PC Perspective Podcast #498 – 05/03/18
Join us this week for discussion on Microsoft Surface Book 2, Intel 905P Optane, 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: Ryan Shrout, Allyn Malventano, Jeremy Hellstrom, Josh Walrath
Peanut Gallery: Ken Addison, Alex Lustenberg
Program length: 1:31:26
Podcast topics of discussion:
-
Merch! http://bit.ly/pcpermerch
-
Week in Review:
-
Casper
-
News items of interest:
-
1:09:00 Seven brown new fans from Noctua
-
1:13:30 Jim Keller joins Intel
-
Picks of the Week:
-
1:21:20 Ryan: Oculus Go
-
1:22:20 Jeremy: Great deal on a B350
-
1:23:15 Josh:New and Improved and On Sale!
-
-
Closing/outro
F4 Falcon? The F16 is the
F4 Falcon? The F16 is the Falcon. The F4 is the Phantom, sometimes also called the “Iron Pig” by ze Germans 😉
Oh man, did I really say
Oh man, did I really say "Falcon"? What the hell is my brain doing.
So basically a company with
So basically a company with only a few million profit (usually just losses) is beating the #$@#@ out of a company with Billions of profit. Mmm, this is interesting and quite embarrassing for Intel I’d say.
Intel needs a high fat/high
Intel needs a high fat/high margin diet that’s currently around 63% gross margins while AMD can exist on a few bread crumbs and show an 11 cents per share profit on just 36% gross margins as of its latest Quarterly Statment. AMD is a Fabless chip maker so no high overhead chip fab upkeep to worry about with GF and TSMC/others able to spread that expensive Billions in Chip Fab upkeep, billions in chip fab equipment, and process node R&D that also costs billions around on a better economy of scale across the entire fabless chip industry of CPU/GPU/Other Processor designers that make use of the third party chip fab industry.
So AMD can now start to show some excess revenues remaining after paying all its quarterly bills and actually show a profit on a low gross margins rate of 36%. And AMD can do so at such a small gross magins rate that Intel would never be able to subsist on with Intel’s sky high overheads required for that expensive Chip fab ownership, fat middle management pay perks, and other overall expenses that have to be managed on a quarterly basis.
AMD’s Epyc Price/Performance metrics totally outshine Intel’s and even AMD’s older Opteron SKUs that in their day garnered AMD around 23% server market share and pushed Intel’s gross margins below 50% at one point in time. That low of a gross margin rate pushed Intel’s share value and market cap downwards as Intel had to lower its Xeno per unit markups to keep Opteron at that time from takeing even more market share.
AMD’s Epyc is much closer in IPC to Intel’s Xeon than Opteron ever was and Epyc’s price/perrmance metrics are even more of a threat this time around with that Epyc/SP3 Motherboard feature set with 8 or 16 full mamory channels, 1P amd 2P MBs respectively, that regardless of their 1P or 2P socket configuration offer the same available 128 PCIe lanes at a standard cost.
Intel can not pivot fast enough to fully be able to compete with Epyc’s Price/Performance and Price/Feature metrics that also include a much lower initial purchase cost for the Epyc/SP3 MB hardware relative to Intel’s higher markup/margin server Kit. And That Intel Kit comes in Intel’s dizzying array of very segemented offerings that vary by CPU and MB(PCIe lanes, other Features) at increasing costs and is very segemented/binned so that every last little of performance/feature sets comse at that increasing cost to any current or potential client.
Intel can not afford to quickly be able to drop its higher per unit markups/margins without drastically cutting its fixed costs and reducing any fixed costs will take Intel time and money to achieve. One need only look at Intel’s gross margin historcial charts to see the affecr that AMD’s, less performant than Epyc, Opteron had on Intel’s overall gross margins as that Opteron business took more server market share, and Intel has to be ready for AMD’s Epyc to have a potentially much worse affect on Intel’s Much Needed higher gross margins that are required for Intel to remain profitable and are currently above the 60% gross margin metric and can not sfford to be going below the 55%-50% rate for too long a period before Intel will have to quickly make some drastic cuts to its very high overhead and high fixed costs business operations.
