[Anandtech] Intel's Architecture Day 2018

[PeterScott]

Intel's Architecture Day 2018: The Future of Core, Intel GPUs, 10nm, and Hybrid x86

I know some bits of this were getting some mention in other threads, but it seems like it was big enough to warrant it's own discussion.

Highlights for me were a new architecture (finally) Sunny cove, that looks to have potential (4-5 wide allocation) to deliver Intels first real IPC improvement in MANY years, and also what should be a nice improvement in the the IGP, and a look at Intels multichip solutions(Foveros) and Intel Big-little.

The Q&A was also somewhat interesting, like where they admit being node-locked on their designs really messed them up, and they will be more node agnostic in the future.

Overall, more actual news than we have seen from Intel in some time.

 

[Nothingness]

We have a new method inside the company to decouple IP from the process technology.

That's hilarious. Fabless companies have been doing that for 20 years. It looks like Intel have discovered non custom design

Marketing speech aside, I guess they are less relying on custom flow now and that helps them switch from one process to another, but to reach their perf level they surely still have custom blocks that needs and will need porting.

 

[Arachnotronic]

That's hilarious. Fabless companies have been doing that for 20 years. It looks like Intel have discovered non custom design

Marketing speech aside, I guess they are less relying on custom flow now and that helps them switch from one process to another, but to reach their perf level they surely still have custom blocks that needs and will need porting.

Given Intel’s resources it would be irresponsible for them not to spend the $ to do ports/backup plans. Otherwise you end up with Skylake for far too long.

 

[Dayman1225]

That's not a demo, that's waving around a non-functional chip.

They had a demo system

 

[NTMBK]

They had a demo system

Ah, my bad in that case I was going off the Anandtech coverage, sounds like they didn't see that system.

 

[Yotsugi]

Ah, my bad in that case I was going off the Anandtech coverage, sounds like they didn't see that system.

That demo system was booting Linux, and that's it, I guess.
It's a Q3 2020 product at best so whatever.
At least it's physically alive.

 

[Charlie22911]

I can’t help but wonder when we will see 4 way SMT come to mainstream x86, there is already precedent with phi.

 

[LTC8K6]

They had a demo system

Well, they had a demo 10nm laptop in January of 2017...

 

[ozzy702]

I can’t help but wonder when we will see 4 way SMT come to mainstream x86, there is already precedent with phi.

Intel may have to if they want to keep up with AMD in the moar core war.

 

[Spartak]

Is there a quote where they literally say they will offer Sunny Cove on 14nm, that they have this at hand?

So far, all the info I saw on decoupling arch from process was on future tense, a promise of safer practice:


All this points towards future designs and what they're announcing here is a management/planning method, not a port of Sunny Cove to 14nm+*/-+

Is there a quote where they state sunny cove is not part of it? Didn't think so.
And since they were talking in present form, yes it was a literal statement unlike your assertion.

 

[TheGiant]

IMO bad move from Intel. They announced "we are making it safer rather improving themselves and delivering sooner than anyone expected".

Looks to me that first worse than expected performance makes the non change people in the company to win. I've heard last time that "innovation is disrespect to tradition"....

As investor it is a move to sell unless they announce new product where the capacities are used

 

[moinmoin]

Given Intel’s resources it would be irresponsible for them not to spend the $ to do ports/backup plans. Otherwise you end up with Skylake for far too long.

May I adapt that?
Given Intel’s resources it is irresponsible for them not to spend the $ to do ports/backup plans. Not doing so is how they ended up with Skylake for far too long.

 

[DrMrLordX]

Even Charlie Dem from SemiAccurate says that Icelake SP will be out in volume in Q3 2020 (intel says 2020 for Icelake Server just in case any one is wondering)

And that would pit it against Zen3. Of course, Intel was supposed to have Sapphire Rapids ready by then. But no, it's just IceLake?

o I saw on decoupling arch from process was on future ten

Not sure how your quote got mangled by multiquote (sorry) but I agree, I figured Intel meant this for . . . Willow Cove?

 

[wilds]

If Intel can maintain a sizable memory speed advantage over the competition like they did with DDR4, then it could be a winner. The IMC is something I am very curious about.

 

[Arachnotronic]

And that would pit it against Zen3. Of course, Intel was supposed to have Sapphire Rapids ready by then. But no, it's just IceLake?



Not sure how your quote got mangled by multiquote (sorry) but I agree, I figured Intel meant this for . . . Willow Cove?

Here's the latest roadmap that's leaked:

SPR looks like veeeerrry late 2020.

 

[jpiniero]

Here's the latest roadmap that's leaked:

That's from June. There's another leaked one that's more recent where Icelake Server is taken off (and it's just Cooper), and the "Future Xeon" is pushed into the beginning of 2021. Too lazy to look for it.

