AMD Says Transistor Tech Will Keep Moore's Law Alive For 6 To 8 Years (theregister.com) 40
Chipmaker AMD has hinted that new transistor technology will keep Moore's Law alive for the next six to eight years, but as one might guess, it will cost more. From a report: Meanwhile, the company still plans to market new chips based on its Zen 4 architecture next year, including Bergamo, which is intended to compete against Arm-based chips for cloud-native computing. In an interview with Wells Fargo analyst Aaron Rakers at the financial outfit's TMT Summit, AMD CTO Mark Papermaster talked about future directions and the company's near-term roadmap. Rakers asked about the Zen family and its chiplet-based architecture versus the monolithic architecture seen with Intel's CPUs, and whether this would continue to serve AMD for the next four to five years, or whether another novel approach might be needed.
"Innovation always finds its way around barriers," Papermaster said. "I can see exciting new transistor technology for the next -- as far as you can really plot these things out -- about six to eight years, and it's very, very clear to me the advances that we're going to make to keep improving the transistor technology, but they're more expensive," he said. In the past, chipmakers like AMD and Intel could double the transistor density every 18 to 24 months and stay within the same cost envelope, but that is not the case anymore, Papermaster claimed. "So, we're going to have innovations in transistor technology. We're going to have more density. We're going to have lower power, but it's going to cost more. So how you put solutions together has to change," he said.
"Innovation always finds its way around barriers," Papermaster said. "I can see exciting new transistor technology for the next -- as far as you can really plot these things out -- about six to eight years, and it's very, very clear to me the advances that we're going to make to keep improving the transistor technology, but they're more expensive," he said. In the past, chipmakers like AMD and Intel could double the transistor density every 18 to 24 months and stay within the same cost envelope, but that is not the case anymore, Papermaster claimed. "So, we're going to have innovations in transistor technology. We're going to have more density. We're going to have lower power, but it's going to cost more. So how you put solutions together has to change," he said.
I also heard (Score:4, Funny)
That 2023 is going to be the year of the linux PC.
Lots of exciting things going on.
Re: (Score:1)
In AI-driven fusion-powered flying cars.
Re: (Score:2)
Re: (Score:1)
Yes, everyone knows that flying bot cars should be gasoline based.
Re: I also heard (Score:3)
Re: (Score:2)
I've been using Linux since 2019 as my only operating system. I manage Windows servers, from Linux. I game from Linux. I code in Linux. I edit videos in Linux. There is less and less reason to use anything else.
Well, yeah.
Re: I also heard (Score:1)
I had no choice; my Windows-centric technical background enabled me to realize what an untenable nightmare Microsoft products were.
PC is dying (Score:2)
The year of the Linux PC will never arrive. We'll have to be satisfied with being the dominate operating system in the world for phones, TVs, and servers. We will never have that final trophy of most popular desktop OS, even through Valve is trying really hard to make Linux gaming work.
Re: (Score:2)
Valve is trying really hard to make Linux gaming work.
Games which require kernel-based DRM aside, it's working pretty damned well. It lets machines that refuse to even run Windows 10 play DirectX12 games, for example. The excellence of Proton is almost enough to make me forgive Valve for how shit Steam is.
Re: (Score:2)
Agreed. It's not perfect but it's been night and day difference for gaming on Linux just a few years ago. I believe Proton has improved Wine as well, because my non-Steam stuff is working better now too. I play things like Elder Scrolls Online. Which I'm happy with in Wine (you can also get it on Steam), not perfect always perfect but also getting better all the time. No lag, few graphic glitches, audio works, etc. Hundreds of games in my library basically unlocked by those guys.
Re: (Score:2)
The year of the Linux PC will never arrive. We'll have to be satisfied with being the dominate operating system in the world for phones, TVs, and servers. We will never have that final trophy of most popular desktop OS, even through Valve is trying really hard to make Linux gaming work.
