PC Perspective Podcast #508 – 02/02/18
Join us this week for discussion on Modded Thinkpads, EVGA SuperNOVA PSUs, and more!
You can subscribe to us through iTunes and you can still access it directly through the RSS page HERE.
The URL for the podcast is: https://pcper.com/podcast – Share with your friends!
- iTunes – Subscribe to the podcast directly through the iTunes Store (audio only)
- Google Play – Subscribe to our audio podcast directly through Google Play!
- RSS – Subscribe through your regular RSS reader (audio only)
- MP3 – Direct download link to the MP3 file
Hosts: Jeremy Hellstrom, Josh Walrath, Allyn Malventano, Ken Addison
Peanut Gallery: Alex Lustenberg
Program length: 1:23:16
Podcast topics of discussion:
-
Merch! http://joshtekk.com/
-
Week in Review:
-
AD BREAK: RX Bar
-
News items of interest:
-
0:49:49 AMD Announces Q2 2018 Results
-
0:56:35 Intel Releases Q2 2018 Results
-
Picks of the Week:
-
1:09:45 Allyn: Star Trek Bridge Crew (TNG DLC)
-
1:11:35 Jeremy: I am a fan of big coolers
-
1:15:50 Ken: HD Fury 4K Vertex (Open box)
-
-
Closing/outro
Also there is that Epyc/SP3
Also there is that Epyc/SP3 Motherboard/Platform’s 128 PCIe 3.0 lanes and 8 memory channels per socket so there is a price/feature comparsion that Intel can not currently match. And for both 1P and 2P single/dual socket Epyc/SP3 MB’s that’s 128 PCIe 3.0 lanes and 8 or 16 memory channels for all 1P or 2P Epyc/SP3 platforms. And Intel can not match that PCIe/memory channel feature set currently. So that’s just the Current Epyc/Naples first generation Zen based Epyc platform SKUs. And Zen2 is expected to come with 10%-15% IPC improvment and more than 32 cores per MCM with the Zen2 based Epyc/Rome SKUs in 2019. AMD does not segement Epyc/SP3 SKUs by PCIe lanes or Memory Channels per socket they all get 8 memory channels per socket and 128 PCIe 3.0 lanes.
Also remember AMD’s Opteron had around 23% Server Market share for AMD and Intel was forced to reduce their markups/margins to compete with AMD’s Opteron. Intel’s Gross Margins fell to only around the upper 40s% to lower 50s% range once AMD’s Opteron hit 23% server market share. AMD is a very lean and low overhead operation if it can produce profits with gross margins in the 33%-37% range and Intel needs to keep its gross margins from falling below 55% as Intel has the chip fabs costs and other costs to cover that AMD does not have to worry about.
AMD is forcasted by Lisa Su to have single digit server market share of around 5% by the end of this year and be into the double digits 10% above by years end 2019. And AMD has closed enough Epyc deals already that where just waiting for final Certification/Vetting of AMD’s Epyc/Naples paltform by those server customers. So those deals are now going forward now that Epyc/Naples has made the grade and is now in use for production server workloads.
Lisa Su and AMD’s BOD had aleready made the choice to Focus mostly on the Higher Margin Server/Professional CPU/GPU compute/AI business and those Vega 10, and soon Vega 20, based Compute/AI SKUs are and will be netting much larger markups for GPUs than any consumer GPU sales can even hope to achieve.
Intel’s AVX can not Match Vega 10’s, or Vega 20’s, Floating Point TFlops metrics and AMD was smart to design Zen with less power hungry AVX units and let the majority of their server customers, who do not need heavy AVX, save power instead. So AMD will offer any potential Epyc based HPC customers their GPU Compute/AI Radeon Pro WX/Instinct accelerator products. The majority of the server market workloads do not need or utilize heavy AVX so AMD would be smart to keep Zen’s AVX units smaller and offer heavy FP performance Via AMD’s Professional Compute/AI GPU SKUs.
Vega 20 is gitting AMD back UP to that 1/2 DPFP to SPFP ratio so Vega 20 will be a DPFP TFlops leader for those that need the FP grunt. Intel currently has no competition to AMD’s Vega or Nvidia’s Volta in the market place and that’s years of hard work for Intel to catch up.
Opteron was no where near as performant as the Zen Based Epyc/Naples SKUs and still AMD used that lower Opteron Pricing to take a Price/Performance 23% share of the server market more than 10 years ago with Opteron. I do not Know if each percentage point of server market share represents millions or hundrends of millions of dollars of revenue potential but AMD has to be getting many millions in new Revenues with each percentage point server market share gains. And that AI/Infrencing market means with each Epyc CPU sold there may be the potential to sell 4 Radeon Instinct MI25’s like was used for that Project 47 Refrence inferencing supercomputer that AMD/Server Partner showed off last year. The AI/Inferencing Market looks to be ready to grow more than only the traditional server market over the next 5 years.
So AMD’s is really been under promising but over delivering over the past few years, 41% projected IPC gains for Zen1 turned out to be around 52% IPC gains once Zen1 was to market. Zen/Epyc single scoket(1P) performance is bettering some Intel dual socket(2P) configurations in general server workload benchmarks. Zen2 is being sampled so hopefully some of that performance information will leak out shortly to verify that rumored 10%-15% IPC gains for Zen2.
AMD’s got a hold of a pot full of server CPU gold and some damn good AI/Inferencing GPU IP that Intel Currently lacks, Raja or not. Because GPUs take some years to develop and certify for professional workloads.
A not relatet to Nvidia’s
A not relatet to Nvidia’s Huang is AMD’s Fenghuang! An APU with:
“The mysterious AMD Fenghuang is a complete system on chip (SoC) powerhouse made up of a quad-core Ryzen processor paired with 24 Vega Compute Units and 8GB of GDDR5 memory on a single chip.” (1)
GDDR5 on a signle DIE, maybe that’s not what they meant but really AMD’s making more semi-custom Revenues also now inaddition to XBONE and PS4.
“AMD’s Mysterious Fenghuang SoC Spotted in Chinese Gaming Console”
https://www.tomshardware.com/news/amd-fenghuang-ryzen-vega-soc,37551.html
Here’s a quick question. What
Here’s a quick question. What ever happened to the steam game pad? There was quite a bit of hype, especially about the tactile response thumb pad. Did pcper ever get a test sample? did it just not work so it was buried? Did they just kinda get high and walk away?
Seems like this was going to be a big deal then it just isn’t.