Computex 2017: Intel Core i9 Launches With up-to 10 cores first - Does 4.3 GHz on LCS

Discussion in 'Frontpage news' started by Hilbert Hagedoorn, Jun 3, 2017.

  1. Denial

    Denial Ancient Guru

    Messages:
    13,228
    Likes Received:
    2,717
    GPU:
    EVGA RTX 3080
    You can't delid a soldered CPU without either cracking the solder or melting it. So unless one of these people has a specifically calibrated X-Ray machine, they weren't seeing it, or again, even looking for it. This isn't to mention that I linked an article that you quoted where Intel itself sets up an experiment that shows how they were mitigating the issue but seen it as a problem for future smaller nm dies - it was published in 2006.

    You're just assuming the TIM is cheaper. Everyone just assumes the TIM is cheaper. It's literally the basis for my rant. Prove to me it's cheaper. Give me any site, forum post, anything where it goes into more detail then "It has to be cheaper because it's not metal!" or "it has to be cheaper because it's not helping my OC!"

    I find it really hard to believe that you can't distinguish the difference between those two completely different scenarios. That's embarrassing.
     
  2. D3M1G0D

    D3M1G0D Ancient Guru

    Messages:
    2,126
    Likes Received:
    1,367
    GPU:
    2 x GeForce 1080 Ti
    But if the solder has cracked to any appreciable degree then it shouldn't be as difficult to delid (in fact, this would be one of the indications that it has indeed cracked). Like I said, if this was a widespread issue then somebody would have reported it. When there is no evidence of an issue, the best explanation is that it's not an issue.

    Also, we're talking about HEDT chips here, which have cores from 10 to 18 (these aren't your typical quad cores). As one of the articles you cited says, "This failure mechanism is one reason why small DIE CPUs like Haswell-DT or Skylake are not soldered while the large Haswell-EP CPUs are soldered." And yet, Intel won't solder the IHS on their 10+ core Skylake-X CPUs, which have large dies. So what's your argument again?

    Hey, I'm just explaining how ludicrous your explanations are. Nobody would be convinced that Intel needs to sell dongles to enable certain RAID modes, and nobody would be convinced that Intel suddenly needs to start using paste for their 18-core CPU.
     
  3. Denial

    Denial Ancient Guru

    Messages:
    13,228
    Likes Received:
    2,717
    GPU:
    EVGA RTX 3080
    You realize they are micro cracks right?

    http://overclocking.guide/wp-content/uploads/2015/11/micro_cracks-1024x283.png

    The entire length of that photo is .01 inches - it's not like the entire IHS is coming off.

    The micro voids form based on the density of the heat, not some weird unknown property of core size - it just turns out that smaller cores have a higher power density. Certain regions of chips are more dense than others - the AVX512 registers included on these for example are extremely dense.

    Also this isn't my explanation, it's Intels and it's not something that's new to Intel either - the void problems occur in a variety of different places for various reasons in BGA packaging, the Xbox (RROD Problem), etc.

    Either way this argument doesn't seem to be going anywhere and it's essentially derailed the topic so I'm done.
     
  4. Prince Valiant

    Prince Valiant Master Guru

    Messages:
    723
    Likes Received:
    72
    GPU:
    EVGA GTX 1080 ti
    With the bulk that Intel is buying materials in I doubt it makes much difference. Whichever requires more materials and processing steps is likely more expensive regardless of performance.
     

  5. Fox2232

    Fox2232 Ancient Guru

    Messages:
    11,325
    Likes Received:
    3,092
    GPU:
    5700XT+AW@240Hz
    But they had desktop/server/mobile superiority to feed it. So Again, how would they feed it now? Reserves? Check how small they have them.
    I agree, And AMD delivers exactly same stuff or maybe better, because Bristol Ridge is already touching intel's mobile chips in both performance and efficiency + better iGPU for low power devices. So Raven Ridge has potential to flip mobile demand to AMD's side. This remains in hands of manufacturers.
    Actually AMD started to flush their stock in ~2016Q3 and started to build new stock in 2016Q4. 2017Q1 was just ramp-up and old stuff was gone. (Because AMD is small and "unstable" this can be seen from numbers.)
    They are equal, that makes them an alternative in comparison to the past.
    Intel does not support servers built based on their platform unless it is their code which breaks. You make so called "Software Call" and OS provider will be handling software bugs/violations/compatibility. In case of RHEL it will be handled by guys here in Brno. In all those years I can remember intel's involvement in solution of any soft/hardware problem is like 0.1%. That's because there is almost nothing what can happen to server (in middle of its life cycle) and HW manufacturer can solve it other way than by replacement.
    That's funny way to say it. But it kind of contradicts reality. If AMD had no way to supply their better chips, why intel did use shady tactics and persuaded those mentioned OEMs not to use AMD's chips? Just to be on safe side?
     
  6. Chillin

    Chillin Ancient Guru

    Messages:
    6,814
    Likes Received:
    1
    GPU:
    -
    They still have the desktop, server and mobile to feed it. Ryzen targets strictly a limited set of power users for the moment. The majority of sales are either in mobile (where Ryzen is a complete unknown and has zero presence for the moment), low end desktop (where the margins are slight), HEDT (where Intel has an equally competitive or superior platform, though perhaps at a slightly greater cost), and server (where AMD doesn't even have anything yet, just paper announcements.).

    So let wait and see before jumping to incredible conclusions about AMD's ability to perform in these markets till they even have a hard product launch for them...

    If you look here in this this forum alone and see the growing pains Ryzen is having and how many people said that they chose Intel because Ryzen has too many uncertainties. then apply that to an IT domain professional who wants his work to be as simple as possible and who's management doesn't really care if the difference in cost is a few thousand dollars over dozens of workstations; you begin to see the picture and where AMD struggles.

    AMD in 2003 was able to at best maybe supply around 30% of the market give or take. The Dresden FAB didn't come online until late 2003 (and ironically added to their losses). They also had a large amount of old inventory that wasn't moving which compounded their losses. Like I said earlier, Intel is and has always been a Fab monster first. Further reading:

    http://www.lnksrv.com/community/pod/2002/020122.htm
    https://arstechnica.com/business/20...e-top-of-the-mountain-to-the-deepest-valleys/
     
  7. Fox2232

    Fox2232 Ancient Guru

    Messages:
    11,325
    Likes Received:
    3,092
    GPU:
    5700XT+AW@240Hz
    That's one lie after another, and you either know it or you do not understand it even a bit.
    - Desktop pains are not growing, they are going away and they are only connected to achievable memory clock
    - By "apply that to an IT domain professional" You surely meant, people who do not try to run their servers out of specifications and therefore will have Ryzen running with 2666MHz memory at most and face no issues getting there...

    You do contradict yourself here more than a bit.
     
  8. Chillin

    Chillin Ancient Guru

    Messages:
    6,814
    Likes Received:
    1
    GPU:
    -
    How did I contradict myself in regards to AMD? AMD's old K7 stock wasn't moving, and they didn't have the ramp-up capacity for K8 to meet the potential demand of the large OEM's till a later time (at which point Intel was already moving away from Netburst to what Banias started, the Core architecture). If I remember correctly, in 2002 AMD had one Fab under construction; Intel had 3 if I'm correct.

    Again, you can believe what you want. I work in sales and I am telling you what everyone tells me they want (the CTO's, etc., from small to large enterprises) and what actually sells on that basis. You can either go on believing that your viewpoint of the world is right and that everyone else is wrong, or you can actually listen to what people say and see the results therein.
     
    Last edited: Jun 7, 2017
  9. Fox2232

    Fox2232 Ancient Guru

    Messages:
    11,325
    Likes Received:
    3,092
    GPU:
    5700XT+AW@240Hz
    Rly? Rly?
    Maybe that's root of problem, reason why we disagree so much. It is not me who look left and right and get to great many different shades of information.
    It is likely you, who considers your small space as sum of everything: "what everyone tells me". (Are they really everyone? Does entire IT world get in touch with You? Your sense of self Grandiosity is something you should consider as flaw before it grows too big.)
    And it is likely you again with your belief that everyone agrees with you except me: "everyone else is wrong". (I definitely do not consider everyone around to be wrong, because that would mean I am wrong as much as great many others who would agree with me.)

    Anyway, seems like it turned into opinion based discussion which belongs to pub. So I am out.
     
  10. No1True

    No1True Member Guru

    Messages:
    146
    Likes Received:
    0
    GPU:
    1070
    Since at least half of this thread has amd posts, I want to ask, so.. how many notebooks will contain threadripper? Or its purely for servers?
     

  11. sverek

    sverek Ancient Guru

    Messages:
    6,097
    Likes Received:
    2,955
    GPU:
    NOVIDIA -0.5GB
    Threadripper is aimed for high level desktop, so its not limiting to servers.
    There probably gonna be laptops with it.
    Designers and modelers on the move will appreciate powerful laptop with affordable price tag.
     
  12. Fox2232

    Fox2232 Ancient Guru

    Messages:
    11,325
    Likes Received:
    3,092
    GPU:
    5700XT+AW@240Hz
    Well, normal Ryzen can easily fit into 65W TDP in desktop. Once in notebook, it may keep decent amount of performance even at ~55W.

    But TR is double of that and desktop is going to be 125/140/155W. Yes, it can still be downclocked to be used in "mobile workstation" and eat around 100W. But main thing about TR are additional PCIe lanes over Ryzen. If those to be used, one would expect 2 GPUs. That makes pretty power hungry portable device. There may be 1 brave manufacturer to do it.

    I much more expect Regular Ryzen to be used at bit higher TDP than 65W in mobile as it is more economical and performance wise:
    100W TR 16C/32T would operate within 2.6~3.1GHz
    100W Ryzen 8C/16T can operate within 3.0~3.8GHz
    Apparently 100W TR can go to same 3.8GHz GHz as Ryzen, but once workload goes to more cores, power limit will kick in anyway.
     
  13. Loophole35

    Loophole35 Ancient Guru

    Messages:
    9,781
    Likes Received:
    1,135
    GPU:
    EVGA 1080ti SC
    Still can't wait to see Ryzen mobile. Single CCX and Vega based iGPU. Definitely my next notebook. Intel is about to lose all of it's footing in the mobile market.
     
  14. PrMinisterGR

    PrMinisterGR Ancient Guru

    Messages:
    7,465
    Likes Received:
    495
    GPU:
    Sapphire 7970 Quadrobake
    I know it's a pipe dream, but if they managed to bake some hbm in that package that both the CPU and the GPU could use, that would be amazing.
     
  15. Loophole35

    Loophole35 Ancient Guru

    Messages:
    9,781
    Likes Received:
    1,135
    GPU:
    EVGA 1080ti SC
    I think you are getting into a size constraint there plus cost for such a chip would make the product unappealing to the mass market. Would be nice in the desktop or custom (read console) market.
     

  16. Embra

    Embra Maha Guru

    Messages:
    1,107
    Likes Received:
    310
    GPU:
    Vega 64 Nitro+LE
    There needs to be decent hardware around these APUs this time around. Let them show what they can do instead of handicapping them.
     
  17. PrMinisterGR

    PrMinisterGR Ancient Guru

    Messages:
    7,465
    Likes Received:
    495
    GPU:
    Sapphire 7970 Quadrobake
    I can't see how it would be unappealing. Just 1GB of HBM there along with the Vega memory controller could make a ton of difference for GPU-bound applications. Even Intel does it with some Iris GPUs, but with e-DRAM, which is much more expensive (I think).
     
  18. Loophole35

    Loophole35 Ancient Guru

    Messages:
    9,781
    Likes Received:
    1,135
    GPU:
    EVGA 1080ti SC
    And all of iris Intel chips are expensive.

    I agree it would be nice but I don't expect it until sub 10nm is used for the APU's. (Again speaking on the mobile market)
     
  19. PrMinisterGR

    PrMinisterGR Ancient Guru

    Messages:
    7,465
    Likes Received:
    495
    GPU:
    Sapphire 7970 Quadrobake
    A Surface with an APU like that would be something else. Raven Ridge and the mobile chips can't come soon enough, AMD is like delaying their license to print money.
     
  20. Loophole35

    Loophole35 Ancient Guru

    Messages:
    9,781
    Likes Received:
    1,135
    GPU:
    EVGA 1080ti SC
    I can agree with that.
     

Share This Page