News Intel announces new roadmap at IFS Direct Connect 2024: New 14A node, Clearwater Forest taped-in, five nodes in four years remains on track

Status
Not open for further replies.
I give Intel some credit. It sounds like they were forced to eat some humble pie and are trying again. Haven't seen this much activity from them since 2009.
 
Seems impressive
Would be interesting to see ARM made by intel and how intel will manage consumption/power ratio in comparison to risc processors
Even if it seems boundaries between cisc and risc are no longer what they used
 
Let me congratulate you, for being the very cutting edge of technology journalism. At least what a web-search tells me here, only Tom's Hardware and AnandTech reported in the past 7 hours about Intel's roadmap.

And nice. Intel is part of my pension savings plan, so double-glad to read the article.

Another part of that plan is AMD. Not mutually exclusive, in my opinion, as it currently looks for the overall semiconductors market to grow (leaving calculations aside here, about at which point it may become overvalued on the stock market). So, in such context, there is pie for everyone, albeit how much for whom, that seems to be seen.

Another part seems to be now Nvidia. I bought in half an hour before the stock exchange closed, and before their earnings call - and I've seen within two hours a peak of 11.01% on the bit I put in.

Disclaimer: I am not giving any investment advice. I am just sharing my thoughts on trying to put actively a bit of available money aside for way later, without gambling (too much). Here in Germany the pension savings plans on offer are somewhat meagre. And Tom's Hardware is a way better source of information about a sector, than analysts of banks are.
 
Has anyone looked at whether Intel 7 provides more density than their 10 nm SF node, whether Intel 3 provides better density over Intel 4, or whether Intel 18A provides better density over Intel 20A?

I have a suspicion that Intel 7 is really just Intel 10+++, Intel 3 would've previously been called Intel 4+, and that the Intel of old would've just said Intel 18A was Intel 20A+.

Regardless of what they call them, it's good to see Intel's process nodes improving. I just think what they accomplished is probably a little less miraculous than it sounds. Especially, considering how underwhelming Meteor Lake has been, performance-wise.
 
  • Like
Reactions: usertests
Intel was on fire back then. Conroe, Kentsfield, and Nehalem were all grand slams. If they work hard and do it right, we may see another Nehalem.
Nehalem wasn't nearly as big an improvement as either Sandybridge or Conroe. I actually agree that a reasonable target for them would be another Nehalem.

BTW, Intel-internal leaks of Arrow Lake performance suggest yet more underwhelming gains in single-threaded P-core performance (iso-power).
 
Has anyone looked at whether Intel 7 provides more density than their 10 nm SF node, whether Intel 3 provides better density over Intel 4, or whether Intel 18A provides better density over Intel 20A?

I have a suspicion that Intel 7 is really just Intel 10+++, Intel 3 would've previously been called Intel 4+, and that the Intel of old would've just said Intel 18A was Intel 20A+.

Regardless of what they call them, it's good to see Intel's process nodes improving. I just think what they accomplished is probably a little less miraculous than it sounds. Especially, considering how underwhelming Meteor Lake has been, performance-wise.
Both 10nm SF and Intel 7 has larger pitches than predecessors, so theoretical maximum density goes down. However, they aren't taking full advantage of it due to their obsession of clocks, so in reality there's no difference.

While on paper Intel 3 is a revision of Intel 4, in reality Intel 3 brings in HD libraries, increase in density for HP libraries, and other libraries that allow a full SoC, while Intel 4 is basically only for CPU. Same is true with 20A vs 18A. 18A is the real, full node while 20A exists purely for bringing 18A reliably.

Meteorlake has been plagued with delays, which is part of the disappointment. Also it looks like Intel 4 might be at fault too because again, it's a stepping stone for Intel 3, which will be utilized fully. Same should be with 20A.

It isn't completely correct to call 10nm SF, Intel 7, and 3 just mere plusses either. They bring a rather big gain in transistor performance. The 14nm plusses only brought 3-5% gains, while 10nm SF alone brought 4x 14nm plus worth of gains.

Intel_explains_that_the_latest_advance_in_Intel%E2%80%99s_transistor_technology_enables%C2%A0the_largest_intranode_performance_improvement.jpg


Intel 3 brings a 18% transistor performance gain, so again it's significant. Intel 4 brings 20%, so Intel 3 in terms of performance is practically a new node. Intel 7 brought 10-15% gains.

https://www.anandtech.com/show/1682...nm-3nm-20a-18a-packaging-foundry-emib-foveros

(That's also ignoring the point I made above, that Intel 4/20A seriously lacks libraries to make anything other than CPU tile, no memory controllers, IO, or SoC)
 
Last edited:
  • Like
Reactions: bit_user
I have trouble concentrating on the content.
I'm just baffled that in the year 2023, that Intel still hasn't discovered Unicode. ... or even Windows-1252.

The cross-multiplication symbol is not a letter but looks like this: ×
The lines are orthogonal, there are no serifs and the symbol is raised slightly above the baseline.

The unit ångström (1 ångström = 0.1 nm) has as its symbol the Swedish letter Å. For example: "14Å".
The letter A is the symbol for the unit ampere (current).
 
  • Like
Reactions: bit_user
I have trouble concentrating on the content.
I'm just baffled that in the year 2023, that Intel still hasn't discovered Unicode. ... or even Windows-1252.

The cross-multiplication symbol is not a letter but looks like this: ×
The lines are orthogonal, there are no serifs and the symbol is raised slightly above the baseline.

The unit ångström (1 ångström = 0.1 nm) has as its symbol the Swedish letter Å. For example: "14Å".
The letter A is the symbol for the unit ampere (current).
These slides (and, indeed, I'm sure their whole naming scheme) were made by the marketing department - not engineering.
 
These slides (and, indeed, I'm sure their whole naming scheme) were made by the marketing department - not engineering.

i agree
Since tenth generation, intel products are good to be honest.
The most important thing is competition and choice for consumers
i still remember playing decently gran turismo in a playstation running on MIPS 180nm process The amd jaguar family is a 28 nm
you can still have a decent running Macintel with kaby lake on 14nm

however, node are not everything, the gain is smaller and the costs are higher, progress in material (photonic) and designs are important too.
One day, someone will bring out photonic computers and we'll all bring back our old metallica "nothing else matters" remembering 0.5 nm lithography costly race
those are maybe be engineering department roadmaps 😉


A 90 nm cmos photonic chip has been tested by IBM and it makes the race to lithography less relevant. It's 10 times faster (or way more) and a lot less energy consuming
They still of course need to overcome bottlenecks and make it ready for the market and for general purpose computers but it depends on what research instituts have in store.
 
  • Like
Reactions: t3t4
Status
Not open for further replies.