What’s in a nanometer? Intel may renumber its chips
The chipmaker is contemplating a change in how it how it numbers each generation of microprocessor.
Intel _does_ read Semiwiki!
- Thread starter hkwint
- Start date
Not surprisingly, Intel.com is the number one domain that hits SemiWiki.com.
P
Portland
Guest
Rocket lake being an underwhelming product is a sign of things to come. Intel was in trouble when apple teamed up with tsmc.
So Intel will no longer call it 7nm? Maybe stop calling it some nm feature size?What’s in a nanometer? Intel may renumber its chips
The chipmaker is contemplating a change in how it how it numbers each generation of microprocessor.www.oregonlive.com
P
Portland
Guest
Quarter projections don't look good.
I advised Intel to adopt Scotten's naming so they could lead the pack rather than follow TSMC and Samsung.
semiwiki.com
Much more accurate in regards to density, absolutely.
Intel Node Names - Semiwiki
There is a lot of interest right now in how Intel compares to the leading foundries and what the future may hold. Several years ago, I published several extremely popular articles converting processes from various companies to “Equivalent Nodes” (EN). Nodes were at one time based on actual...
semiwiki.com
Much more accurate in regards to density, absolutely.
P
Portland
Guest
Intel is in an existential crisis. Their problems are much more than nomenclature.
tooLongInEDA
Moderator
Does this really matter ? Intel lost a marketing battle against TSMC by appearing to be behind on the process labels. Big deal. Careless - yes. Embarrassing - slightly. Fatal - certainly not.
Does anyone who actually does <=7nm designs really care whether the process is called "10", "7" or "banana" ?
Those who need to know can easily benchmark a relevant IP block (recent ARM cpus being by far the most popular) in the different processes and find out the comparative density and speed (though not of course the yield which is also critical - and also security of supply and confidence in the foundry). This will tell you quite a bit more than the raw process comparison. But this sort of information is not public.
The only people I suggest are confused are those who don't understand this and don't do such designs - some Wall Street analysts (I memorably recall how 2000 tech boom analyst reports were described as "paid advertising"), SA commentators, etc. 95% of comment on this subject seems to be noise.
The fact that TSMC has the leading edge market tells you all you need to know.
Does anyone who actually does <=7nm designs really care whether the process is called "10", "7" or "banana" ?
Those who need to know can easily benchmark a relevant IP block (recent ARM cpus being by far the most popular) in the different processes and find out the comparative density and speed (though not of course the yield which is also critical - and also security of supply and confidence in the foundry). This will tell you quite a bit more than the raw process comparison. But this sort of information is not public.
The only people I suggest are confused are those who don't understand this and don't do such designs - some Wall Street analysts (I memorably recall how 2000 tech boom analyst reports were described as "paid advertising"), SA commentators, etc. 95% of comment on this subject seems to be noise.
The fact that TSMC has the leading edge market tells you all you need to know.
P
Portland
Guest
Intel may not exist in a decade if it does it will be a different corporation. People that aren't concerned are ignorant.
tooLongInEDA
Moderator
I thought this thread was actually about processing naming ...
That said, I really can't let the previous comment go.
"Intel may not exist in a decade if it does it will be a different corporation. People that aren't concerned are ignorant."
The industry is bigger than any individual company. Always has been. People used to say "what's good for GM is good for America". I don't seem to hear that as much now.
Any semiconductor industry company that doesn't change for a decade won't survive very long. The history of semis is constant innovation and change. Forty years ago Fairchild was still a player, NatSemi and Motorola were huge. They faded and got replaced by something better. That's progress. Companies have a lifecycle, just as their products do - some longer, some shorter (WorldCom, Nokia Mobile Phones).
Fairly certain Intel will still exist in a decade and will be different. I just don't see why "change" is a bad thing.
Will it be dominant ? Harder to say. If they produce great products in high margin sectors that people want, they have a chance. If they don't, someone else will have stepped up to fill the gap.
That said, I really can't let the previous comment go.
"Intel may not exist in a decade if it does it will be a different corporation. People that aren't concerned are ignorant."
The industry is bigger than any individual company. Always has been. People used to say "what's good for GM is good for America". I don't seem to hear that as much now.
Any semiconductor industry company that doesn't change for a decade won't survive very long. The history of semis is constant innovation and change. Forty years ago Fairchild was still a player, NatSemi and Motorola were huge. They faded and got replaced by something better. That's progress. Companies have a lifecycle, just as their products do - some longer, some shorter (WorldCom, Nokia Mobile Phones).
