• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Intel planning "Revolutionary" processor core.

Soldato
Joined
26 Aug 2004
Posts
5,097
Location
South Wales
Source
With AMD proving to be apt competition of late, Intel has apparently begun amping its design chops for the future, though AMD is not necessarily the sole reason of course. In a new job posting, the company is looking for an engineer to join its Hillsboro, Oregon team, building what the company simply calls its "next-generation core", or NGC for short.

Intel plans for this team to help "build a revolutionary microprocessor core to power the next decade of computing and create experiences we have yet to dream up", which sounds great to us. At the moment, the desktop / notebook processor landscape is interesting again, after a long drought of lopsided dominance by Intel. To realize that Intel planning for some revolutionary design is great to see, and likewise, it's good to know that AMD has its Zen 2 and future parts already well in the development stages. The next few years should prove very interesting on the microprocessor front.


Read more at https://hothardware.com/news/intel-...ionary-new-processor-core#wvv5HpsHqeBAAko4.99

This info is some days old and only a couple of sites are making mention of it so far, looks like it's a bit of a way off, so nothing much to talk about just yet i guess.

To quote motley fool
Intel's tired Core processor architecture should be put out to pasture within the next four to five years
 
on other news, processor company plans to keep making processors for the next 5 years....

to be honest we've known for a while Intel will be moving away from the core architecture soon, that's what tiger lake is supposed to be, 10nm+ with the design of the architecture being stripped down and sorted for the first time in a decade.

half of its issues are support for dozens of old instruction sets that aren't used these days, instead of removing/replacing Intel kept adding new ones (like avx2, now avx 512) *apparently * tiger lake is supposed to be a new architecture design (still usong the core basis) with most of this old junk removed.
 
on other news, processor company plans to keep making processors for the next 5 years....

to be honest we've known for a while Intel will be moving away from the core architecture soon, that's what tiger lake is supposed to be, 10nm+ with the design of the architecture being stripped down and sorted for the first time in a decade.

half of its issues are support for dozens of old instruction sets that aren't used these days, instead of removing/replacing Intel kept adding new ones (like avx2, now avx 512) *apparently * tiger lake is supposed to be a new architecture design (still usong the core basis) with most of this old junk removed.

Interesting, do you have sources for this? Links we can read? Or is this merely speculation?
 
4-5 years? Jesus that's a depressing thought.

Interesting, do you have sources for this? Links we can read? Or is this merely speculation?
There's this article from late last year that says it's rumoured Intel are planning a completely new design after Tigerlake. So basically we still have Coffee Lake, Cannonlake, Icelake and Tigerlake in the current line of incremental Core-based designs.

However, even if that were true, Intel may have rethought that since the release of Ryzen.
 
Last edited by a moderator:
I don't really see the Core architecture as being that old or tired :s it was light years ahead of the game when it was introduced and AMD are only really catching up now. It has however as mentioned above collected a fair amount of baggage along the way mostly for compatibility purposes however.

I think the major changes will be introducing changes to make mesh style setups more viable - especially some of the underlying changes in relation to bus and ring crossbars, etc.
 
Ice Lake is also a from the ground build up, new, tech. Likely not next gen though. The problem with the Core series is that it's almost hit a frequency wall and since many apps like high frequency, something needs to change (enter NGC).
 
if ice lake is on schedule for end of next year, and tiger lake in 2019, I guess we'll get some more info after tiger lake release?

I do think we'll see the end of x86 at some point soon though, maybe in the next 4-5 years, Intel wanted to get rid of x86 as far back as 2001 wanting to use a new instruction set as x86 is getting long in the tooth, and is part of all the 'baggage'

I wonder if ARM could take over desktops in the future?
 
Not sure if revolutionary is a word which inspires my confidence since their last truly revolutionary design was arguably Itanium / EPIC!

