I/O in addition to housing the memory controller. There would need to some pretty spectacular engineering going on to sync up the mem controller with all the I/O on one chip. I can't imagine this becoming reality for at least 2-3 years. There is nothing out there on the desktop like this now.
:shrug:
This stuff is coming, and its coming faster than you think. I can't tell you any more than that.
http://www.broadcom.com/products/pro...uct_id=BCM1480
Originally posted by Targon
This 'Modular System Architecture' is basically a chip with ALL north/southbridge I/O in addition to housing the memory controller. There would need to some pretty spectacular engineering going on to sync up the mem controller with all the I/O on one chip. I can't imagine this becoming reality for at least 2-3 years. There is nothing out there on the desktop like this now.
Unlike Programmer, I have no insider knowledge to go from, but:
Remember, IBM is scaling technology down from high-end kit. Given the kind of engineering they've been able to pull off with their MCMs and support chips and very high-bandwidth busses on the POWER series, I'm not inclined to bet against them.
Which is not to say that none of the other companies Programmer mentioned couldn't do it, only that IBM's been at this sort of thing for so long that I can easily imagine them scaling it down, just as they did with the POWER4->PPC970, just as they did with Elastic Bus.
When HyperTransport first appeared on the horizon, I predicted that it would usher workstation-class architectures into the desktop realm. So far, I have not been disappointed in the accuracy of that prediction (which, for me, is pretty exceptional—I've called a lot of things wrong), and it looks like people are just getting warmed up.
The major problem was just getting bandwidth between parts on the board at a tolerable cost, with high reliability and manufacturability. We've got that now. The attendant issues are hardly easy to solve, but many of them have already been solved in workstations and high-end servers. The new bugbear of system design—especially IBM's—is latency.