Intel 7nm ambitions are...lofty

N

Nightfire

2[H]4U
Joined
Sep 7, 2017
Messages
3,279
This one might cause you to snort your drink out of you nose:
https://nl.hardware.info/nieuws/66737/intel-ceo-bevestigt-plannen-voor-7nm-cpus-in-2021
064371_intel-2019-investor-meeting-process-tech-7-nm.jpg


Not really sure what 1272 or 1274 is. In any case, it seems Intel has not learned their lesson on keeping goals and expectations reasonable unless 10nm has really turned out to be a sort of 7nm-- and 7nm-.
 
The crazy part to me is that AMD has 7nm going already while Intel apparently won't have it ready for 2.5 years according to this chart. AMD is going to have time to refine the hell out of Ryzen by the time Intel even gets on the smaller node.
 
bpizzle1 said:
The crazy part to me is that AMD has 7nm going already while Intel apparently won't have it ready for 2.5 years according to this chart. AMD is going to have time to refine the hell out of Ryzen by the time Intel even gets on the smaller node.
Click to expand...

Do they really care though? They still hold an IPC advantage and one would think once Intel moves to 7nm they would have that plus a new architecture with more than just some 2-3% IPC increase.

AMD has dramatically closed the gap and will win back market share but Intel will swing back before they need to seriously worry about losing major market share.

Unless there is a huge switch to EPYC that is. AMD still has to get their laptop chips out and compete.

If Ryzen 2 can beat cheap Intel chips in laptops while remaining power efficient AND win over the server/data then Intel needs to worry.
 
In fairness, they did go a bit overboard with the dimensions of 10nm. They could of called it 7nm by other standards.

I just don't see them going through this many node changes in the next 5 years and we know how perf/w of 14nm++ stacks up against 10nm given the same architecture.
 
Is SunnyCove's desktop release still on schedule? Otherwise, nothing really worth paying attention to.
 
next-Jin said:
Do they really care though? They still hold an IPC advantage and one would think once Intel moves to 7nm they would have that plus a new architecture with more than just some 2-3% IPC increase.
Click to expand...

Just like AMD, Intel will have to take one step back (clock speed) in order to take 2 steps forward (efficiency and IPC).
 
Intel keeps insisting that its process node is more advanced than the competition...it's a way of saying that its 10nm node is roughly on par with AMD's 7nm node...
 
next-Jin said:
Do they really care though? They still hold an IPC advantage and one would think once Intel moves to 7nm they would have that plus a new architecture with more than just some 2-3% IPC increase.

AMD has dramatically closed the gap and will win back market share but Intel will swing back before they need to seriously worry about losing major market share.

Unless there is a huge switch to EPYC that is. AMD still has to get their laptop chips out and compete.

If Ryzen 2 can beat cheap Intel chips in laptops while remaining power efficient AND win over the server/data then Intel needs to worry.
Click to expand...
Actually, AMD has the IPC crown now. The only reason Intel is appears to have that advantage, is core speed, but clock for clock, AMD is faster IPC wise. Intel is going to lose some of it's core speed with the die shrink. How much is the million dollar question.
 
I swore I read a long rant from Charlie about that question in June. His speculation was that the IPC uplift wouldn't be enough.
 
polonyc2 said:
Intel keeps insisting that its process node is more advanced than the competition...it's a way of saying that its 10nm node is roughly on par with AMD's 7nm node...
Click to expand...

well 10nm was suppose to come out three years ago, who knows what they have been up to.
better not come out this holiday, i just upgraded.
 
NWRMidnight said:
Actually, AMD has the IPC crown now. The only reason Intel is appears to have that advantage, is core speed, but clock for clock, AMD is faster IPC wise. Intel is going to lose some of it's core speed with the die shrink. How much is the million dollar question.
Click to expand...
The 10nm laptop part was 30% or more down over its 14nm bretheren. Although it was more efficient in one or two scenarios it was not as fast in most common, shorter peak loading tests due to clocks.

This chart is horseshit with an unlabeled Y axis though..
 
I don't know not that long ago Intels head engineer was talking about how they basically Fd 10nm up and they had to toss a ton of work, that will just never really go.

At the same time they where talking about 3D stacked chiplets. Now I know they are planning to bring lakefield later this year as their first 3D stacked chiplet SOC design to compete against Qcom. Its possible that is all he was referencing.

Still it makes me wonder how aggressive/experimental they will get with their Foveros 3D stacking stuff.

Intel is going to swing back... as I see it one of three things is going to happen early next year.
1) Intel is going to release a chip that is basically just their current stuff shrunk to 10nm but working. Gain some IPC drop some clock speed and outside of a golden snitch super part to win all the benchmarks their product line will pretty much be a wash.
2) They swing for the fences with a Foveros 3D stacked chiplet design, and pretty much set AMD right on their ass.
3) They swing for the fences with a Foveros 3D stacked chiplet design, and fall flat on their face.

Their engineers saying there eating crow and dumping years of work doesn't lead me to believe the first scenario is all that likely. So I believe its going to be an interesting 2020. I am really hoping Intel does have a Foveros x86 chip that will come out swinging... XE GPU chiplets the whole nine. AMD can hit back with a Zen2+ and the big winners are consumers.
 
Data Center market is where the margins are at.

Intel is royally fcked until at least 2021 in data center market.
Their 14nm Xeons are power hog compared to equivalent core count AMD Epyc Rome releasing in September on 7nm.
Intel 10nm yield is atrocious and will likely never support their Xeon data center segment, expect more 4 core laptop chips on 10nm.
According to Intel roadmap above, Intel 7nm comes out in 2021 & this is the next chance Intel will have to compete again with AMD in the high core count data center market .
 
Last edited:
NWRMidnight said:
Actually, AMD has the IPC crown now. The only reason Intel is appears to have that advantage, is core speed, but clock for clock, AMD is faster IPC wise.
Click to expand...

AMD is faster in some workloads and slower in others, and have shown that they simply cannot get clockspeeds up with Zen.

NWRMidnight said:
Intel is going to lose some of it's core speed with the die shrink. How much is the million dollar question.
Click to expand...

Again, Intel is releasing the mobile parts first, and you're proclaiming that those parts will define all of their efforts.

You have no idea.
 
NWRMidnight said:
Actually, AMD has the IPC crown now. The only reason Intel is appears to have that advantage, is core speed, but clock for clock, AMD is faster IPC wise. Intel is going to lose some of it's core speed with the die shrink. How much is the million dollar question.
Click to expand...

Yep, and it could be up to 13% against that 18% IPC gain, according to Intel's own slides:
0015-1480x833.jpg


Max turbo of 4.1 ghz on the 4/8 (it doesn't actually say on the 4/8). Meanwile, the 8569u could hit 4.7ghz with the same 28w tdp.
 
It's actually worse than a 5% difference if you run the math:
4.1ghz x 1.18(ipc) = 4.84
4.7ghz x 1.00 = 4.70

.14/4.7 = 3% :p
 
IdiotInCharge said:
AMD is faster in some workloads and slower in others, and have shown that they simply cannot get clockspeeds up with Zen.



Again, Intel is releasing the mobile parts first, and you're proclaiming that those parts will define all of their efforts.

You have no idea.
Click to expand...

AMD chose to increase their IPC and core count by 2x. If they could have clocked the chips to 5 Ghz they would have but the fact that at 4.6Ghz (sustained) they win almost everything outside of a few outliers (some specific gaming scenarios).


One thing you love to leave out is that AMD Competes at 65W with a CPU that Intel made that draws 2-3x, depending on the clock speed.


They are both great products, but I think AMD gets the nod due to the lack of security issues and the fact they fixed the 2 worst ones in HW.
 
IdiotInCharge said:
AMD is faster in some workloads and slower in others, and have shown that they simply cannot get clockspeeds up with Zen.



Again, Intel is releasing the mobile parts first, and you're proclaiming that those parts will define all of their efforts.

You have no idea.
Click to expand...

Might I suggest you go read what I said again. I stated clock for clock.. AMD wins (IPC wise).. there isn't any of this they win some/lose some when tested at the same clock (any that intel wins has nothing to do with IPC, but memory and fabric latency). You also seemed to miss where i stated that Intel will lose clock speeds with die shrinkage.. we won't know just how much until they release one.

Now, to proclaim anything about mobile parts, i would first have to be talking about mobile parts, and considering nearly all the discussion is about AMD's achievments with ZEN2 which is desktop, I don't know why you keep throwing mobile parts into the discussion. If you are trying to compare AMD mobile against Intel Mobile, currently that is comparing 12nn to 10nn. Not AMD's Zen2 7nm architecture.

Intel won't have anything out for desktop based off 10nm or 7nm, till late 2020 or 2021.. and then we will see how much of drop im clock speed Intel gets hit with. So, we may find out that AMD has been able to get clock speeds up for a 7nm part. You keep wanting to compare 14nm clock speeds to 7nm clock speeds, which can't be done. So, please tell me I have no idea again, because i am not the one who is bringing cake to a pie eating contest.
 
Last edited:
ChadD said:
I don't know not that long ago Intels head engineer was talking about how they basically Fd 10nm up and they had to toss a ton of work, that will just never really go.

At the same time they where talking about 3D stacked chiplets. Now I know they are planning to bring lakefield later this year as their first 3D stacked chiplet SOC design to compete against Qcom. Its possible that is all he was referencing.

Still it makes me wonder how aggressive/experimental they will get with their Foveros 3D stacking stuff.

Intel is going to swing back... as I see it one of three things is going to happen early next year.
1) Intel is going to release a chip that is basically just their current stuff shrunk to 10nm but working. Gain some IPC drop some clock speed and outside of a golden snitch super part to win all the benchmarks their product line will pretty much be a wash.
2) They swing for the fences with a Foveros 3D stacked chiplet design, and pretty much set AMD right on their ass.
3) They swing for the fences with a Foveros 3D stacked chiplet design, and fall flat on their face.

Their engineers saying there eating crow and dumping years of work doesn't lead me to believe the first scenario is all that likely. So I believe its going to be an interesting 2020. I am really hoping Intel does have a Foveros x86 chip that will come out swinging... XE GPU chiplets the whole nine. AMD can hit back with a Zen2+ and the big winners are consumers.
Click to expand...

thing is AMD has already said they're working on their own 3D stacking as well.. i definitely think AMD has actually planned out ryzen properly for once and what they want to do over the next 3-5 years. next 3 years is going to be pretty interesting from both companies.
 
sirmonkey1985 said:
thing is AMD has already said they're working on their own 3D stacking as well.. i definitely think AMD has actually planned out ryzen properly for once and what they want to do over the next 3-5 years. next 3 years is going to be pretty interesting from both companies.
Click to expand...

No doubt... Intel is for sure going to beat AMD to 3D stacking. Intel is already using their 3D tech for their lakefield SOC. It has a 22nm base with 10nm chiplets.
They haven't confirmed they will be doing that for desktop parts next go round... but it seems logical at this point.

Of course its possible they hit some unexpected snags and end up giving AMD even more of a bump. I agree that AMD went the right way around. Get chiplets working... then consider the more exotic 3D stacking. No doubt the 3D stack will be the best option in the future IF it can scale up and not cost more then it's worth.

Ya next year is going to be interesting 3-5 of as well. I think we see Intels chiplets much sooner then people would expect though. With Lakefiled already up and running... as a low power SOC x86 part. It seems logical to me that their next push will be full on. Heck I even expect they will incorporate some form of their XE graphics before the end of 2020. Looking forward to the next 12-16 months. :)
 
IdiotInCharge said:
I did- but you didn't, and you still have no idea ;)
Click to expand...

Obviously you didn't, as you would have never made your comment, since it has already been proven that AMD now holds the IPC crown (we are not talking about clock speed here). When it comes to IPC, there is not win some or lose some.. either they have a faster IPC or they don't (they do). But according to you, I have no idea...
 
NWRMidnight said:
Obviously you didn't, as you would have never made your comment, if you actually did so, since it has already been proven that AMD now holds the IPC crown (we are not talking about clock speed here). But hey, I don't have any idea according to you.
Click to expand...

I contested your assertion- which you seem to take for granted, and then did not back up, and now you're deflecting by trying to divine my motivations, lol. I was direct.
 
IdiotInCharge said:
I contested your assertion- which you seem to take for granted, and then did not back up, and now you're deflecting by trying to divine my motivations, lol. I was direct.
Click to expand...

WTF are you talking about? You don't have a clue!! But keep acting like you do.. makes me laugh.
 
Last edited:
NWRMidnight said:
WTF are you talking about? You don't have a clue!! But keep acting like you do.. makes me laugh.
Click to expand...

You still haven't responded to my first quote of your unsupported assertion. You going to respond to that or turn this into your ad hom deflection thread?
 
The techspot article to me does not show that ryzen has a higher IPC. It shows that the IPC is about the same.
 
next-Jin said:
Do they really care though? They still hold an IPC advantage and one would think once Intel moves to 7nm they would have that plus a new architecture with more than just some 2-3% IPC increase.

AMD has dramatically closed the gap and will win back market share but Intel will swing back before they need to seriously worry about losing major market share.

Unless there is a huge switch to EPYC that is. AMD still has to get their laptop chips out and compete.

If Ryzen 2 can beat cheap Intel chips in laptops while remaining power efficient AND win over the server/data then Intel needs to worry.
Click to expand...

Like Intel hasn't been trying to swing back? When Intel decides a price cut is the best way to respond, that already means they don't have an immediate answer available.
 
drescherjm said:
The techspot article to me does not show that ryzen has a higher IPC. It shows that the IPC is about the same.
Click to expand...

421 vs 461/459 single core.. 40 points difference is not about the same.. it is 9.5% faster. (not the 6% as claimed in the article, they screwed up their math) Most games are not a test of IPC, but rather memory latency and Infinity fabric latency..
 
However Ryzen lost all of the gaming benchmarks. That is why I would call it a tie.
 
It's been said here many times, but "Intel can't even get to 10nm, no way they can get to 7nm" isn't how semiconductor processing works.

Intel's 10nm process is a multi-patterning process based on traditional immersion lithography, which has been the source of most of their woes. They did this to defer the costs of moving to EUV by one more generation. Their 7nm process will be a single-patterning process, which means it's not going to have the yield issues of their 10nm process.
 
NWRMidnight said:
I don't think you understand what deflection is.. can't deflect something when you haven't asked a question, but here you go, since you are so ignorant:

https://www.techspot.com/article/1876-4ghz-ryzen-3rd-gen-vs-core-i9/

The only advantage Intel has is clock speed, and latency with memory and infinity fabric.. NOT IPC.. you get it yet? Or are you going to throw around more mobile BS.
Click to expand...

There we go- you link an article that doesn't support your assertion. Congrats!
 
IdiotInCharge said:
There we go- you link an article that doesn't support your assertion. Congrats!
Click to expand...

It supports it 100%, you just don't understand the difference between IPC and latency. I suggest you educate yourself instead of insulting people. there have been various discussions on the subject, many from high up tech people, who have stated that AMD has higher IPC, but it is clock speed, as well as memory and infinity fabric latency that is allowing Intel to hold the crown in gaming.. But, keep your head in the sand..
 
AMD itself when explaining Zen 2 talked about how it was the latency n of the cache and the memory that b which was holding them back in gaming the most hence the silly Gaming Cache name to the new hierarchy cache size and the way it was associated.


So yes, he's right.
 
You must log in or register to reply here.
Back
Top