
Back off, I'm hyper-nating!
Blending memory with the switch fabric. How novel. Looked at the paper: Hyperflex, hyper-retiming, hyper-pipelining, hyper-optimization, hyper-aware. I'm hyperventilating.
Intel's followed up on its acquisition of Altera by baking a microprocessor into a field-programmable gate array (FPGA). The Stratix 10 family is part of the company's push beyond its stagnating PC-and-servers homeland into emerging markets like high-performance computing and software-defined networking. Intel says the quad- …
Actually Xilinx did it; the Zynq-7000 originally consisted of two 32-bit ARM cores, an FPGA, and a whole mass of controller cores (e.g. Ethernet controllers, USB controllers, ...). Different models clocked the whole thing a different rates and had different size FPGAs. The newer MPSoc is similar but has up to 4 64-bit ARM cores, 2 32-bit "real-time" ARM cores, and an even bigger and faster FPGA.
I remember that somebody produced an FPGA coprocessor for AMD, but I can't remember who it was.
Yes, if you want to play with a FPGA and an ARM processor, then there are a number of little boards with a Xilinx 7010 on them, like the Parallella, which also sports a 16-core Epiphany RISC chip.
But really, doesn't it sort of seem that hell just turned down the temp a bit? Intel? ARM? +FPGA? It's rather nice to see Chipzilla experiment like this.
My guess would be not. With a tranche of really interesting network and server targeted ARM chips just beginning to appear, I expect this is the first shot of the Wintel symbiosis's hyper-protectionist hyper-move to hyper-punish those firms into hyper-bankruptcy.
A disruption-disruption operation.
The Cyclone V SE has been shipping with them for a couple of years now. Intel meanwhile have specialist Xeon parts with FPGA fabric on them.
I'd guess that the design work for this was underway pre the Intel buy-out so there's nothing particularly noteworthy about it having an ARM in there. I understand there are specialist Xeon models out there with FPGA fabric also. No real change yet.
Yes, I think the headline specs for Stratix 10 were already being advertised 2 years ago.
What will be interesting to see is whether future generations retain the ARM cores.
Personally I wouldn't mind if the big ARM cores give way to to powerful x86s - but I really like what Xilinx have done giving their MPSoC Cortex-R cores for hard real-time tasks.
This post has been deleted by its author
Umm, but it's not got an x86? (As well as the ARM), or did I miss something.
Isn't this mostly updated process technology for Altera FPGA? ARM cores in an FPGA have been around for a while. A good way to prototype an ARM SoC especially at the source can produce files for FPGA or ASIC (dedicated SoC)
'The Stratix 10 family brings to fruition Chipzilla's long-rumoured desire to put an x86 on an FPGA ARM core into a 14nm-process FPGA'
Might still be a vapourware desire, but it isn't reality here.
Nope, there is no x86 anywhere near this, or even a x86 instruction set that gets farmed or partially farmed out as ARM RISC Instructions. These just act as programmable intelligent controllers for a regular x86 processors (as one example use, hence Intel's interest).
All I can think of, is this is someway to bamboozle, getting readers to think Intel are still in the game regards IoTs, when ARM is leaps and bounds ahead.
As far as IoTs, Intel x86 Chips are dead in the water, hence the reason to ditch x86 processor development for mobile, pulling the rug from beneath Microsoft's feet as regards Windows 10 Mobile.
The only thing that could destroy ARM invitable progress/success is Softbank itself, or bamboozling marketing by a disruptive Intel, (or comments by Linus Torvalds*).
So yes - its not all plain sailing for ARM, given Intel's colossal marketing budget.
*said with tongue in cheek.
Though I don't think they gave up ARM licence and I think they did keep at least one communications controller with an ARM core.
Lots of PCs have ARM cores on Marvell Ethernet / SATA chips and on the hard drives. The profit margin is too low for Intel on ARM, also "Not Invented Here" syndrome.
They have a dilemma now with Altera. FPGAs are lot more actual profit per chip than a $1 ARM SoC, the main competitor of Altera is Xilinx which has a flagship ARM core FPGA. So Intel can hardly abandon FPGAs with ARM and no X86 (there is only a tiny market to have both cores, people will want one or the other and often later go to volume SoC ASIC unless it's a niche market.) FPGA have two distinct markets, stepping stone to volume SoC / ASIC and niche markets where there isn't volume, but the chip is volume product for foundry as the same chip can serve many different niche markets as well as development as it's user defined hardware.
There is still no X86 Intel Boilerplate Architecture in any shape or form involved in this particular product, assume the author is trying to say, well its a programmable 'do anything' chip, so you could build an X86 Processor using this Arm based FPGA as the foundation for that development, but thats not a serious proposition, obviously.
"Though I don't think they gave up ARM licence and I think they did keep at least one communications controller with an ARM core."
The management engine in Intel vPRO boards is ARM. They've used it for decades in controllers but that's all they've used it in.... up to now.
No, Intel developed the DEC's wonderful StrongARM further, somehow managing to end up with the lowest instructions per clock of any contemporary ARM architecture variant, with lots of needless register use barriers. Let's hope they just leave the ARM core the hell alone this time.