You can login if you already have an account or register by clicking the button below.
Registering is free and all you need is a username and password. We never ask you for your e-mail.
Intel is so diversified as a company they won't have any trouble, they'll just try to save face until their answer is ready to be shown off but we can consider their style of cpu design a living fossil. Their current i core design philosophy is going to die off, were going to see them start talking about a system on silicon or something to that effect which will be a collection of separate components built on separate fabrication nodes which best facilitate the role of those components, they can make parts on 10nm they just can't make very big parts so the answer is simple, make a lot of smaller parts that work and assemble them into a bigger product.
The problem with 10nm for Intel is one of building a massive chip on a process they can't completely control, theres just too many defects per square mm to make enough working chips with their current big designs. All they can make on it reliably right now and sell for a profit is very small dual core i3's, they just don't have the chip designs which are capable of being glued together because the intention was to keep everything monolithic for another couple generations and they aren't going to publicly admit AMD got the drop on them again.
Theres a few Intel slides showing off their plans to integrate multiple separate dies made on separate lithography processes into one package that can be customized as needed. They experimented with this and the AMD vega integrated GPU thing no one bought, it will be that sort of system but customizable. If a customer wants lets say 16 cores and a basic gpu they'll get four quad core cpu modules and a basic gpu on a kind of carrier which links everything together efficiently, the memory controller will probably be a part of the carrier. If another customer wants an ultra power efficient low end part it will be just one of the cpu modules on the carrier, if a customer wants a 64 core part and no gpu than they'll be able to get a carrier with 16 cpu core modules. Want something with 16 cores and a honking great big FPGA beside it? Intel's got you hooked up.
I've got one on my desk atm, Terasic DE10 Nano dev board. Intel has x86 CPU / FPGA combos for the data center, the only reason they have ARM based FPGA SOC's is Intel don't really have the kind of low end and low power hardware that would fit into the power limits for lower end chips, that and Altera was designing these before Intel bought them. Like if you want parity with ARM from Intel you have to go with Atom or Quark and they are still more power hungry than ARM while being only about as fast. As for data centers they don't mind using crazy amounts of power as long as the job is done quicker so it suits Intel to offer you ARM FPGA soc's and data center X86 FPGA combos.
view the rest of the comments →
[–] Gamio 0 points 3 points 3 points (+3|-0) ago
Intel is so diversified as a company they won't have any trouble, they'll just try to save face until their answer is ready to be shown off but we can consider their style of cpu design a living fossil. Their current i core design philosophy is going to die off, were going to see them start talking about a system on silicon or something to that effect which will be a collection of separate components built on separate fabrication nodes which best facilitate the role of those components, they can make parts on 10nm they just can't make very big parts so the answer is simple, make a lot of smaller parts that work and assemble them into a bigger product.
The problem with 10nm for Intel is one of building a massive chip on a process they can't completely control, theres just too many defects per square mm to make enough working chips with their current big designs. All they can make on it reliably right now and sell for a profit is very small dual core i3's, they just don't have the chip designs which are capable of being glued together because the intention was to keep everything monolithic for another couple generations and they aren't going to publicly admit AMD got the drop on them again.
Theres a few Intel slides showing off their plans to integrate multiple separate dies made on separate lithography processes into one package that can be customized as needed. They experimented with this and the AMD vega integrated GPU thing no one bought, it will be that sort of system but customizable. If a customer wants lets say 16 cores and a basic gpu they'll get four quad core cpu modules and a basic gpu on a kind of carrier which links everything together efficiently, the memory controller will probably be a part of the carrier. If another customer wants an ultra power efficient low end part it will be just one of the cpu modules on the carrier, if a customer wants a 64 core part and no gpu than they'll be able to get a carrier with 16 cpu core modules. Want something with 16 cores and a honking great big FPGA beside it? Intel's got you hooked up.
[–] Dfens 0 points 2 points 2 points (+2|-0) ago
You can get an FPGA processor combo now. Typically the processor is ARM rather than Intel.
[–] Gamio 0 points 1 point 1 point (+1|-0) ago
I've got one on my desk atm, Terasic DE10 Nano dev board. Intel has x86 CPU / FPGA combos for the data center, the only reason they have ARM based FPGA SOC's is Intel don't really have the kind of low end and low power hardware that would fit into the power limits for lower end chips, that and Altera was designing these before Intel bought them. Like if you want parity with ARM from Intel you have to go with Atom or Quark and they are still more power hungry than ARM while being only about as fast. As for data centers they don't mind using crazy amounts of power as long as the job is done quicker so it suits Intel to offer you ARM FPGA soc's and data center X86 FPGA combos.