PC Perspective Podcast #528 – RTX 2060, EVGA Nu Audio, and our CES Roundup

3 Comments

1. 

   delcolux on January 12, 2019 at 10:41 pm

   Thank you Jim for the lovely
   Thank you Jim for the lovely outro sequence.

   Reply
2. 

   BigWigsQueryAvalancheForReleaseDay on January 13, 2019 at 2:09 am

   OK let’s get this started!
   OK let’s get this started! What are your projections for Radeon VII’s FineWine(TM) on that 7nm TSMC process node?

   Also I can see a lower Binned variant of that Vega 20 Die coming in the form of something closer to Vega 56’s complement of shaders:TMUs:ROPs(64). So what price range for that and what will be its main Nvidia RTX competition?

   Vega 20’s new set of AI based ISA ententions on Radeon VII/And Any Lower consumer Vega Die 20 Bins! Will AMD be working up some form of AI based upscaling similar to Nvidia’s DLSS and how will those Vega 20 only AI ISA extantions help with that.

   Continuing on, for AMD’s xGMI(Infinity Fabric) IP, will there be the possibility of some xGMI capabilities for Consumer/Vega 20 variants and how will that compare to Nvidia’s Consumer NVLink that probably will have some features disabled on Nvidia’s non Quadro/Professional GPU SKUs.

   AMD’s Infinity Fabric on the 2p Epyc/CPUs makes use of PCIe’s PHY(Physical Hardware layer) that switches out the PCIe protocal/siginaling with the Infinity Fabric protocal/siginaling on Epyc/Naples and Epyc/Rome. Is it possible that on the 500 series AM4 motherboards that some MB makers could certify that method between Vega 20 Consumer variants for select x16 MB slots on any high end AM4 motherboards for any sort of alternative xGMI/Bridge Chip Hardware usage between some cosnumer Vega 20 variants.

   What method do you think will be available for consumer Vega 20, the more attractive xGMI bridge or the PCIe PHY method on select 500 series AM4 MBs. It’s possible that AMD will not make any use of xGMI on Consumer Vega 20 and just stick with PCIe 3.0 or 4.0(500 series AM4 MBs) what do you think is more likely.

   I’d like to see some consumer/prosumer Vega 20 variants that offer xGMI(Infinity Fabric) bridge chip IP to interfave 2 Vega 20 GPU variants vai the Infinity Fabric’s Cache Coherency/Single VRAM Memory Address space adressing, interface for 2/More Vega 20 based Consumer/Prosumer Variants more tightly Infinity Fabric linked to act more like one single larger logical GPU.

   Also, how long after there is a Vega 20 based lower binned DIE variant with closer to Vega 56’s complement of Shaders:TMUs:ROPs before AMD specs out some Dual Vega 20 DIEs/Single PCIe card variant. Will AMD do this and will AMD make use of xGMI(Infinity Fabric) Protocols instead of simply PCIe lanes between the Dual Vega 20 DIEs(Binned to each have similar to Vega 56’s complement of Shaders:TMUs:ROPs)!

   AMD currently has a dual Vega 10 DIEs(Each Vega 56 Like)/single PCIe card V340 cloud gaming variant that also has that new hardware based GPU virtulization management ASIC. I’ll bet that AMD will be replacing the Vega 10 based V340 variant with a Vega 20 based variant.

   These are some question that will probably have to wait for the NDAs to totally lift but it’s good to be ready with as many questions as possible to ask the Big Wigs on launch day!

   Reply
3. 

   CarryTrainer.Editor on January 15, 2019 at 6:57 pm

   Well done outro. I digged it.
   Well done outro. I digged it.

   Reply

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