Podcast #466 - ECS Z270, Clutch Chairz, AMD market share, Lenovo Yoga, and more!

Subject: General Tech | September 7, 2017 - 09:46 AM |
Tagged: z270, Yoga 920, Yoga 720, video, Threadripper 1900x, superfish, skylake-x, podcast, Lenovo, IFA 2017, HP S700 Pro, GTX 1080, gigabyte, ECS, Die shot, Core i7-6700K, Core i5-6600k, Clutch Chairz, Aorus X5, amd

PC Perspective Podcast #466 - 09/07/17

Join us for discussion on ECS Z270 motherboards, Clutch Chairz, AMD market share, Lenovo Yoga, 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: http://pcper.com/podcast - Share with your friends!

Hosts: Ryan Shrout, Josh Walrath, Allyn Malventano

Peanut Gallery: Ken Addison, Alex Lustenberg

Program length: 1:15:50

Podcast topics of discussion:
  1. Week in Review:
  2. News items of interest:
    1. 0:25:05 Casper
  3. Hardware/Software Picks of the Week
    1. 1:09:10 Allyn: FolderTimeUpdate
  4. Closing/outro

Source:

September 7, 2017 | 11:56 AM - Posted by Anonymously Anonymous (not verified)

Josh's O face confirmed

September 11, 2017 | 09:29 AM - Posted by Poker face (not verified)

Pucker up Josh!

September 7, 2017 | 12:19 PM - Posted by Power (not verified)

Intel Skylake-X 18-core Die looks glued together.

September 7, 2017 | 12:58 PM - Posted by 80VeryCherryVega10Dies (not verified)

Sure AMD is gaing market share among the home system builders markets and AMD's Server market share has no where to go but up with that professional market just getting rebooted with Epyc, where AMD stands to gain the most of its revenue share, including the Radeon Pro WX and Radeon Instinct professional comput/AI GPU SKUs.

It's not the miners that are gitting the first dibbs at the best/cherry Vega 10 dies it's that Project 47 beast that is running 20 EPYC SoCs and 80 Radeon Instinct MI25 accelerators, so that's some 80 Vega 10(Very Cherry) dies that have the best thermal/power usage metrics that are then clocked lower to achieve the best TFlops/Watt performance: "1 PetaFLOPS of compute power at full 32-bit precision delivering a stunning 30 GigaFLOPS/W"(1). And the AI market is expected to grow by the billions of dollers over the next few years. So those Vega 10 compute/AI SKUs are getting the Vega 10 dies used by for the professional SKUs with the remaingng consumer Vega 10 dies getting into the retail channels to be snapped up at above MSRP.

AMD's Professional Epyc and Radeon WX/Instinct SKUs are getting the priority as that market will pay the higher markup and Raja Did say(see his twitter statments) that Vega 10 is designed to go up against GP100/GP102 as well as GP104 so that's one base Vega 10 micro-arch that hase to go against 3 Nvidia professional/prosumer/consumer SKUs as AMD lacked the resources to make more than one base GPU micro-arch when Vega was being developed.

The market where AMD has had/will have the roughest time will be the OEM laptop market where AMD needs to get some OEM partners that will not gimp their laptop SKUs for memory channels(starve the integrated Graphics for bandwidth) or gimp feature offerings(Screen Resolution/Etc.) on the OEM's AMD APU based laptop SKUs. AMD also needs to try and get into the mini form-factor desktop market with some Bristol Ridge/AM$ updatable to Raven Ridge/AM4 based Designs with those mini from factor motherboards that are popular for some usage in the TV/Living room sorts of computing for streaming of movie/gaming content.

Some Websites are ignoring AMD's Bristol Ridge offerings that where released last year for OEMs only that have just now become available for the home system builder market. And the Motherboard Makers are missing out on a market of mini/micro form factor SKUs that can be built right now using Bristol Rodge and updated later when the Raven Ridge APUs start arriving late 2017(mobile) and first quarter 2018(Desktop).

So there are plenty of folks on reddit doing their own testing with the A12-9800(bristol Ridge/AM4) and other final Bristol Ridge excavator SKUs with their pre-Polaris based "RADEON R7/GCN 3rd Gen" graphics. So AMD really should have maybe tried to get some mini desktop OEM/non motherboard mini/micro from factor design wins even with Bristol Ridge as that can be updated later to Raven Ridge. AMD needs to try and notice these niche markets that are now not so niche if you look at Intel's NUC/mini desktop OEM offerings.

(1)

"AMD Powers Project 47"

https://instinct.radeon.com/en-us/p47-announcement/

September 7, 2017 | 02:08 PM - Posted by ProcessNoNodeNews (not verified)

10nm woes, and another customer gets away from Fabzilla!

