AMD Excavator on 28 nm, some production moving back to GF?

Page 5 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

norseamd

Lifer
Dec 13, 2013
13,990
180
106
how about argueing about what not to talk about we talk about something.

will the module based processors be extended or will they replace them with a new x86 cpu.

what would you guys like to see even if you think they will not make it
 

inf64

Diamond Member
Mar 11, 2011
3,764
4,223
136
how about argueing about what not to talk about we talk about something.

will the module based processors be extended or will they replace them with a new x86 cpu.

what would you guys like to see even if you think they will not make it
There is a new core under development (post EX) and its code name has nothing to do with construction equipment. It's relatively safe to assume it won't have Bulldozer module as its basis but anything is possible I guess.
 

norseamd

Lifer
Dec 13, 2013
13,990
180
106
There is a new core under development (post EX) and its code name has nothing to do with construction equipment. It's relatively safe to assume it won't have Bulldozer module as its basis but anything is possible I guess.

a big core?

are raja kodura and jim keller in any way involved?

any links or sources?
 

inf64

Diamond Member
Mar 11, 2011
3,764
4,223
136
Linkedin is a gold mine for information like that . It should be a big core but I have no idea if those two people were/are involved in the development process.
 

pTmdfx

Member
Feb 23, 2014
85
0
0
Linkedin is a gold mine for information like that . It should be a big core but I have no idea if those two people were/are involved in the development process.
Just to support your finding by recalling my search cache: there was a description of having recently worked on a new x86 core design (associated with a high-performance attribute) and a from-sketch low-power core with unspecified ISA. There is also AMDer sharing in my mother language, which the message is similar yet specified the existence of custom ARM design and ambiguous on the positioning of x86 core, which was said to not be named after Cat or BD families. So based on mainly this two pieces of information, I would bet this next iteration of CPU cores from AMD to be an x86 high-performance one plus an ARM power-efficient one, as I argued in some other forums. There are always other combinations of speculations available for people to pick, anyway, like GoodbyeX86AndHugARM, or ThingsRemainUnchanged.

If there is really a new core, I would guess "module" to be dropped. More precisely, there is no such a point to share an FPU, given low power cores are still fine with dedicated FPUs, and un-sharing may enjoy better latencies and lower design complexity either. But speaking of many-core scaling, part of the scaling challenge will be moved back to the inter- and intra-chip interconnect architecture, which is a weak spot of AMD today's in-market Opteron. A nice note is that there seems a new multi-core interconnect in AMD's oven which is said to be ambidextrous, and also intriguing research-patent pairs, which served as a base of following hetero-arch researches, from AMD Research in 2010/11 about a region-granularity directory protocol. So let's wait and see how it will turn out.

People looking forward to an immediate FinFET successor to Carrizo could be disappointed, anyway. Very high chance of non-skipping migration, I guess, as foundries are screaming for charging more for newer processes as reported. Perhaps we will see 20nm to back the low-end in parallel to FinFET processes in higher margin market segments?


P.S. Forget passwords doesn't help my old account, uhhhh.
P.S.2. I assume everything above is discoverable on Google with one's creative combinations of keywords, as I got all this pieces from the search engine either. But the warning is the searching costs you time...
 
Last edited:

R0H1T

Platinum Member
Jan 12, 2013
2,582
162
106
I think there is quite some information known about Broadwell, if you look at Wikipedia or all the news since September. We even know its die size (80mm² for Broadwell-Y). Not sure if you know such things about Qualcomm's 2015 products.
Oh you mean these ~
Snapdragon 610 & 615: Qualcomm Continues Down its 64-bit Warpath with 4/8-core Cortex A53 Designs
No wonder Intel's mobile plans are flailing cause they're promising lots of things & delivering little(if any) on time, not very unlike what AMD was accused off not too long ago !

As for GF & Excavator, like I've always said the software side of things needs a major overhaul plus if/when HSA does take off AMD could bring an 8 or a 6 core APU to the market, probably but surely not before 20nm from GF i.e. ~2yrs from now.
 
