KitGuru KitGuru.net – Tech News | Hardware News | Hardware Reviews | IOS | Mobile | Gaming | Graphics Cards
While the whole world plus dog is still looking at the ripples made by the Sandy Bridge recall, Intel is jogging on at a tremendous rate of knots. KitGuru dons a bee-keeper suit and enters the FAB-u-lous world of CPU technology.
In January, everyone's jaw dropped as Intel asked the market to send Sandy Bridge back. Now the company is not only looking at record revenues for 2011, it is also, very much, looking ahead. Far ahead.
Forget the Ivy Bridge launch and whether or not Intel can get the special Nikon lenses from tsumani-hit Japan in time for an ‘early 2012 launch', the scientists at Intel are focused (pun intended) on something far more ambitious.
The year is 2016, it's heading toward autumn and the chip technology that everyone's waiting for is the ‘tick' shrink of Pelican Lake – the architectural successor to Haswell (or at least that's the code name that seems to be going around the interwibble). It will increase performance in a handful of benchmarks, but – overall – the most significant change is the way that it allows Intel to squeeze even more chips from a wafer. Again, this will add additional price pressure on AMD (who will be hoping that failure rates increase exponentially as Intel attempts to add new processes to the manufacture of its processors).
Technologically, Pelican Lake is streets ahead of Sandy Bridge, with levels of parallelism that present designs can only dream of. Intel's engineers will also need to have solved some serious leakage issues – and consider running large areas of the chip asynchronously, because it's almost too complicated to try and maintain a single clock for the whole processor.
We're presently moving into the area of stacked memory and, pretty soon, we could be looking at stacked CPUs – designs which are inherently 3D in more ways than one.
Right now, in 2011, we're all marvelling at how much power we can achieve from a Core i7 2600k processor, costing just over £200, and clocking to 4.8Ghz on air. For moves like Intel's 7nm process change, the present quad core 2600k processors will be made to look like today's Atoms. We're moving that fast every 5 years.
KitGuru says: From today's talk of a system on a chip, the Pelican Lake era of processor will, effectively, be a network on a chip. When the expected move to multi-dimensional designs happens, it will be interesting to see if the clock speeds all fall back again, as they did with the first of the Core processors. As in life, it's not really how fast you work – as much as how much work you do – that's really important. Despite what the chip-marketing people might want you to believe.
Comments below or in the KitGuru forum.
I can’t wait to find out how they’re going to pull this one off. Transistor down-scaling is a lot more than just shrinking the length of the gate. Pretty much every part of the transistor has to be scaled at the same rate, or the performance increases just won’t be there.
Consider the fact that even with high-k gate dielectrics, which are currently 3 nm for the 32 nm process, a ratio of ~ 10:1. To maintain this ratio at 7 nm, the gate dielectric would be 0.7 nm, which is 5 or so atoms thick. Gate to channel tunneling at this distance would be through the roof, resulting in very high power consumption. And this is only one of hundreds of things they have to do!
Then again, the folks at Intel seem to be pretty good at getting around these rules time and time again.
Intel is not going to have DX11 until 2013, and Intel does not have any credible GPU high end technology/experience in-house and never had it either.
2016 is 5 years from now, and by that time CPU’s/GPU’s are going to be obsoleted. AMD already are releasing a 2nd generation native APU’s on 2013, which will make CPU’s and GPU’s obsoleted since the 2nd generation of APU’s are highly specialized processors that perform natively CPU and GPU functions.
AMD and NVidia/ARM have both credible CPU/GPU technologies while Intel only has CPU technology (obsolete) and basic GPU technology (DX10). Noting that Microsoft released DX11 in 2009 and that Intel still will not have DX11 technology until 2013 minimum. That tells any informed person that Intel is in big problems over the next 5 years and time will tell if they will make it. Money cannot buy success if you spends years of bullying competitors instead of improving your technology.
Remember Intel saying that GPU’s were dead.. What a vain mistake..!! More even if the future are APU’s design are dependent on high-performance GPU technology and Intel didn’t have and does not have credible GPU knowledge which is extremely necessary to build the new generation of APU’s…
How can the 2016 timeframe be correct? haswell comes in 2013 on 22nm, broadwell in 2014 on 14,15 or16nm, skylake later on 14,15 or 16nm, skymont later on 11nm, another 11nm should follow that and then perhaps a 7nm.
Not sure if this will format correctly, but this is the info we have from Intel.
Charlie & Co seem to have new names/numbers – which is always possible.
We’ll keep our collective ears to the ground and report back any updates
New Core Shrink New Core Shrink New Core Shrink
2011 2012 2013 2014 2015 2016
32nm 22nm 22nm 16nm 16nm 7nm
Sandy Ivy Haswell Rockwell Pelican Lake ?
Our source was very specific about the 7nm update, but there does seem to be a ‘hole’ around 10/11nm