Slashdot Mirror


Could AMD's Upcoming EPYC 'Rome' Server Processors Feature Up To 162 PCIe Lanes? (tomshardware.com)

jwhyche (Slashdot reader #6,192) tipped us off to some interesting speculation about AMD's upcoming Zen 2-based EPYC Rome server processors. "The new Epyc processor would be Gen 4 PCIe where Intel is still using Gen 3. Gen 4 PCIe features twice the bandwidth of the older Gen 3 specification."

And now Tom's Hardware reports: While AMD has said that a single EPYC Rome processor could deliver up to 128 PCIe lanes, the company hasn't stated how many lanes two processors could deliver in a dual-socket server. According to ServeTheHome.com, there's a distinct possibility EPYC could feature up to 162 PCIe 4.0 lanes in a dual-socket configuration, which is 82 more lanes than Intel's dual-socket Cascade Lake Xeon servers. That even beats Intel's latest 56-core 112-thread Platinum 9200-series processors, which expose 80 PCIe lanes per dual-socket server.

Patrick Kennedy at ServeTheHome, a publication focused on high-performance computing, and RetiredEngineer on Twitter have both concluded that two Rome CPUs could support 160 PCIe 4.0 lanes. Kennedy even expects there will be an additional PCIe lane per CPU (meaning 129 in a single socket), bringing the total number of lanes in a dual-socket server up to 162, but with the caveat that this additional lane per socket could only be used for the baseboard management controller (or BMC), a vital component of server motherboards... If @RetiredEngineer and ServeTheHome did their math correctly, then Intel has even more serious competition than AMD has let on.

107 comments

  1. I don't know. Could it? by Anonymous Coward · · Score: 0

    COULD IT?

    captcha: salesman

    1. Re: I don't know. Could it? by Anonymous Coward · · Score: 0

      Could it? Who cares... Watching the processor wars is like a slow boil. Maybe AMD something maybe someone else

  2. I don't see why not! by Anonymous Coward · · Score: 0

    Of course they could. I don't see a problem.

    1. Re: I don't see why not! by Anonymous Coward · · Score: 0

      An optimist! I bet you like toot your own horn

  3. Re:Hillary for PRISON! by Anonymous Coward · · Score: 0

    Drown the witch!

  4. Re:Hillary for PRISON! by Anonymous Coward · · Score: 0

    Enjoy your fantasies but Trump is a goner

  5. Re: Hillary for PRISON! by Anonymous Coward · · Score: 0

    Wait for it, wait for it

  6. No. by Gravis+Zero · · Score: 5, Funny

    It could not, would not, on a boat.
    It will not, will not, compute your float.
    It will not have them in the rain.
    It will not have them on a train.
    Not in the dark! Not in a tree!
    Not in a car! Listen to AMD!
    It will not run your Firefox.
    It will not run programs on your box.
    It will not be inside your house.
    It will not 'shop you with Mickey Mouse.
    It does not have them here or there.
    It does not have them anywhere!

    --
    Anons need not reply. Questions end with a question mark.
    1. Re:No. by Anonymous Coward · · Score: 0

      Not in a car! Listen to AMD!

      Of course I won't listen to AMD inside my car.

    2. Re:No. by juegosdeautos · · Score: 1

      Thank you for your post, I look for such article along time, today i find it finally. this post give me lots of advise it is very useful for me

    3. Re:No. by Anonymous Coward · · Score: 0

      Anyway, our IT janitor from San Jose would need at least 640 PCIe lanes with all the crap he does. Plus, if he needs to fix his computer, tell me how the hell he could fit his chubby fingers into the lanes with only a width of 162?

    4. Re: No. by Anonymous Coward · · Score: 0

      That's the answer to all Slashdot questions. And remember the "news" never trolls. Trolls are people who don't like fake news.

    5. Re:No. by Anonymous Coward · · Score: 0

      such article to be found on internet post! :) :) :)

  7. Intel follows. by Anonymous Coward · · Score: 0

    As the saying goes, "when in ROME..."

  8. what about quad socket?? with the faster links? by Joe_Dragon · · Score: 1

    what about quad socket?? with the faster links?

    1. Re:what about quad socket?? with the faster links? by Anonymous Coward · · Score: 0

      what about quad socket?? with the faster links?

      AMD will probably not be bothering with quad socket. Even dual socket is likely going to be going away eventually. Intel probably won't get there nearly so soon as they aren't able to put so many cores on a socket yet, but likely they too will abandon quad socket boards once they are on their 7nm node.

    2. Re:what about quad socket?? with the faster links? by Anonymous Coward · · Score: 0

      what about quad socket?? with the faster links?

      likely they will abandon quad socket boards once they are on their 7nm node.

      But why??

      I mean, what benefits will they get if they abandon they quad, or even octo- socket?

    3. Re:what about quad socket?? with the faster links? by Anonymous Coward · · Score: 0

      Same reason why Intel ate the lunch of "big iron" companies years ago with their cheap 1/2/4p Xeon systems.

      In general, it's cheaper/easier to scale out with a limited number of cores on a large number of systems/motherboards than it is to just keep adding more sockets/cores to the same board. Yes, you can cut down on core-to-core latency by going for a higher socket count (versus trying to maintain coherence with cores housed on an entirely different board), but there's a limit to the number of cores that really require coherence with one another.

      AMD is going to be selling 64 high-performance x86 cores in a single socket. Do you really need cache coherency between more than 128 cores? Unless you're in the HPC world (which has high-speed interconnects to tackle that challenge), you generally don't. Ask a cloud service provider why they would need more than 128 cores in a single machine. In fact, I expect that cloud providers will want 1p server blades (or something similar). Why get some 8p behemoth in a 4U rack?

    4. Re:what about quad socket?? with the faster links? by Anonymous Coward · · Score: 0

      There is one primary technical reason for not supporting quad socket - Zen (and this does not appear to change with Zen2) only has limited number of IF links to use. That number is three. This situation is similar to Intel where dies have even less links (2 UPI links) on anything other than XCC (3 UPI links).

      The optimal topology is a link (IF/UPI) to every other core complex.
      - In AMD's case, 1 link goes to the other CCX and 2 go to the other package, 1 for each CCX in there. This limits the optimal multi-socket configuration to 2 sockets. Due to the topology of CCX and memory controllers AMD does not want to go to unoptimal topology for a good reason. Whether this will change with Zen2 we will see but I doubt it.
      - This is also the reason Intel has not been competing with glued CPUs - They simply cannot. 10- and 18-core dies are limited to 2 UPI links, only 4 of these dies can be connected up and even then not optimally. 28-core XCC dies have 3 UPI links and they can be glued together although considering the size of these at a cost. This is what Intel is doing now with 48/56-core Cascade Lake AP/Intel Xeon Platinum 9200.

      There is a lot of marketing around this from both sides. Intel's infamous "glue" remark. AMD's talk about dual-socket being the most used and their target. This is all technically true but really just marketing bullshit.

    5. Re:what about quad socket?? with the faster links? by DigiShaman · · Score: 1

      With the advent of "chiplet" (dies) being crammed into a single package, it makes the multi-socket rather moot beyond dual configuration. A dual-socket configuration is already complicated. If you truly need that much CPU resources in a single chassis, it can be more economical to move to a blade configuration. While the blade server would be expensive per CPU, the net cost would be reduced if you factor in density to reduce the number of enclosures you'd otherwise be leasing at a data-center relative to that of standard rack-mounted 1U servers.

      TLDR: if need quad socket configuration, at that point you're probably looking to scale well beyond that anyways.

      --
      Life is not for the lazy.
    6. Re:what about quad socket?? with the faster links? by Junta · · Score: 1

      Note that this was said the second we had dual core processors, and is generally repeated ever so often.

      We now can have a single-socket system that is a 32-way server, and yet there is still push for 8 socket and beyond. Parts of the industry just stubbornly don't want to let go.

      Some of it is adherence to tradition, but at least some of it is due to some memory capacity intensive workflows that will eat every byte of memory you can feed it.

      A strong argument for the relatively few applications that do this is that they've been derelict in moving beyond single system and if they do not migrate to multi-system model, then they will be replaced by a competitor that will.

      Another regretable situation I've seen is some people thinking that's the way to do virtualization. Thankfully this is no longer a majority opinion as far as I can tell.

      --
      XML is like violence. If it doesn't solve the problem, use more.
    7. Re:what about quad socket?? with the faster links? by Junta · · Score: 2

      The question of course is why this would be different *now* than before. From the second we had our first dual core processor, I've been hearing people proclaim the death of multi-socket designs with every core count bump. Yet here we are...

      The blade v. 1u argument is a bit moot. Already dense form factors can double up the density. I've seen proposals for 'dual system' boards to gain some marginal economy of scale for the physical components but not the complexity of having the sockets interconnected.

      Of course, it is fair to say that if you are an application developer with a cpu/memory hungry behavior and you *don't* support multiple-systems... One day a competitor will displace you that will span multiple systems. It's an increasingly rare segment and the vendors are charging ever more for systems to accomodate that appetite. A competitor that allows your customers to do the same thing but with 4x cheaper hardware can have an easy time of it.

      --
      XML is like violence. If it doesn't solve the problem, use more.
    8. Re:what about quad socket?? with the faster links? by UnknowingFool · · Score: 1

      The question I have is who is pushing for quad, hex, or octo socket servers and what is the use case to it. Obviously such systems are vastly more mostly to build. Simple things like power supplies and and cooling are far more complicated. It is probably way cheaper to use multiple servers at that point.

      I can only see only one real use case. For supercomputers or render farms where high CPU density is a necessity. Or for more limited scenarios where you physically can only have one server rack and you need as many CPUs/cores as possible in a small amount of space.

      --
      Well, there's spam egg sausage and spam, that's not got much spam in it.
    9. Re:what about quad socket?? with the faster links? by Highdude702 · · Score: 1

      I would assume Supermicro and the others. It costs more to manufacture therefor higher markup. Why would you let that slip and not try to push it at every chance?

    10. Re:what about quad socket?? with the faster links? by Anonymous Coward · · Score: 0

      Supercomputers and render farms are generally two-socket minded at the moment and a ripe market to go single socket.

      Those have been living in a too-constrained world where they simply had to go to multi-system.

      The drivers of high socket systems are certain enterprise software vendors and certain outdated virtualization shops that view multi-system management as hard enough to be worth the cost and insane risks of big box virtualization.

  9. Re: Trump actually WILL die in prison, lol. by Anonymous Coward · · Score: 0

    We drink to your impeccable incredible health trump!

  10. I don't know, mr viral marketing intern by NicknameUnavailable · · Score: 1

    But those are some really specific numbers for a "could x be the next version?" post.

    1. Re:I don't know, mr viral marketing intern by thegarbz · · Score: 1

      You'll find that "Could x be the next y" is the type of shit that editors come up with when they rewrite user submitted stories for more clickbait.

      As for how specific the numbers are 5 x 16 per chip + speculation from articles on the server control hub that each chip will get a lane dedicated to it rather than having split up one of the group of 16. Even as speculation that makes perfect sense.

  11. Re: Trump actually WILL die in prison, lol. by Anonymous Coward · · Score: 0

    Enjoy your fantasies but Trump is a goner

     

  12. Trump actually WILL die in prison, lol. by Anonymous Coward · · Score: 0

    Enjoy your fantasies but Trump is a goner.

  13. Trump actually WILL die in prison, lol. by Anonymous Coward · · Score: 0

    Enjoy your fantasies but Trump is a goner!

  14. Trump actually WILL die in prison, lol. by Anonymous Coward · · Score: 0

    Enjoy your fantasies but Trump is a prison boner cozy.

  15. Computational Numerology by Jeremy+Erwin · · Score: 2

    from the article

    The Intel Xeon Scalable 8-socket servers are very popular in China because the number 8 has special significance, as does having the largest server.

    See, that's why Intel is not using PCIe 4.0. The same theories regard 8 way servers as "auspicious" also fear the number fourth iteration of PCIe

    1. Re:Computational Numerology by sheramil · · Score: 2

      Are they like Star Trek films, where alternate ones are good and then cursed?

    2. Re: Computational Numerology by Anonymous Coward · · Score: 0

      So much for that thread

    3. Re:Computational Numerology by Anonymous Coward · · Score: 0

      No matter how big their server, the Chinamen are still needle dicks.

    4. Re:Computational Numerology by Anonymous Coward · · Score: 0

      I'd say more like "suspicious" than "auspicious".

  16. Re: Trump actually WILL die in prison, lol. by Anonymous Coward · · Score: 0

    Enjoy your fantasies but our President will be re-elected next year because of morons like yourself !

  17. apk: Hillary for PRISON! GAY NIGGERS FELCHING FUN by Anonymous Coward · · Score: 0

    Im GAYpk and im gay!

  18. Re: Trump actually WILL die in prison, lol. by Anonymous Coward · · Score: 0

    Enjoy your fantasies but Trump is a prison boner holster.

  19. thanks by juegosdeautos · · Score: 0

    This is also a very good post which I really enjoyed reading. It is not everyday that I have the possibility to see something go movies

  20. Re: Trump actually WILL die in prison, lol. by Anonymous Coward · · Score: 0

    I'm from the future. I came to April 7, 2019 just to make this one comment:
    Trump will win the presidency again in 2020.

  21. Re: Trump actually WILL die in prison, lol. by Anonymous Coward · · Score: 0

    On behalf of everyone sane. whatever.

  22. Re: Subarus Are Cars for Lesbians by Anonymous Coward · · Score: 0

    I guess that makes me a lesbian too because I own two WRXs. Idiot.

    -subie

  23. Re: Subarus Are Cars for Lesbians by Anonymous Coward · · Score: 0
  24. Desktop CPU Lanes by war4peace · · Score: 4, Interesting

    I've been waiting for 3 years now for a relatively affordable desktop CPU with enough PCI Express Lanes.
    My current CPU is an Intel i7 6800K, using the X99 motherboard chipset and it has 28 PCI Express Lanes. The 6850K has 40 PCI Express lanes but otherwise brings no performance increase. The next step in the upgrade process is the 6900K which, albeit on an EOL platform, has enough meat to satisfy my requirements... but costs a fortune. As a matter of fact, it costs as much if not more than a 2nd gen Threadripper (2920X), which has 12 cores and 24 threads available, compared to Intel's 8/16. But that requires changing the motherboard as well, and those are pricy too.

    Only the HEDT CPUs have enough PCI Express Lanes, if you have 2x GPUs and a minimum of 2x nVME SSDs. There are regular desktop solutions which allow you to use such a hardware combo, but one GPU will run at 8X, the other at 4x, one SSD will run at 2X and the other would most likely use motherboard-provided PCI Express lanes, reducing the data throughput or providing variable performance. The 9900K from Intel has 16 PCI Express Lanes. The Ryzen 2700X has 16 lanes as well. You need more PCI Express lanes? Tough luck, cough up a couple grands on CPU+motherboard alone.

    --
    ...gis sdrawkcab (usually not responding to ACs; don't bother posting as AC)
    1. Re:Desktop CPU Lanes by I'm+just+joshin · · Score: 1

      What about Threadripper?

    2. Re:Desktop CPU Lanes by I'm+just+joshin · · Score: 1

      Never mind, missed seeing you refer to it. A new MB is in the cards either way. May as well accept that.

    3. Re:Desktop CPU Lanes by Anonymous Coward · · Score: 1

      You can get a TR4 board for $250 right now. Pricey? Fuck that, the cpu is $800-1200 and you might as well get 8 sticks of ram too.

    4. Re:Desktop CPU Lanes by execthts · · Score: 1

      What if I have a boyfriend?

    5. Re:Desktop CPU Lanes by war4peace · · Score: 1

      She builds furniture and does woodworking among other things. So yes, we both have hobbies, I'm learning her crafts and teaching her mine (PC modding, 3D design, PC hardware).

      --
      ...gis sdrawkcab (usually not responding to ACs; don't bother posting as AC)
    6. Re:Desktop CPU Lanes by war4peace · · Score: 1

      No, friend, YOU can get a TR4 board for 250 bucks. I can get the same one for 350 EUR if lucky. Same with CPU, a $1000 CPU is 1200 EUR here easily.

      --
      ...gis sdrawkcab (usually not responding to ACs; don't bother posting as AC)
    7. Re:Desktop CPU Lanes by AmiMoJo · · Score: 2

      I'd jump to Threadripper. Intel goes through sockets way too fast to be worth sticking with. Threadripper is going to have 64 PCIe lanes, compared to only 40 on the 6900K, and the socket will most likely be around for years with decent support for future models of CPU.

      --
      const int one = 65536; (Silvermoon, Texture.cs)
      SJW, n: "Someone I don't like, and by the way I'm a fuckwit" - AC
    8. Re:Desktop CPU Lanes by Anonymous Coward · · Score: 0

      That's not even the best reason to jump to AMD. Things like Spectre, Meltdown and the somewhat oppressive nature of Intel's "Management Engine" + the possibility of on-chip WIFI (a hackers dream, no doubt) and I'm starting to wonder if Intel are losing the plot. As long as Zen 2 is competitive in performance terms I'm jumping to it. But remember the current situation: my 4790K is competitive on a per-core basis with AMD's 2700X! Obviously it takes the advantage on multicore jobs, having twice the number. Still, my point is AMD haven't really been setting the world alight with their processors. The price is decent though, I'll give them that.

    9. Re:Desktop CPU Lanes by Highdude702 · · Score: 1

      I don't get where you only get 16 lanes from on ryzen. its 16 to the slots 1st slot 16x if you use first 2 its 8x/8x and if you use 3 it then drops to 8x/4x/4x and the NVMe slot has 4 dedicated. Plus the chipset also gives you 4x pcie 2.0 which is fine for most things not GPU/NVME ultimately giving yourself 20 PCIe lanes which can be used as 8x/8x/4x if you use a riser for the NVMe. And as another user pointed out, if you change processors with intel there is a VERY good chance you will be buying a new motherboard either way. Also ECC.

    10. Re: Desktop CPU Lanes by madsh · · Score: 1

      I am pretty sure it is not aimed at the desktop marked.... But interesting post...

    11. Re:Desktop CPU Lanes by Anonymous Coward · · Score: 0

      Some girls get a threesome with two other boys. But them boys be gay! And the girl ends up with a vag full of shit; and it ain't hers!

    12. Re:Desktop CPU Lanes by TheDarkMaster · · Score: 1

      I can not even buy a motherboard + Threadripper processor set without being forced to pay double the price charged in the US, And that's when I can find a store that sells this kind of parts.

      --
      Religion: The greatest weapon of mass destruction of all time
    13. Re:Desktop CPU Lanes by bill_mcgonigle · · Score: 2

      One person here is well-versed in microcomputing components and has good earning potential.

      Another spends his time anonymously attempting to troll on /.

      It's easy to tell which has trouble dating women.

      --
      My God, it's Full of Source!
      OUTSIDE_IP=$(dig +short my.ip @outsideip.net)
    14. Re:Desktop CPU Lanes by Anonymous Coward · · Score: 0

      If desktop Ryzen 3000 with Zen 2 retains the PCIe 4 (no guarantee that it will) then even if it keeps the same number of PCIe lanes on the chip (16+4+4) the bandwidth will have doubled. It'd be 48 lanes equivalent of PCIe 3. The one thing you would want to see is an option to split the x16 into x8+x8, as that will be more than adequate for most consumer GPUs.

    15. Re:Desktop CPU Lanes by war4peace · · Score: 1

      I have 2x GPUs in SLI. I also have 2x nVME SSDs which reach 3.4 GB/s transfer rate (each). I also own a 16x PCI Express card which can fit 4 nVME SSDs (not using it for obvious reasons). If I want to use 8x+8x for GPUs and 4x+4x for nVMEs, I need 24 lanes.
      The 6800K already has 28 lanes available (CPU only), so if I upgrade, I want to upgrade to something better from this point of view as well. Ryzen 2 would be a downgrade, unless I go Threadripper.

      --
      ...gis sdrawkcab (usually not responding to ACs; don't bother posting as AC)
    16. Re:Desktop CPU Lanes by Anonymous Coward · · Score: 0

      Yeah, the troll is the one with trouble dating women, because he's already married.

      That's my bet anyway. Marriage will do strange things to otherwise sane folk.

    17. Re:Desktop CPU Lanes by Anonymous Coward · · Score: 0

      Posting as AC to preserve moderation.

      Ryzen 3000 (presumably) products / Zen 2 will be on PCIe 4.0.
      Even if you only get 16 lanes for the primary slots, that's equivalent to 32 lanes of PCIE 3.0.
      Plus you'll have 4 PCIe 4.0 lanes for NVMe. That's equivalent to 8 lanes of PCIE 3.0.

      Your 2 GPUs and 2 NVMe drives are covered, provided there's a motherboard that carves it up the way you like, or you're willing to use risers/adapters.

      And in a future scenario where we have 4x PCIe 4.0 NVMe drives and 16x PCIe 4.0 GPUs that you want to use, you can fix it by upgrading to a single GPU. SLI and CrossFire are dead ends. Regardless, 16x PCIe 3.0 is hardly a bottleneck. If you stick with Nvidia and SLI, you'll have the SLI/"NVLink" bridge as well, which further mitigates any PCIe bandwidth concerns. CrossFire hasn't had a bridge in a decade, I believe. AMD just uses PCIe for everything.