Last edited:

witeken

Diamond Member
Dec 25, 2013
3,899
193
106
Oh you mean these ~
Snapdragon 610 & 615: Qualcomm Continues Down its 64-bit Warpath with 4/8-core Cortex A53 Designs
No wonder Intel's mobile plans are flailing cause they're promising lots of things & delivering little(if any) on time, not very unlike what AMD was accused off not too long ago !

That article was actually posted after my comment . Anyway, I'm not impressed by the Snapdragon 610. It's actually a regression compared to the S600 and hardly any better than 410. So I'm not srue why you're accusing Intel of failing. If I remember correctly, Bay Trail was actually a 2014, high-end and Windows only product. I guess you're expecting too much. They're still on track to deliver a fourth graph like in this article. While Qualcomm and MediaTek are preparing their 8 core SoCs, Intel's dualcore and soon quadcore offerings on 22nm should easily defeat those, and after that real 14nm products.
 

norseamd

Lifer
Dec 13, 2013
13,990
180
106
Just to support your finding by recalling my search cache: there was a description of having recently worked on a new x86 core design (associated with a high-performance attribute) and a from-sketch low-power core with unspecified ISA. There is also AMDer sharing in my mother language, which the message is similar yet specified the existence of custom ARM design and ambiguous on the positioning of x86 core, which was said to not be named after Cat or BD families. So based on mainly this two pieces of information, I would bet this next iteration of CPU cores from AMD to be an x86 high-performance one plus an ARM power-efficient one, as I argued in some other forums. There are always other combinations of speculations available for people to pick, anyway, like GoodbyeX86AndHugARM, or ThingsRemainUnchanged. If there is really a new core, I would guess "module" to be dropped. More precisely, there is no such a point to share an FPU, given low power cores are still fine with dedicated FPUs, and un-sharing may enjoy better latencies and lower design complexity either. But speaking of many-core scaling, part of the scaling challenge will be moved back to the inter- and intra-chip interconnect architecture, which is a weak spot of AMD today's in-market Opteron. A nice note is that there seems a new multi-core interconnect in AMD's oven which is said to be ambidextrous, and also intriguing research-patent pairs, which served as a base of following hetero-arch researches, from AMD Research in 2010/11 about a region-granularity directory protocol. So let's wait and see how it will turn out. People looking forward to an immediate FinFET successor to Carrizo could be disappointed, anyway. Very high chance of non-skipping migration, I guess, as foundries are screaming for charging more for newer processes as reported. Perhaps we will see 20nm to back the low-end in parallel to FinFET processes in higher margin market segments? P.S. Forget passwords doesn't help my old account, uhhhh. P.S.2. I assume everything above is discoverable on Google with one's creative combinations of keywords, as I got all this pieces from the search engine either. But the warning is the searching costs you time...

think that a apu or soc with 4 full x86 cores 8 arm cores a gpu a audio processor with true audio and a dsp might be interesting
 

norseamd

Lifer
Dec 13, 2013
13,990
180
106
Why would such a Frankenstein be remotely interesting?

that is basically kaveri with a dsp and expanded audio processor and the arm cores.

would be interesting what 4 big cores and 8 small cores could run
 

Soulkeeper

Diamond Member
Nov 23, 2001
6,713
142
106
that is basically kaveri with a dsp and expanded audio processor and the arm cores.

would be interesting what 4 big cores and 8 small cores could run

I'd rather see an 8 core jaguar type APU with larger cache and clocks
 

norseamd

Lifer
Dec 13, 2013
13,990
180
106
And why would you want to mix x86 and ARM?

so you can run less intensive programs with less energy. and that you could also run easier tasks on more threads while having the big cores run a main thread for stuff like games.

not that it would be what i would want most. just that i thought it would be interesting what the apu would be like
 

pTmdfx

Member
Feb 23, 2014
85
0
0
think that a apu or soc with 4 full x86 cores 8 arm cores a gpu a audio processor with true audio and a dsp might be interesting
Why not simply ride on the voltage scale and gate off unused cores? I see a point of big little mix in the generic use pattern of handheld devices with a smart OS scheduler, but I doubt the practicality of such configurations in PC environment.
 

norseamd

Lifer
Dec 13, 2013
13,990
180
106
Why not simply ride on the voltage scale and gate off unused cores?

not enough of an expert to know about this.

not give unused cores any power?

