Comments Locked

88 Comments

Back to Article

  • Mr Perfect - Wednesday, October 7, 2020 - link

    I look forward to Intel offering something newer then Skylake on the desktop. Q1 is a wide window though, such vague launches typically turn out being the end of the window.
  • FunnyRed7 - Wednesday, October 7, 2020 - link

    I wouldn't expect it any sooner than March 31st 11:59pm.
  • qlum - Wednesday, October 7, 2020 - link

    * do note that availability will be good around late August.
  • psychobriggsy - Thursday, October 8, 2020 - link

    Apparently it's week 10, so early March.

    However, launch may not mean availability. Zen 3 'launches' later today, but availability is later this month.
  • Gondalf - Sunday, October 11, 2020 - link

    What do you mean with availability?
    Is AMD capable to supply the OEMs channels? :). No obviously.
    Zen 2 was in real volume ? No, Zen 3 will be in real volume, no.

    To Intel care nothing of this thing. It is irrelevant for the market.
  • Samus - Tuesday, October 13, 2020 - link

    Intel's branding is so fucking confusing right now that I don't even know how OEM channels handle it.

    And quite frankly I don't think AMD has a supply issue on the CPU side. Yes they go out of stock but they are constantly restocked...I've built two Zen 3 systems since August and both cases CPU's were out of stock at the time of order at B&H, but both shipped within a week after ordering. I've accepted in our current world we cannot expect things to happen rapidly like they used too.
  • JKflipflop98 - Monday, October 19, 2020 - link

    It's really not that confusing.

    Would you rather the names be something like "Intel Core i9 without GPU with AVX256 but not AVX512 and also Hyperthreading PCIE 3.0 clock speed 4.2GHz"?
  • Spunjji - Thursday, October 22, 2020 - link

    Using their segmenting bullshit as an excuse for their naming bullshit is a novel concept, I'll give you that!

    Try their Xeon nomenclature on for size, see how charitable you feel after that.
  • eek2121 - Tuesday, October 13, 2020 - link

    Some of this have known about this for quite a while. There are some more hidden gems that indicate Rocket Lake will be a bit of a turd...albeit a necessary one.
  • eek2121 - Tuesday, October 13, 2020 - link

    I can promise you it will be officially launched after 03/01/2021 and before 03/22/2021...official availability though? That's another discussion...

    Note that launch != announced.
  • yeeeeman - Wednesday, October 7, 2020 - link

    Some roadmaps have leaked online and they state it is March.
  • YB1064 - Wednesday, October 7, 2020 - link

    Intel have lost their way. This is sad to see.
  • eek2121 - Tuesday, October 13, 2020 - link

    No, they actually haven't.

    The issue is that Cypress Cove (the backport) has been in the works for quite a while, and you can't just cancel a project that you invest millions into. On top of that, Alder Lake (the true sequel on 10nm) wouldn't be ready until around May/June...So Intel, with it's corporate nonsense, decided that it would extend the gap between Comet Lake and Alder Lake by 8-9 months...by introduction Cypress Cove AKA the 11000 series.

    Unless the bean counters at Intel panic, Alder Lake will launch in March 2022, but it will be ready by June 2021...all because Intel still can't shed their corporate image. If they keep this up, AMD is going to eat them alive.
  • Qasar - Tuesday, October 13, 2020 - link

    " Alder Lake will launch in March 2022, but it will be ready by June 2021 " i will believe that when it actually happens.
  • silencer12 - Thursday, October 8, 2020 - link

    AMD already has pci-e 4.0 because Intel took that sweet slow time again (with new technologies) to implement it. There is absolutely nothing to look forward to with Intel in processors at this time. Maybe graphics cards, ssds.

