1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

AMD Zen 2 (Ryzen 3000)

Discussion in 'CPUs' started by Jamin280672, Jul 20, 2018.

  1. Murphy

    Wise Guy

    Joined: Sep 16, 2018

    Posts: 1,292

    Speaking of yields it seems Ryzen 2 is managing to hit around 70%, apparently that's only 10% less than the old 14nm.
     
  2. CAT-THE-FIFTH

    Capodecina

    Joined: Nov 9, 2009

    Posts: 18,175

    Location: Planet Earth

    AMD Ryzen magazine with Cosplay:

    http://ryzen.co.kr/bbs/board.php?bo_table=download&wr_id=37

    I linked to the original articles earlier IIRC,but TPU has picked up on the 12NM desktop APU refresh:

    https://www.techpowerup.com/254804/amd-ryzen-3-3200g-pictured-and-de-lidded

    7NM costs much more per wafer and then there is the IO die which is a larger chip on 12NM/14NM. Also 200MM2 die on 14MM2 vs 70~80MM2 die on 7NM and the latter is still yielding less. Also I will be interested to see the binning. People are expecting clockspeeds near to 5GHZ with this,whereas server products are typically less than 4GHZ. So its not only getting sufficient 8C chiplets for launch but also those which clock close to 5GHZ as some want here. We will see how it pans out. For most gamers on here its more important that the cheaper 6C and 8C products have good performance boosts over the previous gens. If AMD can release two chiplet 16C consumer products on X570 at good clockspeed it would be nice,but I don't see 12C at launch as being a problem either or the end of the world.

    Edit!!

    Although AMD could launch a limited run 16C FX though at a higher than normal price and they did this with the Athlon 64.
     
    Last edited: Apr 22, 2019
  3. CAT-THE-FIFTH

    Capodecina

    Joined: Nov 9, 2009

    Posts: 18,175

    Location: Planet Earth

  4. Murphy

    Wise Guy

    Joined: Sep 16, 2018

    Posts: 1,292

    Sure 7nm is going to cost more per wafer than 14nm but where are you getting the much more from?

    Also the 200mm2 die includes a load of extras that aren't part of Zen 2, at least not part of the compute die that's being fabricated on 7nm, either way i don't get the negativity, 70% yields are quiet good on a new node, i would have thought that would have been greeted with a little more positivity. :(

    Personally I'm not expecting 5Ghz, IMO it will be more like 4.5-6Ghz.
     
  5. CAT-THE-FIFTH

    Capodecina

    Joined: Nov 9, 2009

    Posts: 18,175

    Location: Planet Earth

    Look at estimates from generation to generation. GF 14NM was also a licensed version of the Samsung 14NM node,so in reality it wasn't that new. If you look at the Zen 2 picture,the IO is definitely bigger than the CPU die. Also,the substrate has room for another chiplet.So that is 2X70~80MM2 7NM chiplets,and another larger IO die. The total silicon area will be larger than 14NM Ryzen.

    The rumour says 12C at launch not 16C,with consumer 16C being launched later and people are treating it like its the end of the world. People are just overhyping this all too much. Ryzen 2 not only "needs" to be 16C,but also clock close to 5GHZ,run fast RAM,probably have reasonable power consumption,not be hard to cool and also be not more expensive than 12NM Ryzen 2000. I would love it to be all that. It would be an Athlon 64 all over again.

    But,I really don't understand why Ryzen launching with 12C or even 16C being initially gated to a limited edition and expensive FX CPU,being a problem.

    What is more important for many of us,is single threaded performance going up,and memory-CCX latency being improved.

    If anyone of these metrics don't quite get there,you know will happen. I would rather go in with slightly muted expectations and be pleasantly surprised. Even with Zen,look what happened when it didn't clock as high as people wanted,or run RAM as fast as people wanted,they got dissapointed even though it was a good product.
     
    Last edited: Apr 22, 2019
  6. Perfect_Chaos

    Mobster

    Joined: Aug 26, 2004

    Posts: 3,864

    Location: In dem hills

    If it doesn't meet expectations/leaks then i will simply hold off for something better, whether that's from Intel or AMD, doesn't matter.
     
  7. sajtion

    Hitman

    Joined: Jul 2, 2005

    Posts: 818

    it has to be 5gh or no buy
     
  8. ljt

    Mobster

    Joined: Dec 28, 2002

    Posts: 2,808

    Location: Birmingham, UK

    Whats the obsession with 5ghz?

    Surely it's performance is more important than what clock speed it reaches?!? We've had 5ghz AMD CPU's before, GHZ != Performance
     
  9. d_brennen

    Capodecina

    Joined: Jan 30, 2009

    Posts: 15,123

    Location: Aquilonem Londinensi

    It could be 2GHz for all I care, as long as the performance is there
     
  10. SupraWez

    Wise Guy

    Joined: Nov 17, 2007

    Posts: 2,025

    This.
     
  11. Scougar

    Capodecina

    Joined: Jan 30, 2007

    Posts: 12,063

    Location: PA, USA

    Lol. Obsessed with a figure instead of actual performance... Welcome sajtion, to the Pentium 4.... :D
     
  12. subbytna

    Sgarrista

    Joined: Oct 22, 2008

    Posts: 8,642

    Location: Belfast

    Exactly
     
  13. EsaT

    Soldato

    Joined: Jun 6, 2008

    Posts: 5,642

    Location: Finland

    Nope, it's welcome to Celeron and Faildozer.
    https://valid.x86.fr/records.html
     
  14. cliffy

    Gangster

    Joined: Nov 14, 2005

    Posts: 462

    You should have a look around for an old FX9590, came out around 2013 and was the first retail 5Ghz chip. Pretty sure thopugh my 1700x at sto9ck would destroy it though.

    I just dont undertsand the clock speed thing, do people not realise that higher **** speeds does not equal faster chips, single or multithreaded. I some research on IPC should be done!
     
  15. Distracted

    Wise Guy

    Joined: Aug 30, 2018

    Posts: 2,125

    I was thinking 4.7GHz or so, not that it matters given the expected improvements in other areas. They could be running at 2GHz for all I care so long as they deliver on performance, which the demo at the very least implies they should.
     
  16. opethdisciple

    Capodecina

    Joined: May 18, 2010

    Posts: 17,763

    Location: London

    AMD have updated the chipset drivers for the first time since Oct 2018. So something must be inbound soon.
     
  17. Old Man

    Wise Guy

    Joined: Aug 18, 2010

    Posts: 1,520

    Location: Kent

    Wonder how high the price will be?

    The other Ryzen chips (Ryzen 5 2600x) are starting to drop in price :) .
     
  18. chrcoluk

    Mobster

    Joined: Feb 27, 2015

    Posts: 3,204

    If you know how stock turbo works you know it isnt running at 4.1ghz

    MCE doesnt work on those chips either.

    4.1ghz will only kick in when all but 1 core are in C6 state, and that wont be the case when playing games.

    Intel turbo sucks vs XFR.

    There is a reason Intel stopped publishing the turbo clock scaling info as it wasnt putting them in a favourable light with all core turbo speeds been not great. Also £10 to some people is a make or break.
     
    Last edited: Apr 23, 2019
  19. Murphy

    Wise Guy

    Joined: Sep 16, 2018

    Posts: 1,292

    I'd like to start-out by saying i completely agree that people have over-hyped it and set unrealistic expectations, i know why everyone is saying 5Ghz but last time i mentioned it in this thread i got torn a new one. :)

    Expecting the first silicon from a new fab process to have an increase of 600Mhz over an older more mature fab is wishful thinking IMO, however i think we maybe talking over each other or getting our wires crossed as the reason i posted about the 70% yields was to highlight how much more cost efficient it is for AMD than Intel, for Intel making a 32c chip is massively expensive as they may only be getting a 30% yield from one waffler due those 32 cores having to be contained within one monolithic die.

    The 70% yield doesn't really tell us anything about what to expect from Zen 2 but it does show us how much more cost efficient it is for AMD, the rest of what you said i sort of agree and disagree with, 5Ghz is wishful thinking IMO but 16c consumer isn't, not that i understand why someone would want a 16c consumer version. The single thread performance will probably go up but that will mainly be down to architectural changes (things like improvements to the branch prediction, larger µOP cache, increasing the dispatcher and retire bandwidth, doubling of some data pathways to the FPU, and using IF2 that more than doubles the fabrics bandwidth). The CCX latency will still be there it will just be hidden better, not that it should matter much as we're talking about 30 odd nanoseconds.

    I'm going to stop here as otherwise I'll go off on a ramble. :) However i do agree that people have set their expectations way to high.
     
  20. humbug

    Caporegime

    Joined: Mar 17, 2012

    Posts: 30,084

    Apparently its 3.9 All core, but isn't it single core performance that maters in games, isn't the whole point that the controller thread boosts to 4.1Ghz to give you that gaming performance?

    Anyway, Intel don't list the all core Turbo because they base the TDP off the all core base clock, in this case 2.9Ghz, some reviewers have been caught testing performance with MCE and then switching to a low end B360 boards that throttles the CPU to its base frequency giving a power consumption reading within Intel's TDP rating, PCPer and Jayz2Cents to name two, HUB and AdoredTV pulled them up on it. PCPer ignored it while Jay pleaded ignorance claiming he didn't know this.