GrapheneOS and Purism? Shirley, you cannot mean Android.
Re: (Score:2)
Android runs the Linux kernel, the rest of it is not the GNU/Linux we all know but something a bit weird. I do have a fondness for Android's Binder, which never gained traction in the Linux community but it does offer some IPC that would have been a useful bases for some theoretical variant of D-Bus to build on top of. I think Pulse and SystemD would have been done on Binder if it were available as well, but that's just conjecture on my part.
Re: (Score:2)
The thing that infuriated me about Android when I looked at it was it was highly hardware-specific and getting official upgrades was stupidly non-existent. Samsung and Verizon were the worst offenders with locked-off one-off apps that replaced core functionality.
I'm in the Apple ecosystem for now, since it fits my "good enough with zero effort" agenda. Librem caught my eye right away --- it just lacks the hardware availability for my impulse-buy tendencies to grab it. Had several friends with Graphene te
More expensive (Score:4, Informative)
The complexity for minimum component costs has increased at a rate of roughly a factor of two per year. Certainly over the short term this rate can be expected to continue, if not to increase. Over the longer term, the rate of increase is a bit more uncertain . . .
In 1975 he revised it to doubling every 2 years.
Re: (Score:2)
Re: (Score:2)
Agreed, in fact the constant cost is almost the whole point. If I wanted a chip that's twice as fast but also twice as expensive I'd just buy two chips. That's not progress.
"Performance per watt" changes that math a bit. If "1 chip" gives X performance, you you have the options of "2 chips totaling 2X performance at 2X cost" or "1 chip totaling 2X performance at 2X cost but with 20% less energy usage" you'd still be ahead of the game (though not competitive with the traditional 2x performance at same cost every 18 months) to go with the single chip solution.
Re: (Score:2)
Sadly Moore's law hasn't had much relevance to single-threaded execution speed for many years now. Not that it has stalled entirely, but the doubling period is like 10 years.
https://www.man.com/single-cor... [man.com]
Re: (Score:2)
FYI: Moors law is about transistor density, not really about execution speed.
but the doubling period is like 10 years.
Because: it is not a law of nature, but was a simple observation of how industry is/was progressing at his time.
Re: Fusion energy (Score:1, Flamebait)
Re: (Score:2)
Thank you.
Re:Fusion energy (Score:5, Informative)
Re: Fusion energy (Score:1)
Re: (Score:2)
Re: (Score:2)
Moore's calculations were based on anticipations of density, but they are about transistor count. Kind of like how Murphy's law is believed to be about "if something can go wrong", when it's more about "if you can be confused about what Moore's law is"
Why is this coming from AMD? (Score:2)
AMD doesn't actually make chips, they design them. They split off their manufacturing arm into Global Foundries 13 years ago, and their most modern stuff is made by TSMC, which probably have a better idea of where the fab tech is going.
AFAIR, AMD is NOT a chipmaker (Score:5, Informative)
AMD is a chip designer, NOT a chipmaker.
As far as I remember, AMD used to be a chipmaker until 2009, when they spun off all their Chip Making capabilities as a company called GlobalFoundries. Granted, AMD probably has very good reationships with GloFo to this day... buuuuut.... GloFo announced in 2018 that they cancelled their 7nm process node and will remain at 12nm, so, even the indsider info from GloFo that AMD could get is not interesting/relevant for advaces for "the next 6 to 8 years".
AMD fabs on TSMC and, to a lesser extent, on Samsung, so they may get some info there, but no more, no less, than any other Chip designers.
If one wants to know what the future holds for moore's law, one shall not listen to chip designers like AMD, Qualcomm, Apple or Nvidia... There are only 3 reliable sources for the future of moore's law:
Intel, Samsung and TSMC
(in Alphabethical order)
Re: (Score:2)
AMD did sell their foundries, but they did not forget how they worked, and they did not suddenly stop having in-depth conversations with their suppliers about what technologies are impending. AMD committed to $1.6B in product from GloFo alone from 2022 to 2024.
