Intel Delivers a New Product Roadmap

Posted on July 27, 2021 by Paul Thurrott in Hardware, Mobile with 35 Comments

Pat Gelsinger, CEO of Intel Corporation, speaks during a virtual presentation as part of the “Intel Accelerated” event on July 26, 2021. At the event, Intel presented the company’s future process and packaging technology roadmaps. (Credit: Intel Corporation)

With new CEO Pat Gelsinger now firmly in charge, Intel has delivered a new product roadmap for the next several years.

“Building on Intel’s unquestioned leadership in advanced packaging, we are accelerating our innovation roadmap to ensure we are on a clear path to process performance leadership by 2025,” Mr. Gelsinger said during the firm’s Intel Accelerated webcast yesterday. “We are leveraging our unparalleled pipeline of innovation to deliver technology advances from the transistor up to the system level. Until the periodic table is exhausted, we will be relentless in our pursuit of Moore’s Law and our path to innovate with the magic of silicon.”

As part of the new roadmap, Intel also introduced a new naming convention for its process nodes, which have recently stopped mapping directly to the manufacturing processes used by its competitors. So instead of using nanometer-based process node naming, as in the past—7 nm, 10 nm, and so on—Intel is instead naming its coming nodes based on how they match up with the competition.

The current 10 nm node process, which uses SuperFin remains unchanged, but the next-generation node process, previously called Enhanced SuperFin and currently in production, will now be called Intel 7. After that, the first true 7 nm Intel node process will be called Intel 4 and is due in 2022, and then the firm will ship Intel 3 and Intel 20A processes in 2023 and 2024, respectively.

That Intel 20A architecture is apparently the big leap here. Dubbed by the company the “angstrom era of semiconductors,” 20A will be Intel’s first new transistor architecture since 2011. And Qualcomm has already signed on to use Intel’s 20A process technology in future chipsets.

Beyond that, Intel plans an 18A architecture for early 2025 that will deliver “another major jump in transistor performance.”

You can watch the Intel Accelerated webcast here.

Tagged with

Join the discussion!

BECOME A THURROTT MEMBER:

Don't have a login but want to join the conversation? Become a Thurrott Premium or Basic User to participate

Register
Comments (35)

35 responses to “Intel Delivers a New Product Roadmap”

  1. MichaelDarby

    How misleading and clear as mud is that naming convention? It's the AT&T school of misdirection all over again (5Ge = 4G).

    • mattbg

      They are actually trying to deal with "#nm" having become a marketing term rather than a meaningful technical description.


      When the article above says "the first true 7nm Intel node process will be called Intel 4", that's just adding to the confusion. Intel 7 has comparable transistor density to the competition's so-called 7nm node.


      Intel is behind for sure, but they're not as behind as the use of an outdated transistor pitch metric to measure progress would indicate.


      I'm not sure this matters, anyway - it's just what the market has become. What really matters is power consumption, heat dissipation, and performance. Intel is behind in those areas at the moment, and that's what matters.

      • bluvg

        Spot on. Modern CPUs often don't even use the same process node throughout, so it's a near-pointless distinction. At the end of the day, power + perf.

      • bettyblue

        "power consumption, heat dissipation, and performance"


        This! Intel was still in the lead when it came to IPC for single core workloads. However I believe the Ryzen 5000 series just surpassed them to take away their only bragging point.


        I think Intel may have some thermal nuclear models that actually hit a higher clock speed over AMD at the cost of running the SUN in your PC.


        That said AMD no longer actually makes their chips, they design them and TSMC makes them, which it is really TSMC that is using their tech to make the 7/5nm chips. Where Intel still makes theirs. If Intel can solve a few problems, they could make a come back since they currently have FABS to do that. I have heard rumors that they might get out of that business or something a long those lines and rely on a TSMC or whoever to build theirs as well.