Any case I wouldn't say for sure the "Future Xeon" is Sapphire though. Could be Saphhire, could be Granite, could be rehashed Cooper.

 

[Nothingness]

Given Intel’s resources it would be irresponsible for them not to spend the $ to do ports/backup plans. Otherwise you end up with Skylake for far too long.

Yes indeed. I'm surprised it took them so long to do that. I guess process guys (including BK) were too proud to admit they wouldn't solve their issues in time.

 

[DrMrLordX]

Here's the latest roadmap that's leaked:

SPR looks like veeeerrry late 2020.

Hmmmm . . .

That's from June. There's another leaked one that's more recent where Icelake Server is taken off (and it's just Cooper), and the "Future Xeon" is pushed into the beginning of 2021. Too lazy to look for it.

Any case I wouldn't say for sure the "Future Xeon" is Sapphire though. Could be Saphhire, could be Granite, could be rehashed Cooper.

Hmm.

Honestly I'd rather see Sapphire Rapids sooner than later. Sunny Cove and Willow Cove don't interest me anywhere near as much as that project. Might be that Sapphire Rapids is being pushed out to let Keller go over it, if that's even what he's doing there (he may not be).

 

[Arachnotronic]

Hmmmm . . .



Hmm.

Honestly I'd rather see Sapphire Rapids sooner than later. Sunny Cove and Willow Cove don't interest me anywhere near as much as that project. Might be that Sapphire Rapids is being pushed out to let Keller go over it, if that's even what he's doing there (he may not be).

Sapphire Rapids is a server processor; Sunny/Willow Cove are CPU cores. They're not exactly comparable.

 

[Charlie22911]

With the increases in execution resources, is 4 way SMT possible?

 

[LTC8K6]

With the increases in execution resources, is 4 way SMT possible?

With the increasing security issues, is it desirable?

 

[DrMrLordX]

Sapphire Rapids is a server processor; Sunny/Willow Cove are CPU cores. They're not exactly comparable.

So? It's something new. What we see from Sapphire Rapids may set the tone for how Intel holds off AMD and the ARMy for years to come. Or it may amount to nothing.

I don't know how radical is the redesign effort behind Sapphire Rapids. Maybe it's as big as Itanium/IA64. Or maybe not. This is Intel's chance to break away from their legacy dating back to the Pentium III and make an enormous mark on the computing world . . . or do nothing in particularly special.

 

[Arachnotronic]

So? It's something new. What we see from Sapphire Rapids may set the tone for how Intel holds off AMD and the ARMy for years to come. Or it may amount to nothing.

I don't know how radical is the redesign effort behind Sapphire Rapids. Maybe it's as big as Itanium/IA64. Or maybe not. This is Intel's chance to break away from their legacy dating back to the Pentium III and make an enormous mark on the computing world . . . or do nothing in particularly special.

Sapphire Rapids contains one of the new cores Intel is working on (according to the CanardPC guy, it's actually Golden Cove based). But my point is that you are comparing a product/SoC level thing with individual cores. It's not apples-to-apples.

Put another way, if I said: "I'm not really interested in Zen, but I am looking forward to Rome," that's not really an apples-to-apples. The right comparisons would be, "I'm not really interested in Zen, but I'm looking forward to Zen 2" or "I'm not really interested in Naples, but I'm looking forward to Rome."

 

[jpiniero]

Sapphire Rapids contains one of the new cores Intel is working on (according to the CanardPC guy, it's actually Golden Cove based). But my point is that you are comparing a product/SoC level thing with individual cores. It's not apples-to-apples.

I guess the assumption was that Intel was splitting client and server core designs; and the Rapids were using a different core design that Intel didn't disclose. But if that's not the case, then that's really bad for Intel since it makes it all the more likely that they aren't going to ever get ahead of AMD/ARM, let alone far enough ahead to justify their current margin levels. And if they start falling behind they are almost going to have to give up on being an IDM since they wouldn't be able to afford the small nodes that are left.

 

[IntelUser2000]

9% A12 vs 15% ICL is just 'slightly' different from 30% A12 vs 5% ICL.

A12 is 30-40% faster in SpecCPU, which is far more representative of real world workloads than Geekbench, where its sub-tests take 3-5 seconds to run.

 

[Thala]

A12 is 30-40% faster in SpecCPU, which is far more representative of real world workloads than Geekbench, where its sub-tests take 3-5 seconds to run.

3-5 seconds is a long time for a CPU - i do not see any problem running a certain workload for a few seconds in order to get a representative result. Point in case the SpecCPU results are right in line with Geekbench or many other "real world" applications.