AMD Ryzen 16 Core “Whitehaven” Enthusiast CPUs Leaked – 3.6GHz Clock Speed, Boatloads of Cache & Qua

Discussion in 'AMD Processors' started by JustReason, May 11, 2017.

  1. IdiotInCharge

    IdiotInCharge 2[H]4U

    Messages:
    3,917
    Joined:
    Jun 13, 2003
    That is a good point- they're talking about sustaining 3.6GHz max across all cores, 3.5GHz max on the 16-core part.

    I still don't believe they'll squeeze that into 155W real TDP, but <200W is certainly plausible.
     
  2. DuronBurgerMan

    DuronBurgerMan Gawd

    Messages:
    621
    Joined:
    Mar 13, 2017
    155W in AMD's method of calculating it, sure. But real value? Yeah, 155W is too low.
     
  3. SighTurtle

    SighTurtle Gawd

    Messages:
    940
    Joined:
    Jul 29, 2016
  4. bezant

    bezant Gawd

    Messages:
    739
    Joined:
    Oct 7, 2009
  5. SighTurtle

    SighTurtle Gawd

    Messages:
    940
    Joined:
    Jul 29, 2016
    Its ongoing.
     
  6. Schmave

    Schmave [H]ard|Gawd

    Messages:
    1,720
    Joined:
    Jan 2, 2001
    I posted in the other thread but they just announced "Threadripper" (16 cores, 32 threads) as being official and launched sometime this summer. They also said more information would be coming at Computex.
     
  7. Pieter3dnow

    Pieter3dnow [H]ardness Supreme

    Messages:
    4,886
    Joined:
    Jul 29, 2009
    show and tell at Computex 2017 , nothing to worry about missing anything really.
     
  8. Pieter3dnow

    Pieter3dnow [H]ardness Supreme

    Messages:
    4,886
    Joined:
    Jul 29, 2009
    Here is what I like about AMD they did well , there is no reason for them to suddenly drop the ball for this R9 line up. With a little luck the platform on which this ends up has less problems then AM4 by the time it is released.
    Still wondering on the process there using if it is the same as the current R7 ones.
     
  9. gigaxtreme1

    gigaxtreme1 2[H]4U

    Messages:
    2,916
    Joined:
    Oct 1, 2002
  10. Pieter3dnow

    Pieter3dnow [H]ardness Supreme

    Messages:
    4,886
    Joined:
    Jul 29, 2009
    There getting almost the same base performance as the Ryzen R7 1800x with more cores that must be one hell of a process to be able to pull that of.
     
  11. juanrga

    juanrga Gawd

    Messages:
    906
    Joined:
    Feb 22, 2017
    RyZen muarch is more throughput oriented than Intel, but this CPU is using two dies glued together and this comes with a performance penalty when one of the dies has to access the cache in the other die.
     
  12. juanrga

    juanrga Gawd

    Messages:
    906
    Joined:
    Feb 22, 2017
    If you know for sure that 3.5GHz are boost clocks, why then do you repetitively post tables comparing boost clocks on Zen with base clocks on SKL? And the excuse that AMD do TDP differently than Intel is already getting very old. The '95W' RyZen dissipate more than the 125W Piledriver chips from the same company, not because AMD "and AMD do it differently", but because 95W is a marketing label to mislead consumers.
     
  13. juanrga

    juanrga Gawd

    Messages:
    906
    Joined:
    Feb 22, 2017
    The 16C sample was officially rated at 180W, which makes this 155W marketing label still more laughable.

    Evidently it is the same process. The 16C sample used the same B-grade silicon than commercial RyZen.
     
    Shintai likes this.
  14. juanrga

    juanrga Gawd

    Messages:
    906
    Joined:
    Feb 22, 2017
    It is ironic that you start mentioning ignorance...

    First, the problems of RyZen don't start and finish on the marketing TDP doesn't corresponding to the real TDP. RyZen has latency problems, stability problems, some backward-compatibility problem (WM pink screen), IPC deficit, terrible overclocking headroom,...

    The reason why TDP and clocks are being mentioned here is just to illustrate that the recent leak with specs is a kind of joke, not very different to those RyZen joking leaks that promised 5GHz on air.

    Second, it is too late to you to request me to prove that RyZen is not a 95W chip, when reviews already demonstrated that the real, measured, TDP is higher than the marketing 95W, and AMD silently admitting it when communicating the real TDPs.
     
    Shintai likes this.
  15. Gideon

    Gideon Gawd

    Messages:
    989
    Joined:
    Apr 13, 2006
    Reading fail I see. 3.5 is base clocks not boost clocks as much as you love charts you should try reading them. I posted tables so people could see both they both showed the boost and base clocks you fool, except for the Intel 12c that one has no specs yet. It's also not my problem if you dont like the TDP cause I dont care, see a shrink over it. I can straight tell you its less then piledriver since I have a kill-a-watt meter and my new Ryzen draws less power then my old 8350 and they are both overclocked. See I post facts when I find them, if you dont like them I dont care they are simple information charts for everyone to read and have input on.
     
  16. noko

    noko 2[H]4U

    Messages:
    2,863
    Joined:
    Apr 14, 2010
    Four quad core modules connected via Infinity Fabric. Scaling should almost be linear. Maybe Intel uses glue, AMD uses a rather complex design with advance logic hooking everything up. We have to see clock speeds compared to intel.
     
    N4CR likes this.
  17. juanrga

    juanrga Gawd

    Messages:
    906
    Joined:
    Feb 22, 2017
    The problem is not on the 'charts', which are a piece of junk. The problem is on your confusing wording here.

    Neither it is related to liking or not something, but about being serious and posting specs that look at least technically possible instead repeating any piece of junk that can be found on the Internet. Didn't we learn anything about the hype and fantasies published and republished before RyZen launch that now we are going to do the same with ThreadRipper?

    I don't need to prove stuff you are making in your own head. No one here has said anything about "TDP not equating watts". In fact, your whole phrase is meaningless, because TDP is a physical quantity whereas watt is an unit. It is so meaningless like your "95TDP cooler".
     
    Shintai likes this.
  18. juanrga

    juanrga Gawd

    Messages:
    906
    Joined:
    Feb 22, 2017
    Scaling cannot be linear because accessing an L3 slice in the same CCX (sCsD) has different latency than accessing the L3 slice on another CCX on the same die (oCsD) or accessing the L3 slice on an CCX on the other die on the same package (oCoD)

    L_sCsD < L_oCsD << L_oCoD
     
    Shintai and IdiotInCharge like this.
  19. Gideon

    Gideon Gawd

    Messages:
    989
    Joined:
    Apr 13, 2006

    I am not confusing anything you are just lost and are spouting off opinion then any actual facts. The charts are leaks and I got them from techpowerup. Dont like them dont care, they show base clocks and then boost clocks. You and razor are the only ones talking about a CCX issue that will cripple the processor and no one else is. So it's not hype or fantasy I posted what I found and is easily believable to people that actually own Ryzen. I cant control the fantasy that is inside your head. Also I am quite happy with my Ryzen, I am sorry your Intel experience is bitter and you hate your machine and feel the need to lash out at people with AMD rigs. So stop bothering us when you actually have no facts to back you up.
     
    Sharps97, noko, Pieter3dnow and 2 others like this.
  20. JustReason

    JustReason razor1 is my Lover

    Messages:
    2,218
    Joined:
    Oct 31, 2015
    Still dodging I see. So you have NO proof that stating 95W is a problem other than your opinion on the matter. Now when someone states in some forum somewhere on earth that the 95W Ryzen is using less power than a 95W Intel, then by all means you can swoop in and set them straight. But for the time being, AMD stating a particular TDP does not mean they are stating how many watts they can use, but rather what cooler MINIMUM one must use with the chip---AS EVERYONE WITH EVEN THE SLIGHTEST KNOWLEDGE OF AMD AND TDP KNOW THIS.
     
    Zuul likes this.
  21. JustReason

    JustReason razor1 is my Lover

    Messages:
    2,218
    Joined:
    Oct 31, 2015
    Holy crap you are 110% correct. And welcome to the club, lol.
     
    Gideon likes this.
  22. juanrga

    juanrga Gawd

    Messages:
    906
    Joined:
    Feb 22, 2017
    The laws of physics and the specs of the current samples aren't opinions.

    Just because Techpowerup publishes something it doesn't mean it is real or we have to accept it as uncritical zombies. Techpowerup published the next set of slides as a genuine leak

    https://www.techpowerup.com/212112/amd-zen-cpu-core-block-diagram-surfaces

    but all the slides therein were fake. Most tech sites reproduced the slides whereas I questioned them and demonstrated they were fake. One day I received a very funny email in my box:

    My point is that those recent leaks about ThreadRipper look terribly suspicious. The data leaked doesn't fit with the laws of physics neither with what we know about the existent samples.

    Don't attack us only because we can use our brains and take an critical attitude against everything is published on the Internet.

    I have no idea what do you mean by "a CCX issue". Myself, I have stated clearly that the main problem that ThreadRipper will have (on top of the problems it share with RyZen such as the use of 14LPP node, or the memory controller IP) is that ThreadRipper will be a MCM2 chip. This doesn't have anything to do with the CCX structures inside each die.

    The last part of your post is ridiculous, even more if you were to check the specs of the machine on which I am tipping this. Please leave that kind of ridiculous ad hominems for SA forums or for the comment section on some WCCFTECH article and leave this thread to people that want to discuss technical aspects as I did in #74.

    You can continue ignoring that AMD has already communicated us the real TDPs for RyZen and ThreadRipper chips, but it will not change anything.
     
    Shintai likes this.
  23. JustReason

    JustReason razor1 is my Lover

    Messages:
    2,218
    Joined:
    Oct 31, 2015
    This is why I cant take anything you say seriously. You never DIRECTLY answer the questions you are asked.

    As it pertains to AMD TDP is the recommended MINIMUM for a cooler to maintain advertised clocks. You can scream at the sky till it turns yellow and hope AMD posts actual watt usage, but until they do their current definition of TDP is as I have stated. Hell you have yet to post proof that a 95W cooler can not maintain those clocks and thereby able to claim AMD is lying accurately. But since you can not then the lying part is you.

    There are plenty of reviews that show wattage, albeit not necessarily indicative of everyones experience. Not sure why it is such a big deal to you and why you seem so stuck on it.

    As far as your insistence on clocks its not gonna be that simple. Intels new lineup max is 12c/24t against AMDs 16c/32t. Even with an IPC advantage AMDs 4c/8t extra is gonna not only bridge that gap but likely surpass Intels even with clock discrepancies.

    But lets put down to simple truths and facts: At this point in time we have access to a plethora of choices, and for the first time in a while 2 companies have products in a competitive market.
     
    N4CR, Pieter3dnow and Shintai like this.
  24. juanrga

    juanrga Gawd

    Messages:
    906
    Joined:
    Feb 22, 2017
    I find interesting that AMD already admitted the real TDPs for RyZen, but certain people continue acting like if it didn't happen.

    Joining a pair of 1700 dies and adding ~20W from the MCM package gives about 200W TDP for a 16C chip. And the current 16C sample with similar clocks is rated 180W by AMD (real TDP > 200W).

    Yes, it is 16C vs 12C. Factor on the ~20% IPC gap and then consider clocks and you will check it is not so granted that 16C will be faster.
     
    Shintai likes this.
  25. KazeoHin

    KazeoHin [H]ardness Supreme

    Messages:
    6,496
    Joined:
    Sep 7, 2011
    The IPC gap is not 20%. Clock for clock, skylake is only around 10% faster than Zen. That gap grows closer to 15% with SMT dissabled, but nobody is going to buy one of these chips and dissable SMT.

    Intel will still have a huge lead in clockspeeds, however uncertain how huge that lead will be. I think in the end, the top-end RTR chip will compete pretty admirably with the top-end i7/i9 in workstation workloads.
     
    mesyn191 likes this.
  26. Pieter3dnow

    Pieter3dnow [H]ardness Supreme

    Messages:
    4,886
    Joined:
    Jul 29, 2009
    Keep on dreaming ...
     
    JustReason likes this.
  27. juanrga

    juanrga Gawd

    Messages:
    906
    Joined:
    Feb 22, 2017
    Clock-for-clock on workstation workloads Broadwell is ~10% faster than Zen; Skylake is ~13% faster than Zen and Skylake-X has higher IPC than Skylake.
     
