AMD Zen 6 - Should it be ALL X3D?

แชร์
ฝัง
  • เผยแพร่เมื่อ 28 พ.ย. 2024

ความคิดเห็น • 54

  • @notificationguys5128
    @notificationguys5128 วันที่ผ่านมา +39

    God.. I really missed the Ryzen 3 release

    • @morpheus_9
      @morpheus_9 วันที่ผ่านมา +10

      AM4 is basically Ryzen 3 at this point

    • @GreyDeathVaccine
      @GreyDeathVaccine วันที่ผ่านมา

      @@morpheus_9 Nope. Ryzen 9 5900X is still great CPU.

    • @fVNzO
      @fVNzO วันที่ผ่านมา +2

      So you agree

  • @kotztotz3530
    @kotztotz3530 วันที่ผ่านมา +20

    x3D has such a good name that I don't think they'd drop x3D off the name of the CPUs.

    • @Aurora12488
      @Aurora12488 21 ชั่วโมงที่ผ่านมา +4

      Yup, it's basically a double-entendre; it's literally describing the 3D-stacked cache, but "3D" is very evocative of gaming and 3D-rendering, which is what the cache is great for. Having the regular chip, the X-variant being performance-binned, and the X3D be performance-binned and gaming-optimized is the best marketing anyone could have hoped for IMO.

  • @markwilson7013
    @markwilson7013 วันที่ผ่านมา +5

    What non-X3D revenue? 😅
    I think the sales spoke volumes, and non X should absolutely become the non X3D (essentially killing the 'X' models).

    • @Aurora12488
      @Aurora12488 23 ชั่วโมงที่ผ่านมา +2

      Eh, maybe. The non-X chips are clocked lower with lower TDP, which is great for AMD since they don't have to throw away underperforming chips, but is not great for maximum productivity.
      But 3D V-Cache doesn't really help meaningfully for most non-game use cases; it's really only simulations that operate on the same small bit of data over and over as fast as physically possible, with highly random access. Most productivity use-cases have much more predictable memory access patterns and process more data than can sit in 100MB of cache, so AMD consolidating the X3D and X markets would be cause them to have higher overhead with including an additional cache die per CPU when many non-gaming customers are just wanting the fastest chip.
      So having non-X, X, and X3D is actually a pretty damn good tiering system for them. A basic-tier chip, a binned performance-tier chip, and a gaming-tier chip.

    • @markwilson7013
      @markwilson7013 22 ชั่วโมงที่ผ่านมา

      @Aurora12488
      AMDs binning and defect rate will determine what makes the most financial for them, but the stack is getting a bit busy for general or gaming users. I wouldn't have thought productivity users won't be particularly interested in low core count chips (and non-x is overclockable if needed) and modern X3D chips aren't frequency limited, so if you have a slice of users that would benefit you'd probably be better to release a 'Pro' line and not leave resellers with a bunch of stock they can't sell 😅

    • @Aurora12488
      @Aurora12488 22 ชั่วโมงที่ผ่านมา +1

      ​@@markwilson7013 I agree they should consider dropping chips like the 9600X in favor of a 9600X3D; there's essentially no market anymore for low-core, high performance, low cache. But the problem I'm saying is that gaming doesn't cleanly slot into a Pro category. X3D cache is legitimately very overkill for non-gaming, but proven to be critical for gaming. There legitimately are three pretty notable target demographics, and eating the overhead of including a cache die to consolidate seems like a bad idea to me for both AMD and, when the overhead trickles down, the consumer.
      Also, AMD doesn't produce all the chips and then saturate retailers up-front, but will continuously manufacture more and then restock over time. So even if a certain chip isn't moving well, I don't think that'll burn the retailers as they won't be sitting on that much stock anyway, and over the course of the year and with various discounts and bundles all stock will eventually sell through. So as long as retailers have stock of the things that are in demand, I don't think there's too much issue for them having some low-volume stock.
      Overclocking is also very inconsistent; if the chip could have performed better, they would have binned it there, so you're playing the silicon lottery to see how close to the threshold you could overclock the basic chip. And most people who aren't tech enthusiasts just kind of want things as they are out-of-the-box without any tinkering. So I do think the basic-tier can't really realistically do double-duty unfortunately.
      As long as the marketing is clear, they've got a great thing going. Their basic lineup for value-focused customers, the X lineup for performance-tuned beasts, and the X3D lineup as the gaming-optimized variant of X. And X3D being the gaming chip, where "3D" gives heavy gaming vibes, is a crazy-good naming scheme.

  • @3polygons
    @3polygons วันที่ผ่านมา +4

    Yep, I would vote for a super energy efficient, cold, non-x series to be released first (a lot of users want that), typically cheaper as well, and then, all the skus having X3D IF - and only if- the X3Ds do really perform in all cases as well as a non 3D version in professional applications.

    • @guyfawkes0511
      @guyfawkes0511 วันที่ผ่านมา

      If the 9800X3D is anything to go by, then they have resolved that problem already. The 9900X3D variants should be the best example for this, though, so we will have to see.

  • @zzxd7720
    @zzxd7720 วันที่ผ่านมา +13

    Will they release the Zen5C for desktop?
    It could be something like 8 cores full Zen5 with x3d plus 16 cores of Zen5c without x3d.
    24 cores, 48 threads of pure bruteforce.

    • @Traumatree
      @Traumatree วันที่ผ่านมา +2

      p-core is for bruteforce, all the tiny cores are mostly useless for actual work.

    • @roanbrand7358
      @roanbrand7358 วันที่ผ่านมา +1

      Few extra donkeys vs v8 engine

    • @berke__
      @berke__ วันที่ผ่านมา +1

      Maybe but TR5 socket

    • @asm_nop
      @asm_nop วันที่ผ่านมา +4

      ​​@@TraumatreeThat's true for Intel's E cores, but not necessarily AMD's C cores. It's the same core architecture, just optimized into a smaller space with much less cache and slightly lower clocks to keep power density under control. As long as the workloads aren't bottlenecked by small cache size or shared memory bandwidth, they'll have very similar performance to the full-fat Zen cores.

    • @parm2-x7h
      @parm2-x7h 17 ชั่วโมงที่ผ่านมา

      @@Traumatree , bud Amd Ecores are 90% of pcores , they are identical to Pcores with less l3 cache , less clockspeed , but uses 50% less die space .
      intel Ecores are totally different cpu cores and are shit .

  • @absolute___zero
    @absolute___zero วันที่ผ่านมา +4

    They don't put cache everywhere because it is not cheap. Every bit of information requires 6 transistors + control logic and bus lines. So roughly speaking 1Kb of cache will need 50000 transistors, 1Mb will require 50 million transistors, 32 Mb will be 1.5 billion transistors. The CPU die is 10 billion transistors, so 32mb L2 V-cache will have to increase the MSRP by 15%. , gamers are not exactly whealty individuals, but TSMC charges you per transistor.

    • @kevinerbs2778
      @kevinerbs2778 วันที่ผ่านมา +3

      This won't matter as later & later architectures will be seeing increases in all loads from the add 3D cache stacks. Zen 5 already shows that its getting some improvements in multi-threaded loads with 3D Vache were Zen4 did not.

    • @Aurora12488
      @Aurora12488 23 ชั่วโมงที่ผ่านมา +3

      ​@@kevinerbs2778 Cache is only useful if you're needing to frequently access that cached data at the drop of a hat. Gaming's a bit of an outlier that requires essentially thrashing on the same medium-sized bundle of data over and over as fast as physically possible with highly random access.
      If you think about other productivity uses, they're much more about doing bursts of processing of a decently large amount of data, with relatively well-formed access patterns that pre-fetching can help with.
      While more cache is always appreciated, I don't think 3D V-Cache is going to suddenly become the standard. There's overhead in including it, for both AMD and the consumer. While with the compute dies there's some x86 cruft we carry along for simplicity's sake, 3D V-Cache is an entirely separate die that is much more easy to selectively include in your product lineup.

  • @pkpnyt4711
    @pkpnyt4711 วันที่ผ่านมา +1

    It's not gonna happen. It's more expensive to produce. Think about it Zen 5 used 3NM nodes, and the yield rates for those in the beginning were between 55-60%. 2NM, which Zen 6 will be using, right now, is having yield rate issues at tsmc.The process itself is more expensive and is more difficult to produce. They need to produce and maximize wafer yields by producing easier chips to manufacture. This is where the non X3D chips fall in, specifically the lower end.

  • @3polygons
    @3polygons วันที่ผ่านมา +1

    I saw it as a segmentation of the market. Professional and mainstream (no gamers) users market first, then gamers, and probably they knew they were going to beat anything on the market for that.

  • @oappi4686
    @oappi4686 วันที่ผ่านมา

    I think current line up was pretty good. What we got first was basically all purpose pc:s, and now we are geting ones that are for gamers. Only thing I would change is to have 9550E and 9500E or something which have absolutely best efficiency with one ccx. one full ccx and one that has some cores disabled and with very low consumption... something like 15-20W(preferably adjustable to be even lower) and ecc support. kinda like their embedded series, but would have full amount of pcie lanes of AM5 platform. I would buy one for secondary pc that I could leave on all the time to do over night tasks or as general lab pc.
    I would really prefer if they announced whole line up at once with pricing, even if they just ship part of it at the begining and rest comes after 6 months.

  • @KAL-589
    @KAL-589 วันที่ผ่านมา +1

    I want the return of the FX moniker.

    • @gagarin777
      @gagarin777 วันที่ผ่านมา +1

      and I want DFI to return with fun overclocking oriented motherboards with badass radiators.

  • @mondodimotori
    @mondodimotori วันที่ผ่านมา

    At this point, if there's a relevant loss in production performances from installing 3d v-cache, they should drop everything else and just put 3d chips on the market.

  • @spacescatatford
    @spacescatatford วันที่ผ่านมา

    I have a Ryzen 9 7950X machine. When the Ryzen 9 9950X3D comes out, I'm selling RAM, Motherboard, SSD, and the 7950X processor, all in one unit. I'll have all the bugs worked out and it will be a maximized machine. Someone is going to benefit from my knowledge. I suspect X3D will be the future.

  • @TheHangarHobbit
    @TheHangarHobbit วันที่ผ่านมา

    Well they don't really need Ryzen 3 as AM4 will be supported until 2026 so AM4 will be the Ryzen 3 ultra budget baller until then and frankly with 5700x non 3D selling for just $134 last week on Amazon and the 5600 selling for $105? There is no way they can price AM5 that low and still make money so it makes sense.

  • @Mageoftheyear
    @Mageoftheyear วันที่ผ่านมา +2

    Off topic, but I was really happy to hear James saying he had a visceral reaction to the statement that the Radeon team might want to play it safe to avoid engaging Nvidia at the high-end. I can't wait to see AMD reach for the crown with RDNA5!

    • @MooresLawIsDead
      @MooresLawIsDead  วันที่ผ่านมา +3

      No, he meant he thinks they'll be aggressive with RDNA 4...

    • @Mageoftheyear
      @Mageoftheyear วันที่ผ่านมา

      @MooresLawIsDead Ah I understand he was talking about RDNA 4 - I phrased my comment poorly. I meant that in addition to their aggressive intent with RDNA 4 I am glad that that undaunted spirit of competitiveness will be there when AMD competes at the high-end again with RDNA 5.

    • @Aurora12488
      @Aurora12488 23 ชั่วโมงที่ผ่านมา +1

      ​@@Mageoftheyear You're still misunderstanding a bit. He said, when AMD's in it, they're in it to win it. They won't be shy about being in-it-to-win-it in the low-/mid-range with RDNA4, and when they *do* have a competitive product at the high end later they'll also go for the throat, but we'll need to wait and see whether that is RDNA5/UDNA or when. He thinks they'll strike when the time's right, and they'll be working hard to compete in that market in the long-term, but there wasn't any sort of expectation set that the next gen will be the right time.

