AMD is closing the gap with its gigantic and eternal enemy, and from the Intel offices they have thought about that “if you can’t with your enemy, you better join him”, at least when it comes to making processors for laptops. And it is that although Intel has made the most of its dominant position in the market (with more than a fine by means of) to not let the excellent AMD APUs make too big a hole in it, it seems that they have decided to give their arm to twist and stop investing in their own high-end integrated graphics (and they didn’t do anything wrong with the i7-5775c and the i5-5675c) to collaborate with AMD and take advantage of its new graphics architecture by creating what will become known as Kaby Lake-G. It is true that in February we already heard the occasional rumor about a possible collaboration, but it was hard for us to believe that Intel would brand an “Apple -Samsung” and start creating SoCs with one of its arch enemies; now those rumors seem to gain more force.
Kaby Lake-G: AMD could make things difficult for Nvidia.
The future and now more credible Kaby Lake-G would be made with a 14 nm lithography and as you can deduce, they would include dedicated Polaris graphics (Or Vega, who knows) connected directly to the processor through a specific channel PCI Express X8 3.0 and using a new matrix interconnection technology called Intel EMIB or “Embedded Multi Die Interconnect Bridge”. But the thing would not stop here, but AMD could remove at a stroke one of the biggest limitations of its integrated GPUs, which is none other than the memory bandwidth, and for this the GPU would not use the RAM memory DDR4 system, but would incorporate its own HBM2 memory inside the package.
Although a priori this seems great, we find it difficult to believe how a memory that we could only see in the dedicated Fury graphics (given its high price) can make the leap to laptops without making the price of these Kaby Lake-G futures too expensive. Although yes, if we take into account the profit margins that Intel usually manages, the truth is that if there is someone with the potential to do this, it is them.
From what we know of the Intel Kaby Lake-G, these would have a size of 58.5 x 31 mm (The current Kaby Lake-S have dimensions of 37.5 x 37.5mm and 42x28mm Kaby Lake-H), which makes them larger than normal and narrow their scope of use to BGA solutions (soldered on board) created to compete, ironically, with AMD’s own APUs in those laptops that want to have a good level of graphic power without having to resort to a dedicated solution that increases the cost and consumption of the equipment.
You may be wondering why AMD would want to do that, but if we stop to think that this way AMD could make it very difficult for another of its biggest rivals, such as Nvidia, the truth is that the idea does not seem so far-fetched; Since Nvidia would see how all those laptops with dedicated low-end solutions such as the Geforce 920M / 940M and even the 960 / 965M could lose all their appeal overnight in front of more powerful SoCs that improve the autonomy of the equipment.
In short, what AMD could take over the entire low and mid-range laptop graphics market If you use the influence of Intel to get partners this time to mass-build their graphics solutions.
Other known data of future Kaby Lake-G would go through a TDP between 65 and 100W, which although much higher than the 45W that the Kaby Lake handles on the portable market (the i7-7920HQ being the most powerful), is still quite good if we think that we save the consumption and dissipation necessary to mount a dedicated GPU. It will be necessary to see if this news is fulfilled or not, but if so, we could see more than significant changes in a market such as that of laptops, which has been quite stagnant in recent times.
Intel also wants to improve on desktop and server processors with a revolutionary 10nm node.
Intel is not only putting all its efforts in the mobile market, but after several years where we have seen very poor CPI increases between lithographs (the Kaby Lake i5- 7600K performed the same or worse than the Skylake i5-6600K at the same frequency, since they share 14nm), it seems that it could finally be developing a revolutionary new 10nm node capable of producing much higher performance increases than we have been used to for a few years now.
It was on the “Tech Day” held a few days ago where the blues have spoken of a new process that although at 10nmIt would be much more advanced than the one that has been using the most direct competition such as Samsung, Qualcomm or TSMC. To do this, Intel would use a new third generation FinFET process with Hyper Scaling technology capable of producing much smaller transistors and at a much lower cost (up to 30% cheaper).
According to Intel, with this new process they could increase up to 2.7 times the density of transistors inside a chip compared to the current 14nm lithograph, thanks to a ‘minimum gate pitch’ that would drop from 70nm to 54nm and a “minimum metal pitch” that would also see its size reduced from 52nm to 36nm. In this way, Intel would achieve a density of logic transistors of 100.8 mega transistors per square mm, which would be around twice as high as the competitor’s 10nm processes.
Technological terms aside, and focusing on what matters to us, this new 10nm process would achieve a performance increase of up to 25% with a consumption that could drop up to 45%. Moreover, later we would see an improved review of this process called “10 ++” which could increase performance and decrease consumption by a few additional 15/30%.
We are undoubtedly facing improvements that sound so good that they do not cease to arouse some skepticism given the company’s most recent background, but that we will not stop following closely to see if Intel can detach itself a bit from an AMD that has done it. really well with Ryzen and which is finally afraid of him after many reigning with absolute tranquility in the market of domestic processors. We’ll be alert.
Sources: Benchlife, KitGuru