ARM and TSMC are moving ahead at an impressive pace, now predicting 10nm FinFET designs taping out possibly in the fourth quarter of 2015. That could even be possible considering how quickly they incorporated FinFET to move from 20nm SoC to 16nm. The the ARMv8-A processor architecture will have a few less transistors than a high end CPU which does help their process adoption move more quickly than AMD or Intel but with AMD partnering up with ARM there is the possibility of seeing this new ARM architecture in AMD chips in the not too distant future. As DigiTimes points out, there are many benefits that have come from this partnership between ARM and TSMC.
"ARM and Taiwan Semiconductor Manufacturing Company (TSMC) have announced a new multi-year agreement that will deliver ARMv8-A processor IP optimized for TSMC 10nm FinFET process technology. Because of the success in scaling from 20nm SoC to 16nm FinFET, ARM and TSMC have decided to collaborate again for 10FinFET."
Here is some more Tech News from around the web:
- Desktop, schmesktop: Microsoft reveals next WINDOWS SERVER @ The Register
- Apple updates malware definitions to protect OS X users from iWorm Botnet @ The Inquirer
- IBM goes gunning for Intel with Nvidia GPU-charged Power8 servers @ The Register?
- Android Wear can now boot Windows 95 @ The Inquirer
- A Look at Adobe’s Creative Cloud Fall 2014 Update @ Techgage
- Tech ARP 2014 Mega Giveaway Contest
I’m definitly no silicon
I’m definitly no silicon manufacturing expert, but it really seems to me like we are hitting a brick wall fast in terms of efficiency and total performance of silicon chips. I wonder if it’s wise to continue investing in smaller and smaller manufacturing process at this stage or if it would be wiser to push all research into alternative manufacturing processes and different computing platforms. Neurocomputing, different materials (carbon nanotubes, diamond, silicon carbide), multilayer chips, optical busses, etc.
Transistor counts have stopped doubling every 2 years since about 2008/2009. General CPU performance has gotten a little faster since then, but it has dramatically slowed. Efficiency has improved quite a bit, but I wonder if that’s from manufacturing process or improved architecture. GPU’s have still improved quite a bit, but they are inherently parrallel designs that will naturally improve with any density increase, even though clock speeds haven’t really improved.
A silicon atom has a diameter
A silicon atom has a diameter of ~ .2 nm. We are still dealing with only gate lengths of 50 to 100 times larger. I expect slow scaling until we reach a handful of atoms, followed by a push towards carbon which has half the diameter and stronger bonding. Multi-layer chips is already a thing in passive memory storage (Samsung SSDs) and will follow in active memory storage within a few years (HBM), then processors themselves.
GPU’s still have a lot more potential in clocks and architecture as they have long relied on density improvements.
Investments in materials will eventually be a requirement anyways, we might as well invest in the research while still proceeding down the usual path so we have an actual starting point when we hit a real wall.
10nm Q4 2015 or all of 2016
10nm Q4 2015 or all of 2016 for that matter from TSMC is a pipe dream. They’ve been having significant problems/delays since 28nm and it’s only going to get worse with each node.
Samsung on the other hand is close enough to Intel that they may be able to drop 10nm at the same time if not beat them @10nm/7nm.
Yeah, Intel and Samsung are
Yeah, Intel and Samsung are the only two (currently) with the R&D budgets to lead the charge.
Hay ARM Holdings, better get
Hay ARM Holdings, better get some better wider order ARMv8 reference designs out there, 3 IPC is not getting it, when Apple’s stuff has 6 IPC, and Nvidia stuff has 7+ IPC. Poor AMD is stuck using your reference A57, 3 IPC design, they better use twice as many cores, with some HBM. AMD where is your custom ARMv8 ISA running silicon, Nvidia will probably have the 2nd generation of Denver(whatever) on the market, before AMD gets its custom ARMv8 ISA to market. The tablet wars are being won with the custom ARMv8 ISA running cores, and better than Intel graphics, by Nvidia, etc., and AMD is way behind with a stopgap Amur. The RISC designs are besting x86 in low power usage, even though they are a few process nodes behind, but its the custom ARMv8 ISA running designs that will take the most tablet/phone market share, among the competing ARMv8 ISA based SOCs. ARM holdings you are in great need of some wider order superscalar reference designs, and the OpenPower licensed RISC design power8s are coming from the licensees beginning in 2015, and those not by IBM licensed Power* based products have all the RISC advantages of the ARM, MIPS, efficient RISC design, plus the power8’s have SMT, who knows how many Power8 derived SKUs there will be in the next few years, running servers, PC, Laptops, and mobile devices, once the licensees get hold of the powerful reference design of Power8, and slice it down into all sorts of SKUs. Arm holdings you better start, or have been looking at SMT, and wider order superscalar reference designs, as well as your IOT new stuff, as the server market will be more than just taking on x86’s CISC, it will be taking on licensed Power8 based RISC also.
Its amazing that the tech websites, even the Register, fail to mention the future licensed power8s, and will only discuss IBM’s made for IBM power8 SKUs, and costs, but Google is looking to license Power8, and have some custom power8’s fabricated, at only the licensing costs, and some engineering/integration/fab costs(A lot less cost than buying Xeon from the source, at a hefty markup). Wait until the Chinese and Koreans get their hands on a Power8 license, and probably the Japanese and even some of those Blighty based enterprises, and the Germans et all. 2015+ is going the be RISCtastic.
Note: Don’t even mention power8, and PowerPC in the same breath, they are not the same animal.