There are only 3 reliable sources for the future of moore's law:
Intel, Samsung and TSMC
(in Alphabethical order)
Well, it's damned sure not in order of relevance. Let us know when Intel can make volume on their latest process, and also, when they start making sense. Remember when they said the future belonged to thousands of x86 cores in a package? Oh, how we la
Re: (Score:2)
AMD did sell their foundries, but they did not forget how they worked, and they did not suddenly stop having in-depth conversations with their suppliers about what technologies are impending. AMD committed to $1.6B in product from GloFo alone from 2022 to 2024.
What part of
Granted, AMD probably has very good reationships with GloFo to this day... buuuuut.... GloFo announced in 2018 that they cancelled their 7nm process node and will remain at 12nm, so, even the indsider info from GloFo that AMD could get is not interesting/relevant for advaces for "the next 6 to 8 years".
AMD fabs on TSMC and, to a lesser extent, on Samsung, so they may get some info there, but no more, no less, than any other Chip designers.
did you not understand? Because I think I wrote it clearly... Est-ce que tu avais besoin d'un dessine?
To reiterate: Yes AMD engineers still remember transistor tech and still talk with their GloFo colleagues, but that knowledge is stale, more than four years in the case of GloFo engineers, and more than 13 years stale in the case of AMD engineers, and not relevant to "the future of Moore's law in 2022 and beyond".
Also, yes, AMD talks to TSMC and Samsung (and to a lesser extent Intel) but that is
Re: (Score:1)
"but no more, no less, than any other Chip designers"
That is a bold and odd assumption. Why would you assume a foundry shares the same information wooing their largest client as they do with random chip designers?
Re: (Score:2)
"but no more, no less, than any other Chip designers"
That is a bold and odd assumption. Why would you assume a foundry shares the same information wooing their largest client as they do with random chip designers?
While you are technically correct (the best kind of correct), you probably missed a few lines later when I said:
If one wants to know what the future holds for moore's law, one shall not listen to chip designers like AMD, Qualcomm, Apple or Nvidia...
Yes, I 100% agree with you that the smaller designers will get less info on the (present and future) cutting edge process nodes than the heavyweight designers I listed...
But I'd go on a limb and say that AMD, Nvidia, Qualcomm and Apple get more or less the same level of info from the Intel/Samsung/TSMC teams...
Re: (Score:1)
Maybe so. Either way I agree with your conclusion that we shouldn't listen to AMD. This feels like cover for AMD wanting to use its new dominant position to hike prices.
A side effect of Moore's law... (Score:2)
Keeping Moore alive? (Score:3)
Re: (Score:1)
I find it odd that nobody is calling them out on the fact that while at one point they did keep them in the same price envelope the chips are already far more expensive than they used to be.
This is nothing but AMD preparing people for the fact they are going to start increased price gouging now that they are the market leader.
CEO quote translation (Score:2)
"it's very, very clear to me the advances that we're going to make to keep improving the transistor technology, but they're more expensive"
Translation: It'll be cheaper for us, but we'll charge you more. Call it More's law.
Re: (Score:1)
Almost every manufacturer is getting out of making the most cutting edge (smallest) chips because the fabs are getting ridiculously expensive.
The 2nm fab being planned by TSMC is projected to cost a mind blowing $34 billion.
BS (Score:2)
Nvidia said it's time to pay more for less.
Here is what Moore himself said: (Score:2)
Here is the link to what Gordon E. Moore himself said, published in Electronics, v38, #8, April 19, 1965, paper titled "Cramming more components onto integrated circuits".
https://newsroom.intel.com/wp-... [intel.com]
Here is the key statement, Moore's "Rule of Thumb" that has been designated "Law":
"The complexity for minimum component costs has increased at a rate of roughly a factor of two per year (see graph on next page). ...
Certainly over the short term this rate can be expected to continue, if not to increase. Over