AMD's Ryzen CPU's (Ryzen/TR/Epyc) & Vega/Polaris GPU's

Discussion in 'Hardware Components and Aftermarket Upgrades' started by Rage Set, Dec 14, 2016.

  1. TANWare

    TANWare Just This Side of Senile, I think. Moderator

    Reputations:
    2,125
    Messages:
    8,439
    Likes Received:
    2,969
    Trophy Points:
    331
    There is only one defensible position for Intel here I can see. And it requires a bit of history. If everyone remembers back at just before Ryzen 7 launch Intel knew of the HEDT and its plans at the time to keep up with AMD was only up to 12 cores in the next iteration. At the time I am sure they were preparing press material as well.

    I am sure Intel was keeping track of Epyc and TR which early on, like Ryzen 7, was based on stepping 1 of the Zen core. I think with it there were optimizations just not doable under AGESA that were detrimental to performance. This probably left the new architecture much in the same performance boat Bull Dozer would have been in under the same multi core chip.

    This can be bolstered somewhat by all the leaked Benchmarks under Stepping 1. Primarilly of the TR but it shows that silicon did exist at some point and this is what AMD was using. It show where IF was just not giving as good scaling result to the multi CCX complex chips at all. This where is seems with Ryzen7 it scaled reasonably well.

    At some point Intel got wind of the enhanced performance capabilities of both Epyc and TR under stepping 2 of the cores and this started the scramble of both HEDT and XEON chips to now compete with the upcoming AMD chips. Not that Intel lacked the tech but was now going to have to call on it to both upgrade all of their chips, prepare for the price competitive market and possible subsequent market share loss.

    Now this is evident in the partner slide as we know Epyc and Supermicro exist. Ryzen and Gigabyte, MSI at least exist as well up to TR. Microsoft exists, there are I am sure others as well on the slides. So to me these are leftovers from the time Intel felt ZEN just posed no threat, and if it were true were no better than Bull Dozer and a "Glued Chip" as it was.

    Now I doubt this will happen as it means admitting to Ryzen forced reorganization, industrial spying, possible competition, possible revenue loss etc. etc. etc.. But this is just my thoughts here.
     
    Last edited: Jul 16, 2017
    Rage Set likes this.
  2. hmscott

    hmscott Notebook Nobel Laureate

    Reputations:
    3,619
    Messages:
    13,446
    Likes Received:
    15,571
    Trophy Points:
    931
    EPYC™ Tech Day: Gerry Talbot
    AMD Epyc I/O and Infinity Fabric Electrical Interfaces
    New re-issue by AMD, not sure yet what's changed from the first one AMD pulled...

    Original post:
    http://forum.notebookreview.com/thr...ga-polaris-gpus.799348/page-232#post-10564531

    Nino Magazinovic 15 minutes ago
    "Without AMD we would today from Intel get only 32bit 2core cpu for 300$. No 4k videos and games, No VR, nothing. Now, AMD with infinity fabric scalling brings us again BACK TO THE FUTURE, 8K gaming and VR."
     
    triturbo, Kommando and Papusan like this.
  3. hmscott

    hmscott Notebook Nobel Laureate

    Reputations:
    3,619
    Messages:
    13,446
    Likes Received:
    15,571
    Trophy Points:
    931
    AMD Ryzen 3 1200 & 1300X Unboxing
     
  4. Deks

    Deks Notebook Virtuoso

    Reputations:
    872
    Messages:
    3,580
    Likes Received:
    737
    Trophy Points:
    181
    Something is amiss here.
    The RX 580 is 65W with radically lowered clocks (over 30% it would seem) and it still manages keep up with 1060 (120W) while losing only 5-10% performance it would seem in DX11 and maintaining an advantage in DX12.
    The CPU's in comparison are 65W on AMD's part (Ryzen 1600 - 6 core part no less) vs a 7700HQ (4 core, 4 threads and 45W).

    The quarrel I have here is with the premise that a 65W Ryzen CPU is seemingly unable to maintain its turbo speeds and approaches high temperatures as a result, whereas we see Intel and Nvidia laptops (7700 HQ and 1060, both of which have a combined TDP of 165W - while on AMD's side, there's a combined TDP of 130W - 35W less) capable of maintaining their turbo speeds and achieving lower temperatures.
    Why is it seemingly impossible to keep the AMD parts (CPU in this case) from throttling and achieving high temperatures, whereas a higher TDP laptop system that has Nvidia+Intel can more or less manage to do just that at about 15 degrees C lower temps?

    On average, I would imagine AMD would have an advantage on load power draw (which is higher on both systems) and OEM's would have an easier time cooling it.