    • @Mageoftheyear
      @Mageoftheyear 18 ชั่วโมงที่ผ่านมา

      @@Aurora12488 I'm not misunderstanding anything. I'm excited to hear that internally AMD do not intend to pull their punches out of fear of wakening the green beast.
      RDNA 4 can't compete at the high end - so what I (and maybe this isn't you) am excited about is what that attitude will do for AMD when they reach for the crown.

    • @Aurora12488
      @Aurora12488 11 ชั่วโมงที่ผ่านมา

      @@Mageoftheyear Agreed, I was just correcting the part of you calling out specifically *RDNA 5* as the timing. Unfortunately it's not a guarantee it'll only be one generation before they're ready to compete head-to-head on the high-end consumer parts. It might take some growing pains when unifying their RDNA and CDNA architectures into UDNA, especially if they're going to use that as an opportunity to take some new design risks. But I look forward to them going for the crown eventually!

  • @mysticknight487
    @mysticknight487 18 ชั่วโมงที่ผ่านมา

    Only if the Zen 6 9950X3D was v cache on both ccds

  • @ImxxFuZe
    @ImxxFuZe วันที่ผ่านมา

    Yes. It should all be X3D and XT. The hardware market is in desperate need of simplification.

    • @Aurora12488
      @Aurora12488 21 ชั่วโมงที่ผ่านมา

      Not at the cost of, well, cost. The extra cache die isn't free, at all. And not binning your chips to have high-performance and low-performance ones is either losing a bunch of yield or leaving performance up to the silicon lottery.
      The consumer can live with a *little bit* of complexity, if it makes things so much healthier overall. Three tiers is okay. We have the basic-tier with basic performance, the X-tier with binned high performance, and the X3D-tier with the binned high performance but "optimized for gaming". Consolidating to X3D and needing to produce a bunch more cache dies that are only really useful for gaming, and passing that overhead into the price the consumer pays, is not at all a good tradeoff for consolidating the line-up.
      There's no free lunch. As long as people do the *smallest* bit of reading, even just the marketing materials on the box, this really is a pretty good setup for now.

  • @Selloutsatirist
    @Selloutsatirist วันที่ผ่านมา

    11980: 8 Zen6 + 16 Zen6c
    11970: 8 Zen6 vCache + 8 Zen6 vCache
    11950: 8 Zen6 + 8 Zen6 vCache
    11900: 8 Zen6 + 8 Zen6
    11880: 6 Zen6 + 12 Zen6c
    11850: 8 Zen6 vCache
    11800: 8 Zen6
    11650: 6 Zen6 vCache
    11600: 6 Zen6

    • @morpheus_9
      @morpheus_9 วันที่ผ่านมา +1

      Too complicated
      It should be:
      11960X: 8 Zen 6 and 16 Zen 6C
      11950X3D: 16 zen 6 vcache
      11950X: 16 zen 6
      11900X: 12 zen 6
      11800X3D: 8 zen 6 vcache
      11700X: 8 zen 6
      11600X: 6 zen 6

    • @Selloutsatirist
      @Selloutsatirist วันที่ผ่านมา

      @@morpheus_9 You'll need at least one SKU for defective 16-core Zen6C, most likely using a 12 or 14 Zen6C.

  • @1984Captive
    @1984Captive วันที่ผ่านมา

    Bad bin chips and top bin chips take time. No way around it. But x3d should all be released at launch. The issue is pro/cons of hte approach and stock price.

  • @shammyh
    @shammyh วันที่ผ่านมา +3

    Is that a serious question? Because they can charge more for
    the X3D versions...

    • @Aurora12488
      @Aurora12488 21 ชั่วโมงที่ผ่านมา

      And X3D CPUs cost more to make. The cache die isn't a trivial drop-in extra by any means. I'd be surprised if the relative margin of the X3D chips is actually notably higher.

