Intel's upcoming 10nm and beyond

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

  1. jaybee83

    jaybee83 Biotech-Doc

    Reputations:
    4,013
    Messages:
    11,506
    Likes Received:
    9,058
    Trophy Points:
    931
    nah, not unless ryzen 3000 will be an absolute disappointment. expect that stagnation only on the intel side for a while...

    Sent from my Xiaomi Mi Max 2 (Oxygen) using Tapatalk
     
    hmscott and ajc9988 like this.
  2. ajc9988

    ajc9988 Death by a thousand paper cuts

    Reputations:
    1,521
    Messages:
    5,663
    Likes Received:
    7,994
    Trophy Points:
    681
    Couple points.

    1) the desktop roadmap refers to SIPP. That is commercial client deployments. That lags consumer releases. That means you need to subtract a quarter or two from some of the chips to determine consumer releases. Tom's hardware says 9 months, but I wouldn't go that far. They are counting each quarter as 3 months, but are ignoring Intel will release comet lake in October with wider availability months later, always with the possibility of slipping a quarter.

    2) the adoption of Vulkan and DX12, along with ryzen optimization advancements means as time goes on, newer games will scale better. DX12 has CPU optimizations to more parallelize the workload, allowing better core scaling. Although single core is important still, it does improve performance on higher core counts. Vulkan is similar. And, as we've seen with Ryzen game optimizations in certain newer titles, it has added performance where an 8 core has concrete benefits over a six core chip. These will only continue with time.

    3) comet lake may have sunny choice architecture advancements. Looking at each revision in architecture, meaning Ivy to haswell and broadwell to sky lake, Intel has improved IPC around 11%. This isn't overall performance, just IPC.

    This least point is important. AMD is looking at 11-15% IPC over zen and zen+. Zen is around 7% lower IPC than Intel's current offerings. So that means AMD should have 4-8% IPC over Intel, depending on workload. If Intel releases comet lake with an 11% IPC gain, that would put them 4-7% ahead on IPC again. This means AMD really needs frequency improvements to keep the single core lead (they already win for core count and multi threaded).

    Now power consumption to achieve the performance is where Intel will really be hurt! But that is a different story.

    Sent from my SM-G900P using Tapatalk
     
  3. Robbo99999

    Robbo99999 Notebook Prophet

    Reputations:
    3,981
    Messages:
    6,563
    Likes Received:
    5,625
    Trophy Points:
    681
    We'll have to see what kind of game development can really challenge these 8 core 16 thread CPUs, historically games have mostly been GPU bound apart from high refresh rate gaming, but future game development may change that I suppose. I don't know much about game development, but I think the more 'assets' you have on screen at any one time, then the more CPU hungry it is - I think this was highlighted by the DX12 3DMark CPU test, where DX12 showed a massively more complex scene than DX11. Yep, I think in order to take advantage of all these extra CPU cores that it will mean more complex game environments, we'll have to see if the demands scale with the development of these new CPUs.
     
    ajc9988 likes this.
  4. yrekabakery

    yrekabakery Notebook Virtuoso

    Reputations:
    790
    Messages:
    2,447
    Likes Received:
    2,319
    Trophy Points:
    181
    You're forgetting the x-factor. Next-gen consoles arriving in 2020 will sport 8C/16T 7nm Zen 2 CPUs, so as with every new console generation, expect a huge increase in baseline CPU requirements in next-gen games as those boxes are what drive AAA game dev.

    We already saw that happen this gen starting a few years ago, with games like BF1/BFV, Watch Dogs 2, and AC Origins/Odyssey making 4C/4T CPUs obsolete for 60 FPS. Thank god for Ryzen giving Intel the kick in the ass it needed to give us more cores on mainstream/non-HEDT CPUs.
     
    Last edited: Apr 29, 2019
  5. Robbo99999

    Robbo99999 Notebook Prophet

    Reputations:
    3,981
    Messages:
    6,563
    Likes Received:
    5,625
    Trophy Points:
    681
    Yes, I remember seeing reports about those next gen consoles over on Guru3D and it lead me to speculate that it'll put pressure on my 4C/8T CPU in the future! Those are probably gonna be pretty low clocked CPU cores in those consoles I'm thinking - so 8C/16T PC desktop CPUs are gonna pack more of a punch due to frequency (maybe 25% extra?), but then again they don't have quite the same optimised environment as a console, so maybe it's Even Stephens on that front.
     
    lctalley0109 and ajc9988 like this.
  6. ajc9988

    ajc9988 Death by a thousand paper cuts

    Reputations:
    1,521
    Messages:
    5,663
    Likes Received:
    7,994
    Trophy Points:
    681
    Even with the lower clocks looking roughly like mobile CPUs, it looks like they will be adopting the CPU optimizations for DX12 or Vulkan. M$ obviously has an incentive to use their own DX12, but if Vulkan starts being used, even if a tweaked version, for PS5, it will be interesting to see where things go.

