Intel Purchases FPGA Company Altera For $16.7 Billion

Toggle sidebar Toggle sidebar
Status
Not open for further replies.


Any properly run company will grow and purchase businesses that are in their interests, much like how Dell bought EMC who is the parent company to VMWare (a very highly used Virtual Machine software for servers) since they have a big hand in industry grade servers.
 


Desktops are LGA while most of their laptop CPUs are BGA. This is just them buying a company who is in a sector they are interested in.
 

FPGA in this context has nothing to do with sockets. It stands for "Field-Programmable Gate Array" which is a form of programmable logic often used for rapid-prototyping, signal processing, network routers/switches and other applications that require very fast, low-latency processing but either lack the production volume to justify a custom ASIC or where devices must be field-upgradable/customizable.

If you design a switch or router around an ASIC chipset, you have to make every future feature fit within the functions baked into your ASICs. If you use FPGAs, you can make the router or switch do whatever you want whenever you want, as long as the logic design still fits inside available FPGAs and connected resources.

FPGAs are at the boundary between software and hardware engineering.
 
Intel already has CPUs with FPGA built-in. Consider this, Core CPU with Stratix 3 in same chip. When Windows/Linux application needs fast encrypter it will just download "RTL" into FPGA and will run encryption in hardware at few GHz. Next minute (or if enough space in FPGA at the same time) will download decrypter and will also run it in hardware at some GHz.
Consider this in GPU where game can create its own shader!
And, of course, consider viruses using FPGA to hide themselves.
 

Good luck achieving "few GHz" in an FPGA, only the most basic constructs will achieve more than 1GHz. For crypto algorithms which use tons of 32bits multiplications, 500MHz is a more realistic clock target.

FPGA-based logic is not intended to break clock speed records - leave that to ASICs which don't have the massive interconnect overhead. Where FPGAs shine vs CPUs is their ability to run entire algorithms' pipeline concurrently in some series-parallel combination. FPGAs may only have ~1/10th the clock frequency of a CPU but they can work on dozens or hundreds of different things at the same time with execution pipelines tailored exactly for the problem at hand.

I do not think GPU shaders have much to fear from FPGAs: each of your RTL shaders would be several times the size of a single ASIC shader and run at 1/3-1/2 the frequency, which means your RTL shader would need to be 10-100X more efficient per clock than HLSL output to break even.
 


Problem is that their x86 license is non-transferable. A company cannot just buy AMD then pump out x86 CPUs.
 

They can buy AMD, run it as a fully own subsidiary and continue letting it pump out x86. No license transfer involved there. If the parent company wants to produce some x86-whatever hybrid with some of its other products, the AMD branch could provide a customized hardware macro for integration in those other products.
 


Actually, the license is also terminated in case either company changes ownership.

http://www.kitguru.net/components/cpu/anton-shilov/amd-clarifies-cross-license-with-intel-change-of-control-terminates-agreement-for-both/
 
AMD should terminate X64 license to Intel also in that case and we will have an industry stand-off ... i think courts should rule out that thing from the agreements between the companies. Intel at this moment is doing more harm than good to the IT industry.
 

They could, if AMD felt suicidal or decided to bail out of the x86 market.

If Zen fails to impress, AMD might be better off giving up on x86 and sell related IP to Intel to cover their debt payment obligations for 2017-2018.
 


What kind of harm? Last I checked they provide plenty of great features for IT like vPro (which allows for remote management like a Dell iDRAC) and ae pushing WiDi features.
 
If (if) AMD folded, their license would be transferred to someone. Intel wont be allowed to be the only x86 cpu manufacturer when the world literally runs on it.

Getting 64bit mainstream was hard enough, making ARM mainstream on the desktop will be VERY difficult making ARM producers a non issue for Intel.
 


Intel would not be. VIA still has one.

ARMs issue is they are low power which is good for mobile. But to truly compete with x86 they lose that low power advantage making it a hard sell. That and selling people on a new OS that does not work with their 10 year old software.
 

They could, if AMD felt suicidal or decided to bail out of the x86 market.

If Zen fails to impress, AMD might be better off giving up on x86 and sell related IP to Intel to cover their debt payment obligations for 2017-2018.
Click to expand...

I don't think Intel would be interested in any of AMD's recent x86 technology lol
 


AMD has patents that could cause a lot of problems for Intel. That's why Intel and AMD have a cross-licensing deal, each of them has tons of patents on stuff regularly used in modern CPUs. It was even AMD that came up with the x86-64 ISA (by extending Intel's x86 ISA to 64-bit).
 

Intel didn't want to make 64bits x86 chips so AMD created 64bits extensions, Microsoft and the Linux crowd flocked to it and Intel was forced to change their mind.

If AMD's x86-64 ISA and related patents fall in the wrong hands, it Intel might find itself unable to design, manufacture and sell x86-64 chips. That would be substantially worse than not having a second source.
 
Status
Not open for further replies.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom