Intel's upcoming 10nm and beyond

Discussion in 'Hardware Components and Aftermarket Upgrades' started by ajc9988, Apr 25, 2019.

  1. Robbo99999

    Robbo99999 Notebook Prophet

    Reputations:
    3,986
    Messages:
    6,563
    Likes Received:
    5,627
    Trophy Points:
    681
    Which is completely at odds with the supposedly leaked roadmaps! Makes me think the roadmaps are fake. But, if 10nm not coming to Intel until 2022, then that's not good for them vs AMD's purported future with increased IPC & frequency, because AMD are obviously already winning with core numbers & price, if they can get IPC & frequency up significantly then it'll be down to Intel to offer a response, and I'm thinking they'd need 10nm for that.
     
  2. Papusan

    Papusan JOKEBOOK's Sucks! Dont waste your $$$ on FILTHY

    Reputations:
    23,032
    Messages:
    23,733
    Likes Received:
    41,257
    Trophy Points:
    931
    None is happy seeing their competitor eat much more of the cake. Of course they can't wait +3 years to answer the threat. Would you if you run an business?
     
    ajc9988 and Arrrrbol like this.
  3. Arrrrbol

    Arrrrbol Notebook Deity

    Reputations:
    3,062
    Messages:
    710
    Likes Received:
    1,040
    Trophy Points:
    156
    IMO they only keep developing the iGPU to get a head start for when they start producing dGPUs. Technically Intel is the largest GPU manufacturer if you count iGPUs.
     
    Rei Fukai, hmscott and Papusan like this.
  4. hmscott

    hmscott Notebook Nobel Laureate

    Reputations:
    6,572
    Messages:
    19,895
    Likes Received:
    24,718
    Trophy Points:
    931
    Exactly, keep that basic video controller outside the CPU in the chipset where it belonged all along.

    Now with Hybrid packages - now chiplets - it's a matter of multi-layer communication close to the CPU die, but off-die to maintain separate power / cooling - with fancier IHS "heat-pipes"(?) - on separate silicon.
    Intel shouldn't even be on that GPU volume tracking list, it's Intel marketing BS at it's finest.

    Intel's on board iGPU shouldn't count in a discrete GPU tracking comparison, that would make those tracking numbers more sensical without Intel.
     
    Arrrrbol likes this.
  5. ajc9988

    ajc9988 Death by a thousand paper cuts

    Reputations:
    1,526
    Messages:
    5,679
    Likes Received:
    8,018
    Trophy Points:
    681
    Actually, it is not. Ice lake-u quad core is on the roadmap. Also, entry level Xeon on 10nm (115x socket) is as well. But who here cares about a mobile quad core or likely a low core count entry level Xeon? Also, we have to look at yield and fab lines. Yields are so bad currently that only a quad core is possible, up from a dual core. This matches with the "double" comment from Swann this afternoon regarding yields since the changes, although he could have meant a 50% reduction in defect density. To put this in perspective, AMD with TSMC is already at 70% yields, which comes out to a defect density around 0.45/mm2, which is enough for mass production on a new process, which will only reduce with time, and would yield over 500 non defective dies on a 7nm wafer costing $11,000 per wafer, meaning each die would be around $22 for an 8-core. That doesn't account for scavenging usable dies, which would give a higher effective yield.

    With that said, Intel also has the IMC and IO on die, which is hard to shrink. But, Intel only has three lines to produce 10nm and just invested heavily in expanding 14nm, meaning there is a need to recoup cost and they have extremely limited capacity for 10nm lines at the moment.

    As I explained in a different thread, details on the leak say it was a SIP roadmap presented to Dell, meaning commercial machines, not consumer machines, which lags behind by a couple quarters. Then there is a question of how current the map is. So let's give the benefit of the doubt and pull all products in by 1-2Q. That puts comet lake at Q4 to Q1 2020, which around Q3-Q4 were the rumored for comet lake anyways, with the possibility of slipping to Q1 2019.

    With that explanation, does it seem more possible now?

    Sent from my SM-G900P using Tapatalk
     
  6. Robbo99999

    Robbo99999 Notebook Prophet

    Reputations:
    3,986
    Messages:
    6,563
    Likes Received:
    5,627
    Trophy Points:
    681
    Long day at work, so can't follow all the threads of your thoughts/information there, but your initial post where you quoted the links to the various news articles, you summarised in your post before those links with "Bad news for people on the Intel platform: NO 10NM UNTIL 2021 EARLIEST, 2022 DESKTOP.". So, you're saying backend of 2020 for desktop 10nm now, rather than the 2022 you mentioned in your earlier post?
     
    ajc9988 likes this.
  7. ajc9988

    ajc9988 Death by a thousand paper cuts

    Reputations:
    1,526
    Messages:
    5,679
    Likes Received:
    8,018
    Trophy Points:
    681
    Pictures can help.
    Mobile roadmap.png Intel SIPP Roadmap.png
    So, I misstated, if being absolutely critical.

    Above on the client mobile roadmap, you will see Ice Lake U in a 2c/4c offering that is limited coming out this year. In fact, it says half way through Q2. Then, you get Tiger Lake U in 4c offering on 10nm in Q2 of 2020. You also get, in Q3, Rocket Lake U 4/6c with 10nm graphics. On the second image, the Client Commercial roadmap, you can see NO listing for 10nm until Tiger Lake U and Y in Q2 of 2021, both in 4C variants. Allegedly, from reports, certain Xeon E parts may get 10nm.

    So, long end short, nothing above 4C 10nm is planned until after 2021, which places it in 2022. Intel should have 7nm close to around then, so it seems they are skipping 10nm entirely.

    Now, for anyone here, do you want 4C or less? Do you want a "U" or "Y" series low power chip? Otherwise, no, there is NO 10nm planned.
     
    lctalley0109, Robbo99999 and hmscott like this.
  8. Papusan

    Papusan JOKEBOOK's Sucks! Dont waste your $$$ on FILTHY

    Reputations:
    23,032
    Messages:
    23,733
    Likes Received:
    41,257
    Trophy Points:
    931
    Ok, we say 2022:biggrin: Since you are so sure.
     
  9. ajc9988

    ajc9988 Death by a thousand paper cuts

    Reputations:
    1,526
    Messages:
    5,679
    Likes Received:
    8,018
    Trophy Points:
    681
    Intel has given me no reason to believe in them otherwise in regards to 10nm. They have limited production lines and invested in having more production on 14nm. The roadmap allegedly comes from a couple months ago. Although yields have gotten better, they are still low. OEMs are deciding to put out better AMD laptops and integrate them into more desktops, signalling them having weaker confidence in Intel. Intel's 14nm shortage has now been placed into Q3 of 2019. We know Comet Lake will be 14nm. So the only question is if Rocket Lake S will also be on 14nm. Either way, Rocket Lake will be end of 2020 or start of 2021. That is one full year worth of lineup. If it is on 14nm, then it will be confirmed. And 14nm going against 7nm+ on EUV lithography in 2020 lineup is bad anyways, needless to say if they have Rocket Lake on 14nm, which would then go against a likely 5nm EUV chip on the other side.

    But, with little to the contrary, I'll go with the leaks. Not only that, Intel should have 7nm finally ready for volume in 2021, meaning in 2022, they would have those chips, which may be the largest jump in performance on Intel's side in a long time, if ever, if they are going from 14nm straight to 7nm.

    Also, as I have said, that is process tech. They will continue improving architecture over that time period. But Intel being stuck on 14nm, after being late with 14nm then giving Broadwell, I'd have to say yes, they are having huge issues.
     
    hmscott and bennyg like this.
  10. Robbo99999

    Robbo99999 Notebook Prophet

    Reputations:
    3,986
    Messages:
    6,563
    Likes Received:
    5,627
    Trophy Points:
    681
    Thanks, makes sense. So nothing much better than the 9900K likely until 10 core Comet Lake in Q2 of 2020? And that's still on 14nm. (I'm thinking along the lines of 'ring bus' gaming oriented CPUs). To be honest though, in terms of gaming I don't think we'll need anything better than the 9900K for a number of years! I see CPU development stagnating now somewhat for consumer/gaming needs both in terms of requirements and performance offered. I think this because we've been on 4C/8T for so long as an upper end consumer/gaming CPU, then 'suddenly' we get 8700K & 9900K where we double the number of cores. That large increase in performance after all the years of stagnation makes me think there is gonna be a lot of headroom left in these CPUs for a number of years to come. I don't see the 8700K and certainly not the 9900K becoming sub-optimal for gaming for quite a few years now. Perhaps now CPU development will go back to the boring snail pace of before like the long reign of the 4C/8T CPU!
     
    ajc9988 and joluke like this.
Loading...

Share This Page