It isn't new that FPGA chips are being the favourite target of many companies focusing on Cloud Computing and High Performance Servers. Just consider, as example, the choice made by Microsoft to use FPGAs to enhance their own servers related to the search engine"Bing".

This happens because FPGAs are proving to be much more effective than the traditional x86, ARM and Power CPUs or GPUs in certain type of calculations. According to the Microsoft's study "Catapult": "Our study has shown that FPGAs can indeed be used to accelerate large-scale services robustly in the datacenter. We have demonstrated that a significant portion of a complex datacenter service can be efficiently mapped to FPGAs, by using a low-latency interconnect to support computations that must span multiple FPGAs".

The main disadvantage of FPGAs, however, consists in the need of a meticulous hand-made software compilation that needs time and money: "A major challenge in the long term is programmability. FPGA development still requires extensive hand-coding in RTL and manual tuning”. Therefore, for the days to come, FPGAs will be mainly focused on a restricted number of tasks; they will also be adopted in quantity thanks to an extensive use (just think to the number of daily searches on Google or Bing).

Concerning this scenario Intel could have the intention to acquire Altera in order to offer its customers a well rounded portfolio: it offers x86 CPUs (Xeon) and GPGPU accelerators (Xeon Phi), and it is missing FPGAs only (and Altera's high-end models are going to be based on Intel's 14nm FinFET process). Purchasing Altera, Intel might have the most comprehensive server based product portfolio, able to displace even IBM, which is trying to do the same with the OpenPower Foundation. Even Xilinx, Altera's biggest competitor, is proceeding the same way, producing Arm-Based FPGAs (Zynq series).

The big question mark on this maneuvre is the price. With a market capitalization of 10 billion dollars, Altera's outlay should be at least 3 billion dollars: Intel would have to spend something like 13 billion dollars, an astronomical amount if we consider that Intel's liquidity is reduced to approx 5 billion dollars. If the deal goes through, considering Altera's annual profit of 500 million dollars, the investment could take around 25 years to be repaid (at the actual rate of earnings and without any worsening condition).

The good thing is that economy is not based on maths only. With this acquisition, Intel could move all of the Altera's production to its foundries, abandoning TSMC. This would allow, finally, to exploit them at a good pace, given the fact that the x86 production is talking about 60% of capacity (while TSMC and Global Foundries are working, respectively, at 105% and 98%). And, Intel could finally offer the aforementioned product portfolio, allowing to sign contracts otherwise unattainable with the current offer.

In conclusion, the acquisition of Altera could be either a suicide or a brilliant move. We believe in the second option, considering the +6.38% of yesterday.

Autore: Marino Berre
Esperto in: Videogame, notebook, schede grafiche

Si è occupato di informatica ed hardware sin dalla più tenera età ed è stato membro di diversi user group tecnici e redattore per importanti portali italiani.

22 news	32 articoli

Questo indirizzo email è protetto dagli spambots. E' necessario abilitare JavaScript per vederlo.