I see a point of big little mix in the generic use pattern of handheld devices with a smart OS scheduler, but I doubt the practicality of such configurations in PC environment.

alright. so what exactly are vector processors?
 

pTmdfx

Member
Feb 23, 2014
85
0
0
not enough of an expert to know about this. not give unused cores any power?
Yes, and also with dynamic voltage and frequency scaling.

alright. so what exactly are vector processors?
It is just about adding a set of weaker, smaller cores when you have a set of power-scalable cores already. What benefits particularly would you get from such heterogeneous general-purpose CMP with the generic PC use pattern? As of my observation, I would define the PC use pattern as "all about multi-tasking responsiveness", say responsive browsers, office suites, productivity software, creativity tools, etc. You get it on when you need it, and you make it sleep what you are done with it. So honestly speaking, I don't see the necessity of such configurations to save power to the edge as in handheld devices. Nor many-core throughput computing would be the scope of PC.

Anyway, most of the cutting-edge SOCs in the market are neither with a heterogeneous general-purpose CMP config as far as I know. But they likely have some dedicated purpose cores for say always on connectivity or sensor detection, which may be modeled to be a worth of investments, yet this is beyond the scope here as it is about general-purpose CMP, aka the cores you can see in your task manager.

A note is that you need the cores to have an identical feature set in order to form an heterogeneous CMP. Otherwise, it would be a mess at least according to my knowledge (say switching a context with the use of AVX instructions to a Silvermont core). For a cliff-ish difference like the ISA, it should be impossible from day one as the architectures are fundamentally different in where the devil lives.
 
Last edited:

SiliconWars

Platinum Member
Dec 29, 2012
2,346
0
0
Someone over on SA forums saw this:

http://www.linkedin.com/in/kvnagesh?_mSplash=1

20NM Excavator??

October 2010 – November 2012 (2 years 2 months)
Manage India Test Plan and Infrastructure team of Steamroller(28nm) and Excavator (20nm) x86 CPU core processor.



Problem is that's well over a year ago. It's possible and likely that Excavator started off targeting 20nm but was dropped when they realised they wouldn't be able to hit the required clock speeds.
 

Dresdenboy

Golden Member
Jul 28, 2003
1,730
554
136
citavia.blog.de
BTW, I was reading an AMA with someone from Intel, and stumbled across this comment: [...]
Thanks for the link. I liked these answers to the question "How do you feel about AMD? (No really, let it out )":
Intel Engineer at reddit said:
They have fantastic people. I cannot underscore this enough, with the resources they have the fact that they're able to compete in the same ballpark we do shows their quality. Sadly for all of us, execution is key. We want to see an exciting marketplace as much as you do.
answering AMD_GPU_Designer said:
Thanks for this . We at AMD (especially on the GPU side) have a intense amount of respect for the engineers over at Intel. What Intel has done with their recent CPU architecture, along with the constant advances in fabrication technology, they deserve a lot of credit for "keeping the ball moving forward" in our industry.

To support jecb's argument, you often hear of negative press going on between the two companies, but that kind of animosity is largely isolated to the legal, marketing, and upper management levels. The engineers at most companies tend to have many good friends working for competitors, and while we might throw in a friendly jab every now and then, it's almost a universally friendly community.

Thanks for this AMA. It's always cool to hear what it's like on the blue team .
That's the reality I know. And then there are forums (like football stadiums) with certain strong opinions...
 

nenforcer

Golden Member
Aug 26, 2008
1,767
1
76
this would benefit with ddr4 right

Sure it would, but AMD doesn't have the kind of volume to introduce a new memory standard into the marketplace. They pretty much have to piggy back off of Intel.

While DDR4 will be coming for server / enthusiast Intel X99 chipsets this fall it will be in short supply and cost prohibitive. Most of the available quantities will be for large server enterprise customers, anyways.

Even Intel Broadwell mainstream is suppose to still be DDR3 in 2015 which is why AMD Carrizo will also be DDR3 (but quite possibly with a DDR4 memory controller built in.)

Your looking at 2016 before volume / bulk DDR4 is available with CPUs and chipsets to support it.
 
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |