Podcast #521 – Zen 2, 7nm Vega, SSD Vulnerabilities, and more!

7 Comments

1. 

   mammal on November 8, 2018 at 11:29 pm

   We love you Jim you sneaky,
   We love you Jim you sneaky, beautiful bastard.

   Reply
2. 

   unclelala on November 9, 2018 at 2:00 pm

   Was that a 3 second cameo by
   Was that a 3 second cameo by R.S.?

   Reply
3. 

   Brother Michigan on November 9, 2018 at 6:54 pm

   It definitely sounded like
   It definitely sounded like Jim took a hammer to his laptop or something on the podcast. Had to stop and go watch the stream to see what happened.

   Reply
4. 

   Vesperan on November 9, 2018 at 9:30 pm

   When Ryan left I was worried
   When Ryan left I was worried the podcast would suffer.

   I missed Jim’s debut last week, but I loved his presence this week – removes any fears I had about Ryan’s departure.

   Reply
5. 

   Benjamins on November 12, 2018 at 6:48 pm

   I think AMD will have 1 CPU
   I think AMD will have 1 CPU die, 2 IO dies (Server/HEDT, mainstream)

   so mainstream will have its own IO die with 1-2 CPU dies, and the APUs will be the same IO die with 1 CPU + 1 GPU die.

   Giving us 16c for mainstream (they will ship 12c this gen), and a 8c + Vega/navi APU.

   Reply
   • 

     ToEarilyToKnowAtThisPointInTheSchedule on November 12, 2018 at 8:13 pm

     If you want to game it’s
     If you want to game it’s better to have the memory controllers on the CPU Core Die unless of course there is some L4 cache on the I/O die to help reduce some overall I/O to memory latency.

     Every jump over any connection fabric adds latency and there is no getting around that. Now if the games where written in a manner that has the Main gaming logic and Draw Call Code Pinned to a specific Core’s Processors/Threads and a specific CPU’s Die then it really does not matter as much if the memory controllers are on a different central I/O die, provided that CPU has a larger L3 cache and the I/O Die has a big wad of L4 Cache that’s much lower latency than any actual DRAM/Memory access latency.

     It’s not going to be necessary for AMD’s Consumer/Ryzen Zen2 based CPU MCM Layouts to match any Professional Epyc/Rome MCM Layouts. I’m sure any of AMD’s Zen2 micro-arch based APUs will still follow the same design as previous APUs but it would be nice to see some L4 cache or better yet some on module HBM2 for the APU’s Integrated Vega/Navi Graphics.

     AMD will obviously have the funds to maybe bifurcate its consumer and enterprise CPU MCM layout designs to focus better on either gaming or server workloads. For consumer Gaming oriented CPUs more L3 cache may just be the solution along with some actual large L4 cache to reduce the need for any direct to memory accesses on crucial gaming logic that needs the lowest latency possible.

     But we will not know until 2019(sometime in 2019) what AMD will be actually doing on its consumer Zen2 based mainstream consumer CPU and APU variants. AMD’s cosnsumer HEDT variants may look a little closer to what Epyc/Rome looks like but AMD has options there also to engineer more gaming foucused solutions.

     Reply
6. 

   PolarisAndVegaDiscreteMobileForEveryoneElseBesidesApple on November 13, 2018 at 7:28 pm

   Here is a question for PCPer
   Here is a question for PCPer once the RX 590s are up for testing and they probably are already in testing/already mostly tested waiting for the NDA to Expire. Remember back when Polaris 10 was released and Raja was showing those Dual Polaris 10 GPU/CF benchmark figures.

   Would it be possible for PCPer to revisit those Dual Polaris Benchmarks but with Dual RX 590s and revisit those Raja benchmarks. It’s just to see how much the Polaris GPU micro-arch and process node tweaks as well as driver and gaming ecosystem improvments have increased Polaris based GPU performance now compared to day one back in july 2016.

   This will most likely be that last Polaris refresh as Vega has replaced Polaris and Navi is expected next year to replace any mainstram AMD GPUs like Polaris. Who Knows if any new Vega Based Gaming SKUs are planned but it sure looks like Vega 64 and Vega 56 will still be around in the AMD higher end for a while longer while Navi will for sure be replacing Polaris in the mainstream.

   Also will there be any Disrete Vega Mobile/Laptop SKUs outside the Vega Mobile SKU with 4GB of HBM2 was announced for the Apple Macbook refresh SKUs? That Kaby Lake G semi-custom “Vega” mobile is for Intel’s usage only and that first Discrete Mobile Vega GPU only SKU with 4GB of HBM2 for Apple’s Macbooks is probably only an Apple offering.

   I’d like to see more Vulkan/DX12 Explicit-Multi-GPU-Adaptor capable games testing also with the RX 590 SKUs to see if that New DX12/Vulkan Graphics API feature can work better for dual GPU gaming that the old CF/SLI methods.

   Reply

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