Podcast #519 – Core i9-9900K, Changes at PCPER, and more

4 Comments

1. 

   collie on October 25, 2018 at 4:49 pm

   I’m kinda sad I wasn’t able
   I’m kinda sad I wasn’t able to be there live for ryan’s last cast. But that was a sweet send off, You will be missed mr Shrout, I cant even begin to list the questions you’ve answered for me over the years, You and your sight are the reason my pc is so beautiful and powerful I just like to look at it some times. I am sad to see you go, I am very happy for your new oprotunity, and I wish you and your family, all the best, you will never be replaced.

   F

   Reply
2. 

   ServerMarketIsCrazyGoodLatelyInItsQuarterlyIncreasingTAM on October 25, 2018 at 11:01 pm

   Intel’s got more record Q3
   Intel’s got more record Q3 revenue numbers and that Intel CPU shortage has certianly increased Intel’s Quarterly Gross margins up at 64.5% compared to 62.3% in Q3 2017. AMD also has increased its Gross Margin numbers to 40% in Q3 2018 up from 36% in Q3 2107.

   So it’s looks more like a rising Server tide lifts all boats with the server market’s TAM having also rapidly been increasing to record levels over the last 18 months.

   Let’s see how Q4 2018 works out for both Intel and AMD in the server market if AMD can continue increasing its Epyc server market share numbers 1, 2 or even 3(harder to do) percentage points each quarter. The server market TAM Pie is getting larger so each of those 1% incermental server market share gains that AMD can eak out will account for a relatively larger amount of revenues than it would have amounted to only 2-3 years ago when the Server Market was not growing in TAM numbers as quickly.

   AMD Vega 20 will most definitely be needed for any Radeon Pro WX and Radeon Instinct GPU compute/AI market sales paired with Epyc CPUs for the fast growing AI market and the overall server market’s rapidly expanding TAM sure helps any and all server processor makers. AMD just needs to continue with its strategy with first generation Zen and carry that forward with Zen2 and Vega 20 for the Professional Compute/AI market where AMD can continue to see the largest SKU markups relative to any consumer market where the markups and margins are just too small and sales/revenues are not as stable quarter to quarter.

   Folks can see where Intel is earning those record numbers and that’s not really something any consumer CPU market can support. So AMD just needs to carry on and little by little Epyc/Radeon Professional will begin to earn larger and larger amounts of average quarterly revenues beacuse that’s where the real money is.

   Intel spends a lot of its money in the Consumer/Laptop market on incentives and engineering assiatance to the OEMs. And even though AMD’s Vega Integrated Graphics is better performing than Intel’s Integrated Graphics the laptop OEMs will expect incentives and technical assistance from AMD like they can get from Intel so AMD will have to increase its incentives to get wider Zen/Vega Raven Ridge Adoption. Intel earns so much more than AMD currently that AMD has to start competing on the Laptop OEM incentive front in addition to the APU performance front with Intel.

   Intel is just so much larger and there are other market factors in play besides just AMD/Zen Mobile APU to Intel/WhateverLake Mobile CPU performance numbers competition. Intel also invests in low power Laptop/mobile display technology and Thunderbolt technology as well as XPoint Tehnology that can create more value added in the Mobile/Laptop market. So AMD has to begin to try and match Intel’s investments towards Laptop and Mini-Desktop devices extra OEM engineering assistance or AMD will not earn the numbers of laptop design wins relative to Intel. AMD will have to adopt USB 3.2 sooner on it’s chipsets if that Promised Opening UP of any Thunderbolt technology to a wider market of non Intel makers does not pan out. AMD will also have to work on more non APU ralated Laptop sorts of enhancement IP like Laptop Display power saving IP.

   AMD really needs to double down on either Mobile/Vega or Mobile Navi even more so than Desktop GPUs and that should include Zen/Vega or Zen/Navi APUs with at least 2GB or 4GB of HBM2 and enough bandwidth to really provide the integrated graphics with the needed performance boost. In addition AMD has not introduced any discrete mobile Vega/newer GPUs with HBM2(2GB-4GB). And that’s really a shame because Vega’s HBCC/HBC IP has more potential on laptops where memory is not as fast as on desktop and Laptop OEMs are usually not providing for any memory overclocking abilities on laptops anyways. So Vega’s HBCC/HBC IP and the ability to make use of HBM2 like a last level GPU cache would really shine on discrete mobile GPUs as well as APUs(Even at 1GB or 2GB of HBM2).

   TSMC’s 7nm should help AMD in some areas but AMD still has to engineer like its life depends on it and that really the truth up until AMD can increase its Server, and other, market revenues enough to compete on the incentives and OEM engineering assistance front with Intel and Nvidia.

   Reply
   • 

     pdjblum on October 26, 2018 at 7:55 am

     epyc 2 will be much harder
     epyc 2 will be much harder for all the intel entrenched companies to ignore; share gains will accelerate; margins will increase; all will be well

     fuck me, just realized, intel needs a ceo, intel loves to steal stars from amd to run their company, so it makes sense they go after the best ceo around, lisa

     stay the fuck away from her intel

     if ryan can get the job he got at intel, certainly josh can be ceo, so leave lisa the fuck alone; she is happy at amd

     Reply
3. 

   pdjblum on October 26, 2018 at 7:48 am

   most excellent dudes
   most excellent dudes considering it was your first official podcast

   hope you keep the same configuration with jim as MC and ken at the helm

   maybe one awesome consensus pick of the week rather than mostly asides

   makes sense if you are vaping

   anyway ken, know you are going to be great

   Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.