Article "AMD CTO Mark Papermaster: More Cores Coming in the 'Era of a Slowed Moore's Law'" - @ Tom's

[UsandThem]

https://www.tomshardware.com/news/a...ores-coming-in-the-era-of-a-slowed-moores-law

Interesting read from someone we don't always hear from at AMD. It ought to be an interesting few years going forward, with more software making much better use of "MOAR cores!".

"In the near term, I don’t see a saturation point for cores. You have to be very thoughtful when you add cores because you don’t want to add it before the application can take advantage of it. As long as you keep that balance, I think we'll continue to see that trend".

"We're getting great supply from our partner TSMC." Papermaster said, "Like any new product, there is a long lead time for semiconductor manufacturing, so you have to guess where the consumers are going to want their products. Lisa [Su] talked about the demand simply being higher than we anticipated for our higher-performance and higher-ASP [products], the Ryzen 3900 series. We've now had time to adjust and get the orders in to accommodate that demand. That's just a natural process; in a way, it’s a good problem to have. It means the demand was even higher than we originally thought."

 

[Ottonomous]

"Intel, hurry up and leave 14nm"

 

[Nereus77]

I think 16 cores in mainstream CPU sockets is excellent and I hope we get more as the years progress. We can then see games and applications increase threads for greater speed/fidelity. The future is suddenly bright and exciting on that front.

I like how at one point it was normal to have 8 cores on your mobile's processor but 4 in your PC. Which CPU is more powerful, Intel? We were living in a backwards world without knowing it. AMD is shaking things up in the industry which is amazing. When Intel responds (in a few years, lol), we should be in for an amazingly progressive era in the processor space.

 

[soresu]

When Intel responds (in a few years, lol), we should be in for an amazingly progressive era in the processor space.

AMD won't let up in that time, they know they simply can't afford to when they're fighting on 2 fronts continuously.

AMD prices may shoot up and stagnate, but I don't believe the products themselves will.

 

[uzzi38]

AMD won't let up in that time, they know they simply can't afford to when they're fighting on 2 fronts continuously.

AMD prices may shoot up and stagnate, but I don't believe the products themselves will.

As Mark states in the interview, AMD's roadmaps (which actually exist for more a week at a time unlike someone else's ) are set in stone, regardless of what happens to that someone else they aren't going to hold back. Not sure why people might be doubting that, Zen 2 showed that perfectly, but Milan should remove any remaining doubts.

 

[moinmoin]

As Mark states in the interview, AMD's roadmaps (which actually exist for more a week at a time unlike someone else's ) are set in stone, regardless of what happens to that someone else they aren't going to hold back. Not sure why people might be doubting that, Zen 2 showed that perfectly, but Milan should remove any remaining doubts.

Indeed. And the focus on themselves is also a necessity since their core, silicon and architecture design improvements must hold up even in cases where advancements by process nodes are negligible or even fail to mature altogether like in Intel's 10nm case. AMD does very well with this approach so far so there is no reason to get complacent and change it.

 

[Nereus77]

Jim Keller's currently at Intel, and we all know what he's capable of doing. It means Intel's response might be very, very good in a few years. But AMD aren't letting up, so while AMD easily passed Intel who were basically standing still, AMD will still be rushing forward when Intel plays catch up.

 

[krumme]

Jim Keller's currently at Intel, and we all know what he's capable of doing. It means Intel's response might be very, very good in a few years. But AMD aren't letting up, so while AMD easily passed Intel who were basically standing still, AMD will still be rushing forward when Intel plays catch up.

It will take more time.
He left for Intel april 2018. I mean zen 2 is july 2018 what can he transfer so to speak that isnt already known?
Besides he cant move amd production setup to Intel. Amd will still share cost with appIe. I am not so sure he can make a sizeable difference.

 

[Topweasel]

It will take more time.
He left for Intel april 2018. I mean zen 2 is july 2018 what can he transfer so to speak that isnt already known?
Besides he cant move amd production setup to Intel. Amd will still share cost with appIe. I am not so sure he can make a sizeable difference.

People also forget his specialty. It's mostly in process improvement and organizational improvement. He can be and has been an idea guy. But he rarely has been a lead designer. I expect to a leaner, meaner, more on point development team from Intel. They need some one to kick them in the balls and remind them how to innovate. But they need the guys who will develop their uarchs to also be on point. The longer he is there expect some poaching from Apple/IBM/AMD to go on pretty heavily.

 

[uzzi38]

Jim Keller's currently at Intel, and we all know what he's capable of doing. It means Intel's response might be very, very good in a few years. But AMD aren't letting up, so while AMD easily passed Intel who were basically standing still, AMD will still be rushing forward when Intel plays catch up.

You know when Intel will be in a good spot - it'll be when Keller leaves

Jokes aside, Keller's working on Ocean Cove afaik, so 2022 stuff.

 

[moinmoin]

Keller is overrated insofar as many people now seem to expect a quick fix to Intel's situation. That ain't coming, at least not from him.

Let's just keep the timeline in mind. At AMD Keller's second stay was from August 2012 to September 2015. Arguably the first result was the Zen launch in March 2017. So he worked 3 years there, and the first fruits of his work were publicly released 1 1/2 further years later.

He joined Intel in April 2018. The pattern so far his that products containing his work launch after he left again. Also I'd call it very optimistic to expect such products to launch earlier than 4 years after he joined, so expect nothing before 2022.

it'll be when Keller leaves

*snickers*

 

[coercitiv]

Intel's problems are systemic, they won't get fixed through the surgical intervention of an engineering prodigy.

