Earlier this week we heard news about IBM's research into optical transceiver chips and today comes news at The Register of another interesting project to increase the frequency of processors by sheathing current copper interconnects in graphene. This is not the first time the usage of graphene has been investigated for computers, indeed there is research being conducted into improving non-volatile storage and even cooling with the use of graphene. The project being carried out by a team at Stanford University found graphene-coated interconnects can reliably carry data at speeds 4-17% faster than copper without the sheathing. They feel that a 30% improvement is reachable with current process technology; you can read more in the full article.
"Researchers have made tremendous advances on all of the other components in chips but recently, there hasn't been much progress on improving the performance of the wires," said Stanford electrical engineer Philip Wong."
Here is some more Tech News from around the web:
- HP launches its first USB Type-C Windows 10 tablets @ The Inquirer
- JavaScript creator Eich's latest project: KILL JAVASCRIPT @ The Register
- Feature-rich work in progress: Windows Mobile 10 build 10136 @ The Register
- Samsung spins up its latest rusty rotators for release @ The Register
Yes, all very nice, but…
Yes, all very nice, but… There is always one, isn’t it? 🙂
For consumers it’s currently no go. Intel, let alone AMD, won’t invest time and money now. They will try to squeeze ordinary silicone as much as they can before movement toward something really interesting begins. So even with graphene being really cool idea I wouldn’t hold my breath we will see this in the next 5 years from Intel. They have this 3-5% performance improvements strategy and they won’t change it. Times of performance jumps like between 486 and Pentium and Pentium 4 > C2D > Nehalem are gone and never coming back. And certainly they won’t cut into profit margin. Bloody corporations.
Let’s face simple fact. Intel now laying off many, many people (AMD doing that for different reasons) just to save like 300M USD annually. And most funny thing is that Intel as a company has value of like 150+ billion USD. Do you see any logic in this, because I don’t. That 300M for them is like me buying proverbial bag of chips.
Yes Intel is more interested
Yes Intel is more interested in market share, rather than focusing on the engineering that got some of that market share in the first place, monopoly business tactics included! Intel’s contra revenue shows that it as a corporation has digressed to using simple cost supports like a laundry soap manufacturer, rather than drastically improving the CPU/integrated GPU performance to any great extent, year to year. Once a company gets a monopoly market share expect any motivation short of a competitor breathing down their necks, to slowly milk technology for profits. Had Intel spent more of that contra revenue money on improvements to its CPU/GPU hardware, and software driver stack for its integrated graphics, people would be more enticed to upgrade. A Healthy AMD is necessary, and more representation of AMD’s future products need to be offered by laptop OEM’s, and HP needs to offer some upgraded/upgradable screen resolutions on its SKUs that are using AMD’s latest Carrizo based APUs, if not people should complain to there respective state attorney generals, as well as the federal attorney general, and FTC. The OEMs are a little to close to one CPU/SKU/SOC maker, or under the illegal monopolistic influences of one such corporation.
You can bet money that Intel
You can bet money that Intel are already well ahead for the future, as after 10nm, the need for new silicon replacements is abundantly obvious. I thought they had given up on graphene though, as testing was proving too many failures but hopefully this will come to fruition, as Graphene is a decent silicon replacement.
Actually, the place to look
Actually, the place to look for this tech will be HP. This is the exact tactic they are using for their “The Machine” architecture, using fiber optic to connect silicon to memory.
Lets not forget Intel’s heavy investment in Thunderbolt’s fiber optic. And this should also reduce prices and or increase speed for Internet companies.
The range of applications far exceed just Intel and AMD.
And lets not forget Nvidia, whom is already making their own interconnect as well as their own SoC, that could benefit from this.
And, technically, this would make HBM a speed bump if in terms of bandwith. Fewer lines will alway be less expensive for tech, and this could reduce those lines, increase bandwidth majorly or both.
No Nvidia just got the
No Nvidia just got the Coherent Accelerator Processor Interface(CAPI) from IBM and modified it, that big supercomputer contract with U-Sam with IBM’s power9’s and Nvidia providing the Future Volta microarchitecture based GPUs as accelerators. Hopefully U-Sam will be looking at some AMD Zen based APU HPC SKUs for some of its systems also, although both AMD and Intel will have to make their x86 products compete with IBM’s Power9’s in the future for server/HPC workloads. U-Sam does shedloads of big computing contracts and likes to spread the funding around a little to keep things up to date, so hopefully AMD can get some of that HPC/supercomputer action. Lots of technological developments that are funded by the government pay for a lot of the R&D that finds its way into the consumer products, especially in the computing areas. Those power9’s will be beasts of RISC based machines, and the power8 processor has 14 execution pipelines sitting behind 8 instruction decoders for some 8 processor threads per core of processing power, and the power9’s are incoming.
One wonders what Jim Keller’s K12 design team can do for AMD’s ARMv8a based custom microarchitecture should Keller design some fat/wide custom ARMv8a running microarchitecture with plenty of instruction decoders on the front end, and plenty of execution pipelines in the backside, and more than 2 processor threads per core. The Arm RISC ISA, like that of the power8/power RISC ISA, does afford more available die space for transistors in which to cram more execution resources per core in the same die space than the CISC designs. Once AMD gets Zen to market, and can focus on its custom K12 project hopefully AMD could field a custom ARMv8a based SKU that would give Apple’s A9’s some much needed competition, and give third party tablet OEMs an APU SKU that would allow them to compete with Apple’s tablet on a more even processing basis. Nvidia’s Denver cores are no where to be found, and have gone into hiding, or may just wind up in cars, so hopefully Nvidia will be reworking Denver to get it competitive with Apple’s custom silicon also.
Can AMD sit on 14nm for a
Can AMD sit on 14nm for a decade with graphene coating and still produce cheap APU’s that outperform intel…absolutely.