    But, even those games mentioned, many of them didn't get Ryzen CPU optimizations until Feb/March of this year.

    But, definitely gotta be happy for the kick in the pants. It will be interesting to see the Comet Lake 10 core against the Zen 2 12-core. With the rumored 15% IPC, it would still take a Zen 2 based 8-core likely 4.7GHz all core to match a 9900K clocked to 5GHz. The engineering samples only do 4.5GHz, so unless they got more gas in the tank, it would still take the 12-core to beat the 8-core Intel chip. And that is assuming the 4.5GHz is all core OC, not single core boost.
     
  7. jaybee83

    jaybee83 Biotech-Doc

    Reputations:
    4,013
    Messages:
    11,506
    Likes Received:
    9,058
    Trophy Points:
    931
    so even in worst case scenario, the 16 core zen 2 will likely beat everything that intel has to offer in its mainstream lineup :)
     
    lctalley0109, ajc9988 and hmscott like this.
  8. ajc9988

    ajc9988 Death by a thousand paper cuts

    Reputations:
    1,521
    Messages:
    5,663
    Likes Received:
    7,994
    Trophy Points:
    681
    And they know it. That is why Red Gaming Tech has said two things: 1) AMD has no idea how to market and who the target audience is for a 16-core mainstream CPU (some workstation work needs the memory bandwidth, other needs the PCIe lanes to scale out add-in cards, like graphics cards, to accelerate processing, etc.); and 2) AMD wants to wait on the 16-core release to keep hype built up for the process and their products. Technically, a sub-point to number 2 is that they also want to respond in the event Intel's 10-core somehow beats their 12-core CPU.

    Personally, I feel any mainstream CPU at 8-cores and above, regardless of AMD or Intel, has two primary consumers: 1) game steamers, and 2) entry level to intermediate content creators, like YouTubers, hobbyists, etc. The second group is based depending on the amount of content produced, the quality necessary for the content, etc. As they grow, they will likely move toward HEDT builds as it makes sense. But this will help get them used to platforms before that point. And due to more cores and higher performance graphics cards in the industry at large, there is a burgeoning of steamers and content creators online taking advantage of the lower cost to entry.

    But, "that is like [my] opinion, man." (Big Lebowski).

    Sent from my SM-G900P using Tapatalk
     
    jaybee83 and hmscott like this.
  9. AlexusR

    AlexusR Guest

    Reputations:
    0
    I would not say 8-core CPUs are only useful for streamers (who can actually use Intel's QuickSync or Nvidia's NVENC to encode the stream, which would be cheaper and give better performance) - gaming consoles right now use 8-core CPUs so many game developers have to learn to optimize games for maximum scaling for those 8 cores. And this will also affect PC ports of console games or games which are released both on PC and consoles, especially MMORPG games with large PvP battles where you have to do A LOT of CPU calculations for things like position and action of each player (there can be 100's of them in single battle) or physics calculation for large destructible structures without relying on proprietary APIs like PhysX (for example, the upcoming Camelot Unchained MMO has its own proprietary physics engine that is not using PhysX or other existing APIs).

    Now the 16-core is a little bit excessive. Although if AMD would've convinced console manufacturers to use such CPU for next console generation - game developers would've quickly found the ways to use all those extra cores to maximize the performance ;-)
     
    Arrrrbol likes this.
  10. bennyg

    bennyg Notebook Virtuoso

    Reputations:
    1,533
    Messages:
    2,337
    Likes Received:
    2,315
    Trophy Points:
    181
    Can't we just look at what workloads the current 16 core CPUs show good scaling with, and make allowance for better gaming performance on non-NUMA/non-mesh memory subsystems ?

    I get why AMD are not rushing to release 12/16 core mainstream chips... they will just cannabalise sales of their own 12/16 core Threadrippers which will force firesale prices like what we saw at EOL of the 1900X

    If Intel keep the same socket for the rest of their 14nm desktop CPUs... Maybe one day for teh gits n shiggles I'll drop a 12 core in this thing that was originally sold with a 2015 quad core ... 1151 may end up longer lived than AM4 ROFL.
     
Loading...

Share This Page