Wall Street is very sensitive to any Gross Margin Basis Points that are trending in the downward direction and after looking at AMD’s Opteron market affect on Intel’s margins, the Epyc threat to Intel’s higher margins can not be looked at too lightly with a proper historcial market perspective. And that is knowing as everyone knows now about Zen/Epyc’s much better performance metrics and that lower across the board pricing that AMD maintains for Epyc the same that AMD maintained for the less performant than Epyc, Opteron line of server CPU SKUs that so drastically affected Intel’s margins back at that time.
Intel is going to be forced to reduce its markups and maybe even its vast array of very segemented by prices offerings at some point in time just to combat Epyc’s Price/Performance advantage in the market place over the next several business quarters and beyond because AMD is not sitting still and Zen2/Zen3 at 7nm with Epyc Rome/Milan SKUs possibly offering 64/128 Cores/Threads per socket.
AMD’s current Zen/Zeppelin modular scalable die is comimg off the diffusion line at better than 80% die/wafer yields at an enonomy of scale that Intel’s lager monolithic CPU die designs can not match. So that’s just more pricing latitude for AMD to increase Epyc’s Price/Performance metrics above anything that Intel can currently afford to do currently without Intel making some very painful cuts to its high overhead ooperations that come with very high fixed costs.
Here is the ThunderX2 Custom
Here is the ThunderX2 Custom ARM core(1) with SMT4 capabilities from Cavium(Soon to be part of Marvell) that was created using that sold off by Broadcomm Vulcan 64 bit core design instead of being based on Caviun’s original ThunderX IP.
So Cavium purchased the Vulkan core IP from a Broadcom that was acquired by Avago(Very dumb for selling that Vulkan ARM core IP to Cavium by the way) and now Cavium(Marvell) has a very nice 32 core custom ARM ThunderX2 design based on the old Broadcomm Vulkan custom ARM core IP, and that’s 32 cores and 4 threads per core on those ThunderX2 SKUs.
“The new bits of data about the Vulcan ThunderX2 cores include the fact that the Vulcan chips will support SMT modes with 1, 2, or 4 threads active per core, and Hegde says that the two-thread SMT boosts throughput by about 20 percent and the four-thread SMT pushes it up by around 30 percent. That’s about as good as Intel gets with two-thread HyperThreading on its Xeons and about as good as IBM gets on four-thread SMT on the Power8 and Power9 chips, although IBM can push it up to eight threads per core if customers want to do that on the Power8 chips and certain models of the Power9 chips and get more oomph for workloads that are amenable to lots of threading – think web application servers, Java and PHP runtimes, and databases.” (1)
Watch out Intel and AMD(Better keep that K12 IP handy) as there is still some players that will remain after all this wave of consolidations continues in the AMD makers market. And Broadcom wants to acquire Qualcomm also but really Avago made a rash decision selling off Broadcom’s Vulkan Core design and IP after it acquired Broadcomm.
(1)
“ThunderX2 Arms Hyperscale And HPC Compute”
https://www.nextplatform.com/2018/05/07/thunderx2-arms-hyperscale-and-hpc-compute/
Edit: Vulkan
to:
Edit: Vulkan
to: Vulcan
Vulkan is a Graphics API.
Edit: AMD makers market
To:
Edit: AMD makers market
To: ARM CPU makers market
my proof reading sucks!
It would be nice to see
It would be nice to see female podcasters 🙂
You missed the years where
You missed the years where Colleen was running the board. You aren't wrong though, just hard to find people who will put up with us for any length of time.
Ryan,
We’re getting close.
Ryan,
We’re getting close.
https://www.pcgamer.com/add-a-large-capacity-2tb-solid-state-drive-to-your-pc-for-dollar269/