    This seriously baffles me.
    Is Asus really that bad with cooling in their laptops or is this just a pre-draft issue that needs and can be sorted?
    Could it also be thermal paste related? (doubt it would result in this much discrepancy).

    It would be pretty stupid for Asus to do an amazing thing like putting together a proper FULL AMD system in such a long time (with 6 cores no less on the CPU end) and then skimp out on the cooling, resulting in poor performance and throttling?

    Why can they seemingly keep GPU's with much higher TDP's and overall power draw stable and properly 'fed' in terms of power, but have issues with a CPU?

    I'm going to wait for the final release and see if these issues have been or can be addressed.
    I really want an all AMD Ryzen (8 cores/16th) RX 580 laptop... but not if Asus botches the cooling.
     
    Last edited: Jul 16, 2017
    hmscott likes this.
  5. hmscott

    hmscott Notebook Nobel Laureate

    Reputations:
    3,619
    Messages:
    13,446
    Likes Received:
    15,571
    Trophy Points:
    931
    That's a lot of doubt for a simple first preview of an engineering sample, maybe wait for real information and production reviews before jumping overboard.

    All the vendors botch the CPU cooling to some extent, and that's why so many OCD'OC'rs repaste.

    You can get a good one, and most Intel CPU's you can live with by undervolting, but with AMD Ryzen it's unknown how that's going to play out.

    I've had 7 Asus laptops, and haven't needed to re-paste any of them.

    Only one laptop had a fault develop with a GPU (AMD), and Asus RMA fixed it in 2 weeks with no further problems - and it's still working.

    Let's try to stay positive until facts arrive to prove otherwise. :)
     
    Deks likes this.
  6. Deks

    Deks Notebook Virtuoso

    Reputations:
    872
    Messages:
    3,580
    Likes Received:
    737
    Trophy Points:
    181
    I'm aware of the issues with pre-release engineering samples... that's why I said I'd wait for the final release before drawing any conclusions.
    It's just that even with an engineering sample this close to release, you'd think ASUS would have things like the cooling sorted - but I have to admit I was indeed skeptical of pre-release info to begin with.

    As for undervolting Ryzen... don't know... but I tried a quick search on it and came up with this:
    https://twitter.com/BitsAndChipsEng/status/843871819585208322

    Those are total system power draw (so actual CPU would be pulling much less), and someone was able to undervolt a Ryzen 1700 on 3.7GhZ to 1.05V it would seem on Reed forums.
    Now if we can get that kind of silicon into these Asus units... that would be NICE for undervolting which would of course result in lower power draw and temperatures - but Asus should really work on this to get those in proper ranges on stock speeds from the get go - no more skimping from OEM's.

    I'm also aware of the thermal paste issues commonly present with laptop OEM's. I already have a quality thermal paste which I'm using in my Acer.
     
    Last edited: Jul 16, 2017
    hmscott likes this.
  7. TANWare

    TANWare Just This Side of Senile, I think. Moderator

    Reputations:
    2,125
    Messages:
    8,439
    Likes Received:
    2,969
    Trophy Points:
    331
    Last edited: Jul 16, 2017
    ajc9988 likes this.
  8. ajc9988

    ajc9988 Death by a thousand paper cuts

    Reputations:
    1,009
    Messages:
    4,291
    Likes Received:
    5,743
    Trophy Points:
    581
    All with 3 cores makes more sense with how they connected epyc partners on different sites and different chips.

    Sent from my SM-G900P using Tapatalk

    Edit: until we see something showing a work around to the mirrored connection, I'd be dubious. Now, the 3x2 complexes could mirror each other. This may be how they want to approach it, but we don't know if the ccx communication to the other complex on the same die is core dependent as well, which also uses IF.
     
  9. hmscott

    hmscott Notebook Nobel Laureate

    Reputations:
    3,619
    Messages:
    13,446
    Likes Received:
    15,571
    Trophy Points:
    931
    Intel's 3DXpoint Optane 32GB - Running on AMD's AM4 Platform with Ryzen R7!!! It works!! But..
     
    ajc9988 likes this.
  10. TANWare

    TANWare Just This Side of Senile, I think. Moderator

    Reputations:
    2,125
    Messages:
    8,439
    Likes Received:
    2,969
    Trophy Points:
    331
    I lean more toward the 3x3x3x3. I think this is the reason for just the two skews. They took the optimal configurations and are releasing just those. The 12 core was though to be a $599 or so price point but as we see it is quite a bit more. it may be the off counted CCX cores scaled weird such as say a 3x2x3x2 for 10 core. With only $200 between the 1920x and 1950x it does not make much sense for a 4x3x4x3, or does it?
     
    ajc9988 likes this.
Loading...

Share This Page