    (All american rejects - move along)
  • Samus - Tuesday, October 13, 2020 - link

    AMD usually beats Intel when implementing new standards. I can think of a dozen examples in between x86-64 and PCIe 4.0

    And in manufacturing breakthroughs I can think of a dozen examples between copper interconnects (thanks to AMD-IBM-Motorola partnership) and the use of off-process chiplets.
  • JKflipflop98 - Monday, October 19, 2020 - link

    Actually, Intel was the first to market with copper interconnects. "Hey! We made this one transistor in a lab as proof of concept" doesn't really count.
  • Spunjji - Thursday, October 22, 2020 - link

    That would be "not true", on so many levels. First up, IBM did more than just make "one transistor in a lab as a proof of concept":
    https://www.ibm.com/ibm/history/ibm100/us/en/icons...
    "On September 1, 1998, IBM announced the shipment of the world’s first copper-based microprocessors. The IBM® PowerPC® 750 was originally created as a standard aluminum design operating at up to 300 MHz. By applying IBM’s copper manufacturing process to what was essentially the same chip, the company was able to produce semiconductors featuring speeds of at least 400 MHz"

    Then came AMD in 1999 with the K7, the first CPU designed with copper interconnects in mind:
    https://en.wikipedia.org/wiki/Athlon#:~:text=One%2...

    Meanwhile, Intel in 1999:
    https://www.anandtech.com/show/399/2

    Intel didn't move to copper interconnects until 2002 with the Northwood Pentium IV.
  • yeeeeman - Wednesday, October 14, 2020 - link

    you guessed it right, it will be in march.
  • nandnandnand - Wednesday, October 7, 2020 - link

    "putting those rumors to bed"

    So, Rocket Lake in mid-March, at the end of Q1.
  • eek2121 - Tuesday, October 13, 2020 - link

    March 21 is one date and March 10th is the other I think. One was Tiger Lake H and the other was Rocket Lake S, exhausted so I can't remember which is which...
  • Jorgp2 - Wednesday, October 7, 2020 - link

    >there’s no specific indication at this time that there will be an increase in PCIe lane counts from the CPU, although that has been an idea that has been floated.

    Most z490 boards specify an additional 4x lanes from the CPU. In the form of 8+8(+4)
  • Slash3 - Wednesday, October 7, 2020 - link

    In addition to the expected dedicated CPU-connected NVMe M.2 slot, I'm curious to see how the increase in DMI bandwidth is handled. Early Intel slides showed it as an eight lane DMI 3.0 (8GB/s) link, up from the existing four lane DMI 3.0 (4GB/s) link - and not an expected four lane DMI 4.0 (8GB/s) link.

    Has anyone done any board level analysis to determine if the extra traces already exist between the socket and PCH on Z490 motherboards? If not, this is a pretty solid indication that we'll see another chipset revision land alongside Rocket Lake to be able to address the extra functionality, even if the CPUs are still compatible with some existing boards.

    It would make Z490 boards sort of a B550 equivalent. Gen4 to the CPU connected lanes, and one dedicated NVMe M.2 slot (depending on model). For full functionality you'd need a Z590 (similar to X570's full Gen4 connectivity).

    Messy.
  • Slash3 - Wednesday, October 7, 2020 - link

    *Z570
  • Slash3 - Wednesday, October 7, 2020 - link

    *Nevermind, misread my own comment. I gotta quit drinking. Or start.
  • repoman27 - Wednesday, October 7, 2020 - link

    Comet Point-H, despite being nothing more than a warmed over stepping of Cannon Point-H, can in fact support DMI x8. Although it was removed in later revisions, the ballout diagram in the CNP-H datasheet used to show connections for DMI x8. In other words, DMI x8 has been sitting there unused in Intel chipsets since April 2018 because they have yet to produce a CPU that can take advantage of it. Neither CNP or CMP support PCIe Gen4, which is why Intel went with DMI 3.0 x8 instead of DMI 4.0 x4.

    Comet Point-V, which is just a rebranding of the 22nm Kaby Point-H, only has DMI x4.

    Motherboard support will vary by manufacturer and model, but visual inspection should be sufficient to confirm whether the traces are present or not.

    If there is a 500 series PCH for Rocket Lake, it would likely be based on Tiger Point-H. But if Tiger Point-LP is any indication, it still wouldn't support PCIe Gen4. So Rocket Lake would only have the x16 PEG lanes plus the additional x4 from the CPU for PCIe Gen4.
  • Quantumz0d - Wednesday, October 7, 2020 - link

    DMI is still at 3.0 from the leaked Z5xx design block diagrams. So that 4.0 Chipset is not there ? If that's the case and they are only trying to get more lanes because to maintain parity with X570 AMD chipset, but if AMD's Zen 3 achieves 5GHz clocks and improved IPC then Intel is toast, add X670 chipset refresh with ASMedia without fan this time instead of re-purposing the I/O die with more lanes or any improvements over X570 then Intel's gaming leadership and OCing for the past decade+ will come to an end.
  • Kevin G - Wednesday, October 7, 2020 - link

    I think part of it is that Intel has been outsourcing production of their chipsets of late. Moving them to PCIe 4.0 while leveraging a 3rd party's older process node would tip the power budget too much.