  28. gigaxtreme1

    gigaxtreme1 2[H]4U

    Messages:
    2,916
    Joined:
    Oct 1, 2002
  29. noko

    noko 2[H]4U

    Messages:
    2,863
    Joined:
    Apr 14, 2010
    No - it really depends on the work load. At some things Intel does much better as in FP stuff, integer stuff AMD is better in general. One just need to investigate which one will do better overall. As for perf/w they are so close that I just don't see it mattering that much. Plus price can come into play as well, if you can get two RyZen systems for the price of one Intel then the game is over.
     
  30. KazeoHin

    KazeoHin [H]ardness Supreme

    Messages:
    6,496
    Joined:
    Sep 7, 2011
    Citation required for the idea of Skylake-X having higher IPC than Vanilla Skylake.

    Also, there are MANY tasks where Zen destroys Broadwell-E clock for clock, and they aren't fringe cases. Zen is essentially broadwell IPC.
     
    mesyn191 likes this.
  31. juanrga

    juanrga Gawd

    Messages:
    906
    Joined:
    Feb 22, 2017
    Aren't many and Zen wins for low-single-digit percents on those (interestingly the same kind of workloads that AMD chose for the public demos: Handbrake, Blender,...), whereas that when Zen loses it does by a larger amount, giving an average IPC well below Broadwell.

    [​IMG]

    [​IMG]


    Skylake-X is a different muarch than Skylake, with different caches and 512bit datapaths. For instance there is a huge performance gain (>10% IPC) on workloads such as CineBench.
     
    Shintai likes this.
  32. os2wiz

    os2wiz Limp Gawd

    Messages:
    411
    Joined:
    Nov 20, 2011
     
  33. os2wiz

    os2wiz Limp Gawd

    Messages:
    411
    Joined:
    Nov 20, 2011
    I see you are up to your mischief again. Since when is any synthetic benchmark, now matter how well respected, a workload?? Workloads are designated for real world applications. Perhaps you never learned that?????




    A virtual intellect in a real world.
     
    N4CR likes this.
  34. griff30

    griff30 [H]ardness Supreme

    Messages:
    8,128
    Joined:
    Jul 15, 2000
    Quad channel memory, and 32 threads.
    If it overclocks, it's a buy from me.
     
    N4CR likes this.
  35. N4CR

    N4CR [H]ard|Gawd

    Messages:
    1,908
    Joined:
    Oct 17, 2011
    Same! Been waiting for something like this long enough. Boatloads of ram, cores, good IPC, decent clocks. Maybe water or massive custom passive/or 18db air.
    This cpu will make mincemeat of video editing loads and make 4k much more accessible. Maybe even 8k with Vega and Navi helping..
     
    griff30 likes this.
  36. juanrga

    juanrga Gawd

    Messages:
    906
    Joined:
    Feb 22, 2017
    Reading accusations about the use of standard benchmarks used even by AMD in public demos of Zen performance was funny, but what is totally hilarious is these funny accusations are coming from a poster that claimed 'world-record' CPU-Z buggy scores from "a techy sales guy named Jim" (LOL) to pretend that RyZen performance was much higher than really it is...
     
    Last edited: May 19, 2017
    Shintai likes this.
  37. juanrga

    juanrga Gawd

    Messages:
    906
    Joined:
    Feb 22, 2017
    I don't know any reason why it couldn't overclock.
     
    griff30 likes this.
  38. juanrga

    juanrga Gawd

    Messages:
    906
    Joined:
    Feb 22, 2017
    Actual info

    [​IMG]
     
    CSI_PC, griff30 and Shintai like this.
  39. Gideon

    Gideon Gawd

    Messages:
    989
    Joined:
    Apr 13, 2006
    Yep who knew they would hit even higher speeds, much higher then what you said either. But I dont make a blog about predicting things right either. But thanks for pointing out things turned out even better then that list I posted minus the 14 cores.
     
    ManofGod likes this.
  40. drescherjm

    drescherjm [H]ardForum Junkie

    Messages:
    13,467
    Joined:
    Nov 19, 2008
    I think he was commenting about the TDP difference. Although it does not make sense to have all 3 models with the same TDP. With that said I don't trust AMD's TDP ratings at all.