Re: Intel Active Management Technology (AMT): not necessarily your friend

On 05/06/2017 02:32 AM, Rick Moen wrote:
> Quoting Ivan Sergio Borgonovo ([email protected]):
>
>> After
>> https://www.theregister.co.uk/2017/02/03/cisco_clock_component_may_fail/
>>
>> time to buy AMD, or POWER or RISC-V?
>
> [Article at TheReg fingers the dodgy part as an Intel Atom C2000 SoC
> that Cisco uses in sundry untis.]

> I sat on your query for a while to try to make sure I gave a thoughtful
> answer.  The question is, of course, what new pitfalls await after you
> flee more-familiar ones such as (say) Intel AMT or Intel SMM.

> The thing is, chips designed for the commodity marketplace are going to
> tend towards featuritis, because there's a big market reward for
> throwing in value-add funcionality like AMT.  Omitting them wouldn't
> shave much cost, so there they are.  For example, AMD has over the same
> timespan I've been writing about -- roughly this past decade -- built
> into _its_ competing chips a close equvalent 'remote management' feature
> called the Platform Security Processor, which an ARM core embedded in
> AMD's (recent) x86_64 CPUs.

> POWER and (IIRC) RISC-V don't give you anywhere near the same CPU grunt
> for the dollar that you get with either Intel or AMD's x86_64 designs.
> This is why, you might recall, Apple Computer (er., Apple, Inc., now)
> abandoned the PowerPC flavour of POWER for all of its desktop and server
> gear that runs MacOS X.  But if you don't mind having lower MIPS for the
> buck, sure you can go with one of the surviviing RISC flavours.

I'd say these kind of bugs are less frequent in AMD CPU compared to 
Intel but I haven't done any serious research and it may be due just to 
larger diffusion of Intel making them more scrutinized.

CPU and software market have changed, in part thanks to Linux.
POWER and RISC-V have a different licensing, RISC-V being more Free.

I think the problem in the MIPS/bucks ratio is mainly a problem of 
produced units and software support. x86 architecture is definitively 
not superior to alternatives.
The way software is developed in the open ecosystem and Linux made it 
much easier to support different architectures and mobile and embedded 
market provided the numbers.

You could shortly have a WinARM PC:

http://www.pcworld.com/article/3191401/computers/qualcomm-first-windows-10-arm-pc-coming-in-the-fourth-quarter.html

and yeah they failed before.

Few years ago if you had to think who was building the fastest smaller 
gates you could just think about Intel.
Now you've Samsung and TSMC at least.

It is much more common that silicon manufacturing is outsourced.

Google has put into production their specialized neural network CPU.

The fastest supercomputer is Chinese, running on Chinese designed RISC 
CPU. It's pretty impressive they were able to squeeze out record 
breaking performance from a unique architecture adapting Linux.
Of course building up a supercomputer for a bunch of specialized tasks 
is different than supporting a disparate generalist software ecosystem, 
but still impressive considering it is not just a supercomputer, but a 
record breaking supercomputer.

There is a market for media player/smart TV that requires fast CPU/GPU 
to play high resolution video.

There is growing interest for ARM in the datacenter

https://www.nextplatform.com/2017/05/05/red-hat-gatekeeper-arm-datacenter/

and when and if RH will decide the time has come you'll start to see 
patches for server related features and optimizations coming to the 
kernel and relevant software at a much faster speed.

I don't know about architecture advantage of j-core vs RISC-V but 
considering j-core is just on FPGA but you can buy real RISC-V and 
considering who's behind RISC-V I'd bet on RISC-V.


-- 
Ivan Sergio Borgonovo
http://www.webthatworks.it http://www.borgonovo.net