        • bluvg

          "I believe the Ryzen 5000 series just surpassed them"


          Depends which benchmark and also mobile vs. desktop, but yes, AMD generally is ahead of Intel in both single- and multi-core with Zen 3. Zen 4 also looks to offer a substantial gain over Zen 3.


          "I have heard rumors that they might get out of that business or something a long those lines and rely on a TSMC or whoever to build theirs as well."


          Intel has said they will outsource some, but they also are apparently in talks to acquire GlobalFoundries, and they plan to offer foundry services for others (much like TSMC). They just announced a $20B plan for two new fabs in Arizona, so they're definitely not getting out of manufacturing, and they said they will be getting some ASML (EUV) equipment ahead of TSMC and Samsung for CPUs coming in 2025. All this news telegraphs they have no interest in being anything other than #1.

  2. behindmyscreen

    So...the "Angstrom era" is just a rescaling of the transistor size....they hit 2nm (performance...in their new naming)...but 20A sounds better because they can then go down a tenth of a nm and make it seem like they have made a large leap.

    • bluvg

      It's not about sounding better. It no longer makes sense when you're down to 5, 3, 2, etc. They did this when they went from µm to nm also. But regardless, the "nm" label has not been an accurate metric for a long time.

  3. dstrauss

    Anyone else think of the Wizard of Oz - "Pay no attention to the man behind the curtain!"


    Or maybe this one - "If you can't dazzle them with your brilliance, baffle them with your bull****!"

    • whiplash55

      I think Intel is on the right track, but only time will reveal if that's true. Thanks to AMD, they have no choice, and they have an engineer in charge again. As it should be.

    • bluvg

      This seems like reflexive hatred of Intel. Whatever your feelings about the company, this is actually a good move for the industry, similar to the focus shifting away from clock speed.

  4. johnh3

    Atleast Intel got a Ceo now that know what he talking about and a clear plan for the future. A big step from the previous leadership.


  5. jonsimon

    But what J.A.R.V.I.S.? (no response expected)

  6. lvthunder

    Hopefully, they can deliver on this roadmap.

  7. brettscoast

    Good post Paul

    Intel has some catching up to do when it comes to new chip design and implementation.

  8. thewarragulman

    This new naming scheme reminds me of the old "Performance Rating" system AMD used back in the Athlon XP/Athlon 64 days, where they used a name like "3200+" to state that their processor was equivalent to a 3.2 GHz Intel processor or faster, despite not actually being 3.2 GHz. This "Intel 7" and future names they're going to give is the exact same thing, their processor won't actually be 7nm, but it's their idea of what's equivalent to everyone else's 7nm.


    It's silly and misleading just like the old PR naming scheme during the GHz wars. Stop it, Intel. Just make a good product without the confusion and people will buy it if it's better than the competition.

  9. bluvg

    "Intel 7" and future names they're going to give is the exact same thing, their processor won't actually be 7nm, but it's their idea of what's equivalent to everyone else's 7nm.


    That's the opposite of what's happening. The "nm" label doesn't matter--for any manufacturer--and it hasn't in a long time. It's not a useful metric. Intel's 10nm is equivalent in transistor density to TSMC's 7nm. Neither the "10nm" nor the "7nm" represent a physical reality as they did at one time. This name change takes Intel from a useless marketing term ("10nm") to something that is equivalent to TSMC, not just wishful thinking.


    If anything, Intel is going from something that was misleading (and not to their benefit) to something that is more accurate.

  10. chrishilton1

    This naming convention sounds dumb, how to confuse the consumer and hide your flaws

    • bluvg

      No, 10, 10+, 10++, etc. was far more confusing, and ultimately not useful since it is actually comparable in transistor density to TSMC's "7nm" node (actually NC7--they do it too, as the whole "nm" thing has been a useless label for a long time). This is not Intel pulling a fast one, it's actually TSMC and Samsung that received the benefit of Intel's previous process labeling. Intel is now aligning with the rest of the industry.