Revolutionary is always risky but at least Intel have deep pockets.
 
half of its issues are support for dozens of old instruction sets that aren't used these days, instead of removing/replacing Intel kept adding new ones (like avx2, now avx 512)
They don't anymore carry dedicated hardware for every old now abandoned instruction, but handle some of that old stuff with microcode.
Even if emulation always involves performance penalty that old stuff no one uses anymore would likely run better than during time it was used.
 
There's this article from late last year that says it's rumoured Intel are planning a completely new design after Tigerlake. So basically we still have Coffee Lake, Cannonlake, Icelake and Tigerlake in the current line of incremental Core-based designs.

However, even if that were true, Intel may have rethought that since the release of Ryzen.

No, i specifically want to read about Intel "stripping out old stuff" as i have not heard of this, and i guess once again its just speculation being peddled as fact by our new Intel supreme fanboy.
 
I do think we'll see the end of x86 at some point soon though, maybe in the next 4-5 years, Intel wanted to get rid of x86 as far back as 2001 wanting to use a new instruction set as x86 is getting long in the tooth, and is part of all the 'baggage'
"Itanic" was that supposed CPU revolution free from old legacy limits, first conquering servers and workstations and then spreading to desktops.
And when it was finally available it was very power hungry and didn't offer that much advantages over x86 CPUs which had kept evolving...

Then with AMD64 (with major involvement from Jim Keller) updating instruction set of x86 CPUs to modern era evolving x86 pushed into servers stomping Itanium under its feets.
http://www.pcworld.com/article/3196...d-to-replace-x86-in-pcs-hits-end-of-line.html
So much for all that hype 20 years ago...


Supercomputers have never had any use for anykind "Wintel" compatibility so if other architectures/ISAs had kept offering clear performance/efficiency advantages those would have been used.
But still x86 CPUs took absolutely dominating position also in them.
Especially after Core 2 updated performance and energy efficiency of Intel CPUs to at the time modern level.
https://en.wikipedia.org/wiki/Itanium#Competition


Something theoretically better but not available in reality isn't any good.
Especially if it can't run existing software.
Just like in nature things which keep evolving thrive best.

And similarly any other architecture/ISA would face problems in increasing single thread performance when running out clock speed potential of materials and transistor budgets stopped being limiting factor.
Only reason why ARM has increased its performance so much in this decade is forgetting its original power consumption limits.
Limit its TDP to that of ten years ago and I suspect it isn't closer to x86 in performance than it was then.
And thinking that it could efficiently emulate x86-64 to run existing software...
 
on other news, processor company plans to keep making processors for the next 5 years....

to be honest we've known for a while Intel will be moving away from the core architecture soon, that's what tiger lake is supposed to be, 10nm+ with the design of the architecture being stripped down and sorted for the first time in a decade.

half of its issues are support for dozens of old instruction sets that aren't used these days, instead of removing/replacing Intel kept adding new ones (like avx2, now avx 512) *apparently * tiger lake is supposed to be a new architecture design (still usong the core basis) with most of this old junk removed.


If that is the case that they are planning on dropping obsolete instructions and simplifying the decoder, they would have to allow enough time for the market to adapt without risking breaking the multitude of software complied to run on existing/recent x86. Perhaps the tail end of the core based Lake architectures could take some first steps in preparation for a new 'revolutionary' architecture rather than be its debut. They would need to set and implement a transition phase with widespread use of a new compiler to end up with a sufficiently large userbase with hardware that is able to run software that is no longer reliant on obsolete legacy instructions or any others that they might wish to remove. That sounds a long-term game, not something you could introduce overnight with a single CPU launch.

Reminded me of Agner's blog, old but interesting: http://www.agner.org/optimize/blog/read.php?i=25
and also: http://agner.org/optimize/blog/read.php?i=421#421 -most of its too technical for me to understand the implications of the suggestions but some interesting plainer bits at the end.
 
Last edited:
Wasn't they supposed to be using Graphene at some point to help CPU development? I'm sure IBM developed a CPU with graphene which showed promise.
 
Back
Top Bottom