Fairly certain Intel will still exist in a decade and will be different. I just don't see why "change" is a bad thing.
Will it be dominant ? Harder to say. If they produce great products in high margin sectors that people want, they have a chance. If they don't, someone else will have stepped up to fill the gap.
P
Portland
Guest
After this quarterly earning and the next there will be a lot stronger comments. Intel is so stuck on 14 nm and tsmc process lead has become so far that designers have given up and are leaving. Intel is a broken company and we'll see if there is a will inside it to restructure and turn it around. What once was an existential threat has become an existential crisis.
hist78
Well-known member
It's my hope Intel won't turn any new naming scheme into a marketing campaign and then confuse themselves.
Not long ago, Intel adopted a strange terminology called "Contra Revenue" in perusing their market share in the mobile device industry. At the end it worked out badly.
Intel to hit 40 million mobile chip units with aid of subsidies | ZDNet
Intel is buying its way into the mobile market with subsidies to vendors, but that investment — on track to lose $4 billion in 2014 — is needed if it's going to threaten the ARM ecosystem.
www.zdnet.com
P
Portland
Guest
Narrative n**is are trying to push their narrative.
Is.anyone listening?
Is.anyone listening?
Exactly. It's the naive herd who care about the labels; and the naive herd is sooner or later influencing and feeding the blogosphere and this also reaches investors.
Samsung 5nm is a good example: I think Daniel already mentioned how it's way less advanced than TSMC N5 but Samsung feeds the press not to be too critical about it, and Scotten Jones calculated the TSMC equivalent node would be 6nm.
But Samsung is spending $100 billion next few years, and shareholders maybe wouldn't approve of that CapEx if TSMC was at 5 and Samsung 'lagging' at 6.
Groupthink and public image is important.
If you ask the US administration and the EU for tens of billions in subsidy, are you going to ask for help on a 4nm process? Or for P1278 or so?
Of course, designers don't care. If you're a car-fleet manager, you want to know about technical availability, fuel efficiency and TCO. No matter if BMW, Audi or Mazda labels a car 1, 3, 5, 7 or 8. People with ego's care, but the fleet manager looks at technical merits and specifications.
This reminds me of the 14nm / 16nm debacle. TSMC named their process 16nm based on the density being less than Intel 14nm. Samsung however chose 14nm even though their density was comparable to TSMCs. Then Apple used both TSMC and Samsung 14/16nm chips in their iPhone 6+ and people went crazy (Chipgate).
So yes, semiconductor insiders know the difference between processes thanks to SemiWiki and Scotten Jones but consumers do not and sometimes it does matter. So Intel is definitely selling themselves short and it has to stop. specially now that AMD is eating some of their lunch, absolutely.
it’s actually the opposite from what you’ve described, people preferred TSM’s 16nm over Samsung’s 14nm. people went crazy not because of the naming of the nodes, people went crazy after battery drain tests showing that iphone 6s with TSM chips last “2 extra hours”. An average Joe is not as dumb as you think.
P
Portland
Guest
Let's be honest Intel is already designing products to be made from tsmc. If you work with tsmc you don't have to be around the person that complains to human resources that you like to run alone if anyone at that campus jogs.
Actually it's not since I did not describe it in that way. The truth is that the SoC was designed for Samsung because TSMC 16nm was late. TSMC/Apple then ported the design to TSMC 16nm. The 16nm version was actually bigger since it was ported over but it was the same design. So tell me how a slightly larger die size chip consumes less power?
tooLongInEDA
Moderator
"So tell me how a slightly larger die size chip consumes less power?".
I think it's possible - depends on a number of factors. The libraries are almost certainly different - the 16 TSMC library might be lower power (or faster at the same power) than the Samsung 14nm used in some important details. Also possible that leakage might be different. It also depends what we mean by "ported". If this is a fresh synthesis and layout on the new process/libraries, then you're also at the mercy of the designers and tools - it may be the design implementation in one case was better than the other. Not saying any of these are the case - but such things are possible. 5-10% smaller die area might not always mean lower dynamic/leakage/total power.
I think it's possible - depends on a number of factors. The libraries are almost certainly different - the 16 TSMC library might be lower power (or faster at the same power) than the Samsung 14nm used in some important details. Also possible that leakage might be different. It also depends what we mean by "ported". If this is a fresh synthesis and layout on the new process/libraries, then you're also at the mercy of the designers and tools - it may be the design implementation in one case was better than the other. Not saying any of these are the case - but such things are possible. 5-10% smaller die area might not always mean lower dynamic/leakage/total power.