People who still expect "Conroe" or "Keller" moments from Intel still do not understand what put this company in the position it finds itself today.

 

[CentroX]

When zen architecture reaches 5ghz, intel will be in deep ****



Profanity is not allowed in the tech boards.

AT Moderator ElFenix

 

[Ajay]

When zen architecture reaches 5ghz, intel will be in deep ****



Profanity is not allowed in the tech boards.

AT Moderator ElFenix

Not happening. Read the interview.

As potent as TSMC's engineering team is, there's possibly a diminishing point of frequency returns, if not frequency declines, on the horizon as it moves to the smaller 5nm process. Papermaster is confident in AMD's ability to offset those challenges, though.

. - probably on background since it’s not a direct quote.

 

[lobz]

When zen architecture reaches 5ghz, intel will be in deep ****



Profanity is not allowed in the tech boards.

AT Moderator ElFenix

just say deep waters

 

[ehume]

just say deep waters

Or deep mud. Or deeply mired . . .

 

[Jimzz]

Not happening. Read the interview. . - probably on background since it’s not a direct quote.

If small nm production does lower frequency I could see them using a hybrid approach as well in the future. AMD did that before on older chips; use small tech in some areas and larger in other parts.

I think this is one of, many, reasons now holding intel back. Their 14nm is so matured that the new 10nm may not perform as well against it. Thats why they are only doing low power parts that don't need to scale as high.

 

[jpiniero]

I think this is one of, many, reasons now holding intel back. Their 14nm is so matured that the new 10nm may not perform as well against it. Thats why they are only doing low power parts that don't need to scale as high.

It's not a performance issue per se with Intel, it's a yield issue... which also indirectly hurts performance since Intel has to loosen the bin as far as possible so raising clocks means jacking up the voltage to ensure that all 10 nm chips that come out clean can be used.

 

[turtile]

Jim Keller's currently at Intel, and we all know what he's capable of doing. It means Intel's response might be very, very good in a few years. But AMD aren't letting up, so while AMD easily passed Intel who were basically standing still, AMD will still be rushing forward when Intel plays catch up.

It will really depend on how easy it was for him to reorganize the mess at Intel. He probably can't get as much done in a larger company in comparison to a smaller company like AMD. I wonder if he has a 3 year contract with Intel as well.

 

[RetroZombie]

I think this is one of, many, reasons now holding intel back. Their 14nm is so matured that the new 10nm may not perform as well against it. Thats why they are only doing low power parts that don't need to scale as high.

Good point, right now amd with TSMC 7nm have designs that peak at 4.7Ghz. Intel 10nm (TSMC 7nm equivalence) only does 4.1Ghz. Intel at 14nm peak at 5.1Ghz. AMD with GF14nm does 4.1Ghz max (here's Intel where AMD were in 2017), and AMD with GF12nm does 4.4Ghz.
Intel 10nm is an huge step back over the previous node, power hungry, yields poop, and performs miserably.

 

[RetroZombie]

People who still expect "Conroe" or "Keller" moments from Intel still do not understand what put this company in the position it finds itself today.

I agree Intel can't do the same trick, Conroe was based on an high performance mobile 35 Watt design, expanded and scaled has an 65 Watts desktop chip, and if that wasn't already enough they glued two together for the 130 Watts quad core monster.

Dr. Mark Papermaster worked at IBM for the PowerPC 4 (or 5) 'chiplet' approach, and Dr Lisa Su on the many cores Playstation 3 cell processor. You put the two together and you get the unstoppable more cores more chiplets approach

 

[lobz]

It will really depend on how easy it was for him to reorganize the mess at Intel. He probably can't get as much done in a larger company in comparison to a smaller company like AMD. I wonder if he has a 3 year contract with Intel as well.

Absolutely this. In a behemoth of a company you always run into oppositions from bureaucrats or just petty local strongmen from the most unexpected places.

 

[soresu]

I agree Intel can't do the same trick, Conroe was based on an high performance mobile 35 Watt design, expanded and scaled has an 65 Watts desktop chip, and if that wasn't already enough they glued two together for the 130 Watts quad core monster.

Dr. Mark Papermaster worked at IBM for the PowerPC 4 (or 5) 'chiplet' approach, and Dr Lisa Su on the many cores Playstation 3 cell processor. You put the two together and you get the unstoppable more cores more chiplets approach

Wait, Lisa Su worked on the Cell?

Huh, well either way they aren't strictly speaking 'cores' as you put it.

Cell was more like a single full POWER ISA based core coupled with 7-8 128 bit SIMD engines called SPE's - kinda like a very early proto APU, I'm pretty sure I saw an article claiming some games used the SPE's for extra graphics oomph.

If anything Lisa was the mother of the APU in that case.

 

[lobz]

If small nm production does lower frequency I could see them using a hybrid approach as well in the future. AMD did that before on older chips; use small tech in some areas and larger in other parts.

I think this is one of, many, reasons now holding intel back. Their 14nm is so matured that the new 10nm may not perform as well against it. Thats why they are only doing low power parts that don't need to scale as high.

That is utter bull excrement. They are only doing low power parts because they probably not making any profit even on those very small chips either.

 

[jpiniero]

Intel doesn't have a uarch problem, they have a fab problem. They need to decide - and quickly - if they are capable of keeping up with TSMC or even Samsung, and if not, they need to at the very least start dual sourcing.