    18. Re:Desktop CPU Lanes by Highdude702 · · Score: 1

      Well, AC pretty much summed it up better than i would have. But from everything you are trying to run then a desktop platform is not fot you and yiu should be jumping to workstation ot server platforms. There is no other way unless you dont use them all at once in which a pcie switch may work for you but those are not cheap at all and would be cheaper to go TR4.

    19. Re:Desktop CPU Lanes by DRJlaw · · Score: 1

      Posting as AC to preserve moderation.

      Pull the other one.

      Ryzen 3000 (presumably) products / Zen 2 will be on PCIe 4.0.
      Even if you only get 16 lanes for the primary slots, that's equivalent to 32 lanes of PCIE 3.0.
      Plus you'll have 4 PCIe 4.0 lanes for NVMe. That's equivalent to 8 lanes of PCIE 3.0.

      No, it's not equivalent, because you still have only 16 lanes for the primary slots, and still have only 4 lanes for NVMe. You must have a PCIe 4.0-capable device for there to be any bandwidth equivalence, otherwise you won't even have that.

      If you want a PCIe 4.0-capable GPU, there's reportedly Navi releasing in July. That's it. No prior-generation AMD(ATI) products, no current or prior generation NVIDIA products.

      If you want a PCIe 4.0-capable NVMe SSD, there's a Phison prototype that's reportedly releasing in 19Q3.

      Your 2 GPUs and 2 NVMe drives are covered, provided there's a motherboard that carves it up the way you like, or you're willing to use risers/adapters.

      Why are you using the present tense for an unreleased CPU combined with an unreleased GPU and unreleased SSD? Why do you assume that someone who has an issue with spending a few thousand on a HEDT CPU and motherboard will replace the MB, the CPU, 2x GPUs, and 2x SSDs just to achieve a lateral-grade to a 20 lane PCIe 4.0 platform from a 28 lane PCIe 3.0 platform?

      And in a future scenario where we have 4x PCIe 4.0 NVMe drives and 16x PCIe 4.0 GPUs that you want to use, you can fix it by upgrading to a single GPU. SLI and CrossFire are dead ends.

      Let me know when there's top-of-the-line PCIe 4.0 GPU that beats 2x first tier current (or last) generation PCIe 3.0 GPUs. Your "future scenario" is 4-5 years out.

    20. Re:Desktop CPU Lanes by thegarbz · · Score: 1

      I have 2x GPUs in SLI.

      Even if they were 2x Titan X GPUs the difference between running 8x/8x and 16x/16x is a rounding error at best on all but a very peculiar and very specific set of circumstances even on incredibly graphic intense games.

      The NVMe case is far more compelling, not that I see a reason to have that many high speed devices for anything other than posting brag numbers. If there was a real reason you'd likely be running an EPYC or Threadripper.

    21. Re:Desktop CPU Lanes by Anonymous Coward · · Score: 0

      some guys like being cuckolded by misery in modern feminist marriages..and misery loves company.

    22. Re:Desktop CPU Lanes by Highdude702 · · Score: 1

      There are already options out to do that. And I'm sure after some devices use PCIe 4 there will be a spliter for it.

    23. Re:Desktop CPU Lanes by war4peace · · Score: 1

      That's not the point. The point is 8x+8x is still impossible to achieve if you have even ONE nVME SSD. I don't need 16x+16x for the GPUs. I need 8x+8x, hell even 8x+4x would be fine, provided both my SSDs get 4x CPU lanes each. That's not the case with any current desktop platform except Intel's X299 and AMD's X399 platforms.

      And there's another point I am trying to convey. Back in 2017 when I bought the CPU and motherboard, they set me back around $700 (both), in my country, which means USA prices would have been at least 25% lower, say around $550 for both. I got 6 cores, 12 threads, 4.25 GHz frequency and 28 PCI Express lanes. All good and dandy. Fast-forward to today, this kind of deal doesn't exist any more. Here in my EU country, the lowest price for a 7900X CPU is $1350. That's 2.5 times I had paid for the CPU+motherboard combo a couple years ago. Threadripper 1950X (first generation) costs $784 at some shady stores. The larger stores price it at close to a grand. Just the CPU.