    Post Rocket Lake is where I see Intel moving DMI to PCIe 4.0 link rates. If they were wise, they'd prep for PCIe 5.0 at the same time. Remember that PCIe 4.0 adopting has been lacking as Intel's initial plans for adoption have been hindered with the rest of their processor roadmap. The 10 nm delays set them back in more than one way.
  • repoman27 - Wednesday, October 7, 2020 - link

    Where are you getting that idea from? Intel still fabs all their own chipsets AFAIK. CMP-H is on intel 14nm and CMP-V is on good ol' Intel 22nm.

    They're still on PCIe Gen 3 because they're still selling the essentially the same chipsets they designed for release in 2016 alongside Cannon Lake / Kaby Lake for crying out loud! PCIe Gen5 for client PCs is an extremely dubious proposition in the near term. Might as well add 24G SAS while you're at it.
  • eek2121 - Tuesday, October 13, 2020 - link

    Intel fabs their own chipsets. AMD does not.
  • Slash3 - Wednesday, October 7, 2020 - link

    Radio silence on the AMD chipset side would suggest that ASMedia still hasn't been able to deliver a product fully capable of taking the baton away from AMD's X570 solution.

    At this point it wouldn't surprise me if they simply have vendors squeeze out a few new Rev 2 B550/X570 boards to ship with the appropriate Zen 3 capable AGESA while they're focused on AM5 / Zen 4.
  • Flunk - Wednesday, October 7, 2020 - link

    There has been talk of a X670 chipset that adds USB 4 support. If it's ready it will be announced tomorrow with the CPUs.
  • Notagaintoday - Thursday, October 8, 2020 - link

    "There has been talk of a X670 chipset that adds USB 4 support"

    Whomever is doing that talking, clearly isn't very well informed!
  • eek2121 - Tuesday, October 13, 2020 - link

    I can say this is false except for the fact that there won't be a new chipset.
  • eek2121 - Tuesday, October 13, 2020 - link

    Zen 3 hits 4.9 GHz. Who cares about 100 MHz?

    What kills Intel is the fact that they can't roll out a 10 core Rocket Lake (according to Rumor).

    That means that Rocket Lake will be faster than Comet lake, but will trail AMD's planned XT refresh and will be neck in neck with the 5xxx launch.

    Note that AMD started playing this game a few months ago. Their marketing slides indicate they plan to continue it. AMD took an NVIDIA tactic and is using it to slap Intel...meanwhile, AMD is baiting NVIDIA to use that same tactic in order to launch faster GPUs after NVIDIA pushes out the Ti models (or maybe super, though rumors indicate Ti).
  • ikjadoon - Wednesday, October 7, 2020 - link

    People asking Intel for a 10nm CPU with a TDP over 28 W: "Are you not ashamed of yourself? Are you not embarrassed? This is really embarrassing."

    As if NVIDIA only released mobile GPUs for the 10-series, the 20-series, and the 30-series, and desktop users kept buying re-heated GTX 980 Ti GPUs.

    It's uncanny how Intel's CPU division feels closer and closer to AMD's GPU division. "Marketing. Mainstream movie/game gimmicks. More marketing. Drips of unverified, unusable details."

    At least for reviewers, though, hopefully you'll have more time between launches to dive deep.
  • Jorgp2 - Wednesday, October 7, 2020 - link

    >People asking Intel for a 10nm CPU with a TDP over 28 W: "Are you not ashamed of yourself? Are you not embarrassed? This is really embarrassing."

    Imagine being such a shill that you have to bend reality to fit your world view.

    It's a 12-28w configurable TDP, just like AMD has a 10w-25w configurable TDP. And the AMD part runs at 18w when set to a 15w TDP, throttling hard all the while.
  • schujj07 - Wednesday, October 7, 2020 - link

    The OP is wanting a 10nm CPU with a 35/45/65...etc... TDP. The only 10nm CPUs from Intel are in the 12-28W range until server Ice Lake is released. However, that doesn't fix the lack of H branded mobile or desktop 10nm CPUs.
  • TheinsanegamerN - Wednesday, October 7, 2020 - link

    A) WOOOOOSH. Jesus christ man.

    B) Funny you focus on AMD throttling when Intel needs that 28 watt to compete with the 15W "throttling" AMD part, and tools exist to up that TDP to whatever your heart (and heatsink) desire.
  • Jorgp2 - Wednesday, October 7, 2020 - link

    >Funny you focus on AMD throttling when Intel needs that 28 watt to compete with the 15W "throttling" AMD part, and tools exist to up that TDP to whatever your heart (and heatsink) desire

    It's funny, because it manages to compete with 8 core AMD 15w and 45w parts in various tests.
  • Spunjji - Saturday, October 10, 2020 - link

    "Imagine being such a shill that you have to bend reality to fit your world view."

    Projection much? They said 10nm CPU with a TDP *over* 28W. If you're aware of such a thing (not counting the 50W boost, lol) then please do let us know.
  • eek2121 - Tuesday, October 13, 2020 - link

    To add to this. TDP != power. There is not a single 15W chip that uses less than 30W of power for the duration of the benchmark suit. AMD included.
  • Spunjji - Saturday, October 10, 2020 - link

    "As if NVIDIA only released mobile GPUs for the 10-series, the 20-series, and the 30-series, and desktop users kept buying re-heated GTX 980 Ti GPUs."

    That's actually an excellent way of putting it.
  • eek2121 - Tuesday, October 13, 2020 - link

    agreed.
  • eek2121 - Tuesday, October 13, 2020 - link

    Tiger Lake H is releasing in March, and (not so) amazingly, it will render anything below the Ryzen 5000 series, desktop or otherwise, obsolete. The issue is that now Intel has to contend with the 5000 series...however again they are in lockstep and neither has been able to pull away.
  • Pneumothorax - Wednesday, October 7, 2020 - link

    Wow another 14nm chip. I’m so excited. Intel in 2020 is IBM in 1986. While everyone was releasing 386 computers, IBM kept on releasing 286....
  • edzieba - Wednesday, October 7, 2020 - link

    Who cares what process node it's fabbed on if it delivers the performance you need for your application in the power budget you have available?
  • baka_toroi - Wednesday, October 7, 2020 - link

    No corporation can bend the law of physics. Process node is 100% relevant to a CPU's performance.
  • JayNor - Wednesday, October 7, 2020 - link

    smaller nodes can be lower power. Higher clock rates, not so predictable. AMD's Norrod says they may need to go to 3D fabrication to improve performance ... Rice HPC conference, 2019.
  • TheinsanegamerN - Wednesday, October 7, 2020 - link

    The 14nm node is obsolete at this point and will hold backw hatever proposed improvements Intel has made. Or will require a nuclear reactor style heatsink.
  • Spunjji - Saturday, October 10, 2020 - link

    "Who cares what process node it's fabbed on if it delivers the performance you need for your application in the power budget you have available"

    Well that's the problem - it's unlikely to do that; at least in comparison with competing products. If they manage to clock the cores at a rate that will provide competitive performance then it's unlikely to be at a reasonable power level. This is assuming the cores will need more power than they do on 10SF (~50W for 4 cores at 4.8Ghz boost); if they somehow don't then WOW that node is still bad.

    Cost is also a factor - those chunky Cove cores without the scaling benefits of 10nm are going to result in a fairly large die. They'll surely yield well on 14nm, but a large die is a large die, and they're still competing with their own server products for those wafers. The upshot is that even if they want to, they can't trivially sink a bunch of 14nm wafers into this and eat the cost as they might otherwise be tempted to do.

    There's a reason Intel cared a lot about process node, right up until they couldn't execute. They built their entire organisation around it.
  • azfacea - Wednesday, October 7, 2020 - link

    putting what rumors to where ?

    intel's next product will be another 14nm recycle, short on core counts, short on pcie, long on power consumption, long on security holes, short on mobo compatibility, long on price, short on supply. why did they need a new product for this ??
  • yeeeeman - Wednesday, October 7, 2020 - link

    Unfortunately that is the truth...
    Intel should have scrapped rocket lake and move straight to alder lake, 10nm, pcie 5, ddr5, xe graphics, big little, etc. Instead they are refreshing the refreshes and then when they actually have new hardware (as in rocket lake) they are still making it on 14nm.
  • TheinsanegamerN - Wednesday, October 7, 2020 - link

    PCIe 5 isnt ready yet, DDR5 is just barely starting production, big.little x86 is unproven, ece.

    But yes Intel should just abaondon their next gen, piss off their SIs and OEMs, and wait for the new tech to be ready. Worked great with 10nm right?
  • JayNor - Wednesday, October 7, 2020 - link

    They just opened fab42 for 10nm, but only have 3 10nm fabs for now. They have plenty of 14nm capacity. Rocket Lake will be a desktop chip with avx512, pcie4, 5GHz turbo and a nice core IPC update.

    10nm Tiger Lake-H is rumored to be coming also.
  • Spunjji - Saturday, October 10, 2020 - link

    "They have plenty of 14nm capacity"

    Citation needed on that one - last I knew they were capacity constrained.
  • nandnandnand - Wednesday, October 7, 2020 - link

    What do you mean by short on PCIe? Less lanes?

    If the backported Willow Cove improves IPC enough, they might be able to catch up to Zen 3, or get close enough and drop prices to undercut.

    It would be an improvement on Comet Lake, and likely work on the same motherboards. If they don't release Rocket Lake, it's Comet Lake vs. Zen 2/3 for many months which would be a disaster.
  • brucethemoose - Wednesday, October 7, 2020 - link

    I'm more interested in the lack of professionalism in this post.

    Careening between topics and throwing out buzzwords? Posting on Medium? The weird font change y'all got?

    I appreciate candidness and all, but this doesn't sound like something that should come from the Intel VP/GM for Desktop/Workstation/Gaming.
  • Spunjji - Saturday, October 10, 2020 - link

    It has 4 claps, too. 😬
  • Slash3 - Wednesday, October 7, 2020 - link

    I can't wait to read about their next big Xe DG2 feature, no doubt posted to Geocities and Tumblr as a sparkle GIF diary entry.
  • Spunjji - Saturday, October 10, 2020 - link

    🤣
  • morgandg - Wednesday, October 7, 2020 - link

    I invented SKT LGA 1200! Go Rocketlake!!
  • 1_rick - Wednesday, October 7, 2020 - link

    No link to the post on Medium?
  • repoman27 - Wednesday, October 7, 2020 - link

    https://medium.com/intel-tech/intels-commitment-to...
  • 1_rick - Thursday, October 8, 2020 - link

    Thanks!
  • lmcd - Wednesday, October 7, 2020 - link

    Given that Rocket Lake will likely launch before Microsoft's 2021H1 update, that'll give Intel a few months to tout its Hyper-V advantage and get a few developer desktops sold.

    After that, Intel will have to finally strip its iGPUs from high-end consumer to match the core counts AMD puts out. It's exhausting watching Intel lose mindshare to AMD while doubling their own die size with an often-unused iGPU.
  • Unashamed_unoriginal_username_x86 - Wednesday, October 7, 2020 - link

    Q5 2020, get hype Gamers
  • outsideloop - Wednesday, October 7, 2020 - link

    Of course it will be released in Q1 2021! Just like Intel told us two months ago that Ice Lake-SP will be released by Q4 2020:

    https://wccftech.com/intel-unveils-ice-lake-sp-xeo...

    I would never doubt anything they say. Such honest people over there at the Intel corporation. Fantastic.
  • lmcd - Wednesday, October 7, 2020 - link

    Why would you doubt an Intel 14nm product's release date?
  • repoman27 - Wednesday, October 7, 2020 - link

    You'd think they'd at least have that under control, but look at how late Comet Lake-S was.
  • lmcd - Thursday, October 8, 2020 - link

    Probably the WiFi 6 integration, which I personally assume needed backported from planned 10nm. Why that impacted desktop is to question the last 7 years of Intel strategy.

    By contrast, this round there should only be Willow Cove integration + PCIe 4.0. Less teams involved in that.
  • lmcd - Thursday, October 8, 2020 - link

    Also worth noting that the 10 core getting cut should probably help yields a ton. Cutting the GPU from anything above 6 cores would probably also help the yields, but I'm not a product manager used to screaming quicksync in every Intel benchmark demonstration over the last 8 years
  • repoman27 - Thursday, October 8, 2020 - link

    No. Wi-Fi 6 was integrated into the PCH, not the CPU. As was 2.5GbE, which was what they reportedly screwed up. Regardless, they burned through several early steppings of the CML-S 10+2 die, so clearly there were issues there.