  11. timwakeling

    I seem to remember when AMD stopped needlessly ramping up the GHz on its processors, but knew customers still used GHz to compare, they introduced model numbers based on Intel's GHz. Seems like exactly the same thing now with the boot on the other foot, so to speak. :)

    • bluvg

      I think that was more associated with Intel when Pentium was their flagship (Netburst architecture)--they introduced a 1.7 GHz model around the time of Windows XP, and that was a huge clock speed leap over the competition of the era (I think 1 GHz). There were few advantages to that approach, and it was mostly a failure. IPC started to enter the consciousness of savvy buyers because of it.


      This is not akin to that. This change aligns better with buyer perception, rather than distorting it.

    • mattbg

      Agree, but the transistor gate pitch metric is far less meaningful. AMD was at least aiming to get a measure of Intel vs. AMD performance equivalence. If you go looking for a CPU and say "I want a 7nm CPU", you could get a 10W part or a 120W part and there is going to be a huge difference between them.


      The vast, vast majority of people don't care about speed anymore, and especially not about pitch. Most devices are fast enough for most people. Further, the vast majority of people don't even care about power consumption and heat dissipation. They care about how the device looks, fit and finish, how it performs, and whether or not the battery life is good enough for their use case. There are proxies in there for pitch, but the pitch itself doesn't matter. Intel suffers in some of those areas, but who can say whether or not it's enough to matter outside of tech websites. It probably doesn't matter, and that's why Intel is putting marketing around it with things like "Evo".

    • nkhughes

      Yes, I have an old HP box here from 2006 - it's got a dual-core 2GHz 3800+ Athlon 64 X2, and I think the 3800 refers to a theoretical Pentium 4 single core speed?

  12. bluvg

    "mapping directly to the manufacturing processes used by its competitors"


    The process nodes "nm" descriptors don't really map directly for TSMC or Samsung, either.

  13. winner

    New Powerpoints. How cute.

  14. eric_rasmussen

    I can buy an Intel system that performs tasks at speed 'X' while consuming 275W of electricity.


    I can buy an AMD system that also performs tasks at speed 'X' but only consumes 65W of electricity.


    I know most people don't give a crap about electricity consumption, but when you talk about a billion devices in the world you're talking about the difference of trillions of tons of CO2 released into the atmosphere. We should be very interested in power consumption if we're responsible stewards of this planet's resources. Therefore, let's start marketing CPUs based on real power consumption values, not fake marketing terms.


    Of course doing this would mean Apple's M1 would mop the floor over both Intel and AMD.

    • bluvg

      Everyone on the silicon side is concerned about this, and they're addressing it with both process and architecture. Intel is also going the same hybrid "big little" route as Apple/ARM. The M1 benefits greatly from on-chip memory, but that has some obvious tradeoffs not well-suited to the general CPU market (want M1 with more than 16 GB RAM? Get two computers).


      Some other wins would be on the software side: consumers calling for reduced use of gratuitous graphics (4K on mobile phones and laptops is often just wasteful) and demanding lower-resource software and DEMANDING ads that do not suck up CPU resources (this site and Petri.com have both been just a total assault on CPU usage recently; yes, there are ad blockers, but I don't use them on this or Petri.com). Another huge win would be addressing cryptocurrency mining, but with the money at play there... tough sell.

    • lvthunder

      They need to market heat as well. Chips that run cooler can be quieter and feel better to the touch.

    • garymount

      There is a great benefit to having more CO2 in the atmosphere, it helps green the planet as after all it is plant food.

    • crunchyfrog

      Adding to this, all of that higher wattage increases heat generation as well which has to be cooled properly to avoid thermal throttling.

      • Oreo

        Another issue is noise: my 16” MacBook Pro is unusably noise for when I want to record lectures. An M1-based machine (or my 2018 iPad Pro) can do the same work without a fan potentially.

Leave a Reply