      Eventually I will upgrade my current platform, but now is not the time.

      --
      ...gis sdrawkcab (usually not responding to ACs; don't bother posting as AC)
    24. Re:Desktop CPU Lanes by Highdude702 · · Score: 1

      Well, with ryzen you could have 8x/8x/4x

      I currently have a 1070 at 16x and an NVMe at 4x. I dont understand how youre not getting the math on this. If I was to put another gpu in I would then be at 8x/8x with a 4x nvme nothing sharing lanes.

    25. Re:Desktop CPU Lanes by Highdude702 · · Score: 1

      provided both my SSDs get 4x CPU lanes each Sorry I missed that. In this case you could stick a NVME in the board, one in your add-in card then do 8x/4x/4x + 4x nvme.

    26. Re:Desktop CPU Lanes by thegarbz · · Score: 1

      The point is 8x+8x is still impossible to achieve if you have even ONE nVME SSD.

      Err x8+x8 + dedicated x4 for NVMe is the main selling point of Ryzen.

      I need 8x+8x, hell even 8x+4x would be fine, provided both my SSDs get 4x CPU lanes each.

      There's desktop motherboards that provide what you want. Asus's DIMM.2 system uses PCIe lanes from the memory leaving the GPU free for 2x x4 lanes to M.2 slots, and many motherboards on the market have PLX chips that switch idle PCIe lanes between slots allowing you to put 2 or even 4 M.2 slots at full speed on a motherboard with x8+x8 graphics at the same time.

      Your requirement doesn't seem to include transferring at 12GB/s while maintaining full speed through the GPUs.

      Back in 2017 when I bought the CPU and motherboard, they set me back around $700 (both), in my country, which means USA prices would have been at least 25% lower, say around $550 for both. I got 6 cores, 12 threads, 4.25 GHz frequency and 28 PCI Express lanes. All good and dandy. Fast-forward to today, this kind of deal doesn't exist any more.

      You have an odd set of requirements that isn't marketed at the desktop by the CPU manufacturers anymore precisely because PCIe idle lane switching is a thing now. It's also a very specific use case (preferencing PCI Express lanes over CPU cores) that is typically handled by server motherboards. Consider yourself lucky that you bought that development quirk.

      But really ... you actually expect to upgrade hardware in 2 years? This isn't 1995 anymore. If I bought a PC back in 2017 and spent *half* as much as you did I would expect to not even be looking at hardware articles let alone consider buying something until at least 2022. So I am right there with you, now is *not* the time for you to upgrade.... Unless you desperately want to annihilate money, in which case send it my way and I'll spend it on hookers :-)

    27. Re:Desktop CPU Lanes by Anonymous Coward · · Score: 0

      And why'd that be a misery. Being cuckolded is fun!

    28. Re:Desktop CPU Lanes by Anonymous Coward · · Score: 0

      You need more PCI Express lanes? Tough luck, cough up a couple grands on CPU+motherboard alone.

      Thing is, with AMD, you're more likely to get more bang for buck out of the motherboard than Intel where you almost surely need to upgrade for another Skylake architecture.

  25. They could also run of farts by Anonymous Coward · · Score: 0

    Just wait for it to come out, spreading rumors is just irresponsible

  26. Re:Drumpf for FOURTH REICH!!! by Anonymous Coward · · Score: 0

    I celebrate the anniversary of that faggot's suicide, lol. What a bitch.

  27. No. Rubbish. by Anonymous Coward · · Score: 0

    This whole thing is based around the fact that the IF speed is doubling, and you can provide the previous intercore bandwidth with half the links.

    This ignores the increase in the number of cpu cores and the bandwidth requirements addedby pcie4.

    This is ignorant fanboism at best and veiled marketing at worst.

  28. If 2 extra PCIe lanes matters.... by madsh · · Score: 1

    ...you are doing it wrong!

    1. Re:If 2 extra PCIe lanes matters.... by thegarbz · · Score: 1

      Or maybe *you* are understanding it wrong.

      The difference between having 2 extra PCIe lanes is that the SCH gets a dedicated lane to each chip. What this means in practice is that the otherwise group divisible by 16 which is split down into 2x8 or 4x4 lanes now suddenly has the bandwidth to support 2 additional full speed NVMe drives (2 x x4 lanes as required for NVMe) instead of giving you a 2x x3 interface (which for NVMe would only run at x2 for compatibility reasons) and splitting 2x x1 lanes off for system management.