    The sentence you started with, “By contrast,” made my head explode. I don’t even...

    And at this point Intel’s 14nm yields are probably higher than any other process node in history. I mean, how can they not be? And we’re taking about relatively tiny chips here. CML-S 10+2 is only twice the size of Apple’s A13 which is on TSMC 7nm!
  • Spunjji - Saturday, October 10, 2020 - link

    "only Willow Cove integration" really was a reach, and particularly funny in the context of blaming WiFi integration for a CPU being late. 😅
  • Spunjji - Saturday, October 10, 2020 - link

    Because they've never made Cove cores on 14nm before..?
  • hansip87 - Thursday, October 8, 2020 - link

    One thing that we don't want is to have Intel lagging anymore, otherwise AMD will push the price sky high because of demand. Simple economic law.
  • Spunjji - Saturday, October 10, 2020 - link

    Truth.
  • alufan - Thursday, October 8, 2020 - link

    Curious does Anandtech even know Ryzen 3 is out Today or that Nvidia 30 series launched a while back?
    Has Intel bought the site all I can see recently are articles focusing on Intel products.

    Anand must be so sad over what has happened to his site
  • lmcd - Thursday, October 8, 2020 - link

    You clearly can't even shill correctly, given that Ryzen 3 is not out today. Zen 3 is out, at least, being revealed, but not released, but not yet revealed either!

    Which sounds like news for when it's revealed.
  • alufan - Thursday, October 8, 2020 - link

    Idiot and what word do we shorten to get Zen?
    R Y Z E N ooohhhh look numpty talk about shill and Today the chip known as Ryzen 3 (or 3rd generation to make it easier for you) will be launched at 9am in the states.
    Do you see any links etc on this site to it? No? Hence my comment.
  • Qasar - Thursday, October 8, 2020 - link

    um, cause the launch isnt until 9 am pdt. and due to the fires in california, the rtx 3000 series review has been delayed. dr cutress has mentioned this in other threads. technically it is zen3, and either ryzen 4000 or 5000.
  • PopinFRESH007 - Sunday, October 11, 2020 - link

    For someone calling someone else an idiot, you sure are the most ignorant AMD fanboy I've seen in a while. Ryzen is a CPU product line. Zen is a CPU architecture. They are not the same thing. Ryzen 3 is also not the same as 3rd generation Ryzen, it is a low end model line within the Ryzen CPU product family.

    Ryzen 3 is not equal to 3rd gen Ryzen
    3rd gen Ryzen is not equal to 3rd gen Zen

    3rd gen Ryzen is a 2nd gen Zen CPU and there were several Ryzen 3 models of 3rd gen Ryzen. Also, wow look at this the 4th gen Ryzen 5000 series didn't actually "launch" at 9am on the 8th. Its crazy I know. Why would Anandtech post an article with nothing but speculation hours before a press event or product announcement? Maybe you should go check out wccftech, that is probably more your speed.
  • Spunjji - Saturday, October 10, 2020 - link

    Advice on shilling from lmcd should not be taken lightly
  • Spunjji - Saturday, October 10, 2020 - link

    What is this nonsense? They covered the Zen 3 release, and they're not in control of wildfires.
  • Warrior24_7 - Tuesday, October 13, 2020 - link

    Good, can’t wait! I NEVER buy AMD! Their video cards are horrible, and Intel STILL has the fastest gaming CPUs on the market! This is with old tech. This is a fact. Clocks beat cores. Intel wins, and is still winning. PCIe 4.0 will bring their board where they should be.
  • Spunjji - Thursday, October 22, 2020 - link

    Can't tell if fanboy or mocking fanboys 🤔
  • Machinus - Wednesday, October 14, 2020 - link

    I think the honest, actual designation for this product is 14nm++++++. Right?

    Everyone is going to have Zen3 by then, so no one will buy these.
  • Spunjji - Thursday, October 22, 2020 - link

    Don't underestimate the power of brand loyalty!

Log in

Don't have an account? Sign up now