  • @forog1
    @forog1 18 ชั่วโมงที่ผ่านมา

    If the x3d chips can clock just as well as the X counter parts AMD should just kill the X lineup and make the non X models the king of efficiency and if the consumers still want a bit more they can still take it upon themselves to unlock their power limits from say 65w to 105-120w as a more budget option from the x3d if cache is now what you need. X models are quite pointless as im seeing it going forward.

  • @cicio7777
    @cicio7777 วันที่ผ่านมา

    The real question is who is foolish enough to buy a non 3d cache CPU? Another question is does AMD realise that releasing a zen 6 3d cache cpu with more cores than 16 even if only 24 will guarantee a buying frenzy in amd ryzen zen6 cpus?

    • @Aurora12488
      @Aurora12488 21 ชั่วโมงที่ผ่านมา

      Someone who doesn't play games. A bunch of L3 cache is actually not terribly useful for other tasks. It's best for having a handful of data that you need to essentially thrash on and randomly access and mutate as quickly as possible. That's primarily just gaming. For burstier, throughput-centric work the non-3D amount of cache included in the main CCD is 95% effective. And adding a whole extra cache die is not going to be just a 5% difference in manufacturing overhead, so there's going to be a notable price difference.

  • @mattpulliam4494
    @mattpulliam4494 วันที่ผ่านมา +2

    Ideal yeah but can't be cause it would cause issues with world record overclock chase

  • @flimermithrandir
    @flimermithrandir วันที่ผ่านมา

    Look... i would love to see both CCDs with V Cache as well but i dont like the Price Hick that would come with it. I dont plan to pay 1000 Dollars for an 7950X3D. And they would sell it for that.
    They just need to get their Shit together and Fix the Scheduling. Or whatever the Issue is. 2 Times 8 Cores with 3D Cache would not fix everything for Gamers anyways. Some Apps would benefit even then. I have no Doubt about it. But its not an Shiny "all Issues are now gone" Trump Card. No. They need to fix it in Programming. And i doubt its that hard. Basically (VERY basically i know but i am sure they can figure it out... without Issues) its like this... "If Game is loading then use 3DVcache". Done. Nothing else matters. Just Done. No i am not an Programmer. This is like the only Thing i remember from Programming (back from the C64 Days). But really... its that easy. Again... its not that easy really i am sure of but i cant see how its taking them 4 Years (5800X3D launched in like November 2020. Ok something something... it takes them 1 Year to fix it. Fine. Thats terrible long but ok. Whatever. 4 YEARS? Not for Years but 4 full YEARS and they still struggle with this Stuff? BS.
    Just Fix it. And sure in an Ideal World i would love to see an 2CCD X3D too. No Issues with that. But i know that this would not fix or magically increase Gaming Performance by 15% now. The Issues with the 3D Cache would still be there if the Game needs to go through the I/O Die. THIS is the Main Issue. And that doesnt magically go away because... You would maybe get 3% better Performance sometimes and still loose some % in other Games.

  • @ChiquitaSpeaks
    @ChiquitaSpeaks วันที่ผ่านมา

    Shutup Moore’s law lol

  • @djnes2k7
    @djnes2k7 วันที่ผ่านมา +1

    What kind of a dumb question is this? V cache has more heat and lower clock speeds, and most games don’t use more than 8 cores. It’s literally a waste. Especially if the gpu isnt going to outpace it. So the price would rise, you’d need better cooling clock speed won’t be lower…. All for what? I think we are losing the plot in gaming and are just racing toward how much can we spend.

  • @johndoh5182
    @johndoh5182 วันที่ผ่านมา +3

    No, it should not. These packaging and layering technologies cost money. I guess for the YTr that has everything handed to them I can see how they might think so but when most the world can barely afford an entry computer, companies need to be respectful of what a customer can afford.

    • @GreyDeathVaccine
      @GreyDeathVaccine วันที่ผ่านมา

      Agree. Host sometimes says bollocks.