  29. Filtering on slashdot? by madsh · · Score: 2

    Why do I get to see the meta data on the post that are hidden by filter??? Really... Are you trying to impress me but showing all the hard work you are doing? How good you are at counting? Do your job, shut up, fuck out of my attention span! Am I the only one having in head conversations with user interfaces :)?

    1. Re:Filtering on slashdot? by Highdude702 · · Score: 1

      Yes, yes you are..

      *backs away slowly*

  30. Desktop GPU Lanes by Anonymous Coward · · Score: 1

    Except that Crossfire (multigpu) setups in gaming are a bust and didn't work out like people wanted.

  31. 163 lanes by goombah99 · · Score: 1

    I don't know now one gets to 162. 128+32+2 ? 2*9*9 ? since it's not a power of 2 or a simple sum of a power of 2, or a simple multimple of a power of 2 it seems unusual. Even if they were trucking around parity bits then it would have been something like 2*8*9 not 2*9*9. Maybe they are planning for having processor cores in multiples of 3?

    Anyhow Intel loves meaningless spec wars like Megahertz counts, so I'm sure we'll hear about the one with 163. Or more likely they will just double up two of their 82 lane systems and have 164

    --
    Some drink at the fountain of knowledge. Others just gargle.
    1. Re:163 lanes by edwdig · · Score: 1

      If the two processors are both accessing the same region of memory at the same time, they need to be able to pass the changes back and forth to ensure memory state is consistent. I believe they do it by directly connecting a bunch of PCIe lanes between the processors.

      You've got two issues at play here. When you get to the processors with tons of cores, there's often several chips inside the package. A 32 core processor might have 4 chips inside it, each with 8 cores. Those chips need to talk to each other, which takes some lanes. If the motherboard supports multiple sockets, you need lanes connecting the sockets.

      Whatever is left after all that can be used by the rest of the motherboard.

    2. Re:163 lanes by Anonymous Coward · · Score: 0

      in current epyc its 128 lanes per cpu, with 64 reserved for the interconnect... leaving 64 per cpu for other uses depending on motherbaord maker how many of those are then exposed to the user as expansion lanes/slots

      in this tard hat artical its just a ass speculating that cause pcie 4.0 is faster only 32 lanes will be used for interconnect so that leaves 96 per cpu.. so magic?!

      at end of day it doesnt matter as its up to mobo makers to then some how deliver these to us, and some actual use case for so many lanes....

    3. Re:163 lanes by thegarbz · · Score: 1

      10x x16 lanes + 2x x1 lanes used exclusively for the management interface (one for each chip) so that the x16 interfaces don't need to be robbed of a lane. That may not seem relevant until you realise there are no PCIe x15, PCIe x7 or PCIe x3 devices out there.

    4. Re:163 lanes by fintux · · Score: 1

      AMD uses Infinity Fabric for the inter-chiplet communication, so those don't actually use PCIe lanes. But when there are two sockets, those will at least in the current implementations require the usage of PCIe lanes (which would essentially encapsulate the Infinity Fabric between the two sockets).

  32. Obviously the Golden Ratio by Anonymous Coward · · Score: 0

    This is AMD's tribute to Phi, the Golden Ratio.

  33. Re: what about quad socket?? with the faster links by Anonymous Coward · · Score: 0

    PCI/e used to be measured in mhz. Now they are measured in ghz. The length or the wire, in cm vs mm, is becoming a factor in the speed. A 4ghz CPU is doing multiple instruction in a quarter of a nanosecond. The speed of light has been creeping into things as a real limit.

    So adding ten or twenty cm is socket interconnects is stressing already tight constraints. Bigger chiplets already need more interconnects which makes the motherboard layout even more complex as you have to "avoid crossing the stream".

    Essentially, we already have multiple socket boards. It was just cheaper to embed the sockets onto the chip and then drop that assembly ubto a single socket.

    Or, in 90s bad tech movies. Mr. Johnny mnemonic already used Ram Doubler before you tried compressing it further.

  34. Will I still be waiting for this slow ass computer by Anonymous Coward · · Score: 0

    ...because all I hear are faster chips, more RAM, SSDs...but here I am still waiting on this slow ass computer. Every fuckin' day.