"Intel foundry customer bails out
Another partner silently goes AWOL"

https://semiaccurate.com/2017/09/06/intel-foundry-customer-bails/

September 7, 2017 | 06:58 PM - Posted by TheWholeDevicesMarketSmellsSuperFishy (not verified)

That Lenovo Superfish legal Judgment requirements needs to become a standard across the PC/Laptop/Phone market with every OEM/Carrier(Phone/other) required to submit to some OS/firmware/CPU(embedded security processor) screening to protect the public's privacy. You hear that Google! Lots of SuperFishy things going on below the Ring-Zero OS functionality that is baked into the CPU's hardware/firmware(micro-code) that needs to be audited very closely by more than TLAs!

Oh and Google looks to be trying to acquire HTC's phone business. So will it just be another Motorola Redux for the Search Engine Colossus. Maybe the 2nd time around will take and the Folks in Redmond will be watching because they do Know something about platforms burning. But at least Google has its Android O platform out Now that is still in need of some tweaking and a move towards some proper non carrier limited methods of pushing out security updates. The Android "OS" still needs more features that put the phones owners in the driver seat, and I do not mean cars and such.

"Google is very close to buying HTC's smartphone business"

http://www.businessinsider.com/google-to-buy-htc-smartphone-business-rep...

September 8, 2017 | 01:45 AM - Posted by Lorash651

If the podcasts were ever named, this one would surely be called "Pucker up, Buttercup"

September 8, 2017 | 05:46 AM - Posted by Br01 (not verified)

I think more than one of you are wrong.
This card can be used in SLI if it's stated in the manufacturers documentation.

September 8, 2017 | 05:58 AM - Posted by Br01 (not verified)

Why do we need more power?:)

September 8, 2017 | 06:17 AM - Posted by Br01 (not verified)

PS. Alan is The shiznic:) Love im.

September 9, 2017 | 05:52 AM - Posted by RandomGuy (not verified)

Ryan, Actually 1900X have 4-0-4-0 cores/ccx configuration acording to TPU (can't post link cause spam filter...)

Check out cache sizes - 1950X/1920X have 38-40MB cache (L2+L3) while 1900X has only 20MB - because there are only two CCX active so L3 is 8MB+8MB=16MB and rest is L2 (4MB - 8x512kB)

September 9, 2017 | 04:18 PM - Posted by FuzzyMethodsNot (not verified)

Seeing quite a few Riddit posts lately still pushing that single core IPC issue and refrencing tests on software that may not have been optimized for the Zen/Ryzen, Zen/TR's, or Zen/Epyc CPU SKUs. So folks need to be called out for not listing the testing software's build numbers and what SDK/Compiler/compiler build flags that were used for the testing software refrenced in the quoted testing runs.

I know that Phronoix does list information like compiler/SDK/IDE used and compiler flags/setteings for its Phronoix benchmarks but some of the windows testing results need to have the same compiler/SDK/IDE compiler flags/setteings listed that where used or the testing is qoing to have to come with a big asterisk. The scientific method appears relatively unknown to many gamers when they go about coming to conclusions whith out proper vetting for unknowns!

Some Applications have not yet been fully optimized and most likely they will not be optimized fully for a good while until the whole software ecosystem can be brought up to speed for the Zen micro-arch based deigns.

Also folks trying to sell you on Threadripper platform for workstation workloads are probably benefiting financially more from Threadripper/Assoicated Products sales that any Epyc/Assoicated Products sales. And AMD's Epyc platfrom can counterintuitively be the better feature for feature deal for workstation workloads when the cost per core metric comes into play. For example, for the 24 core/48 thread Epyc 7401P($1075) costing only $76 dollars more than the 16 core/32 thread Threadripper 1950X($999) with the Epyc 7401P costing $44.80 per core and the Threadripper costing $62.44 per core.

The workstation Epyc motherboards may cost more than most gaming motherboards but the total cost of the motherboard divided by say the number of PCIe lanes offered can also come out in Epyc's favor even with an Epyc motherboard that costs $200 more than a X399 TR motherboard because the Epyc motherboard offers twice the PCIe lanes(128) over the X399 TR MB's 64, Ditto for the cost of the respective TR and Epyc motherboard/cost divided by the number of memory channels offered with the Epyc motherboards offering twice the number of memory channels(8 memory channels) than the TR/X399 motherboards'(4 memory channels).

So its better to look at any Platforms' cost per feature analysis over simply the overall cost becaues, as counterintuitive as it appears, AMD's Epyc workstaion single socket solutions can over more features per dollar than AMD's Threadripper platform for folks looking to do mostly workstation workloads. AMD is actually the inverse of Intel in that regard with AMD's Epyc platform being feature for feature an even more affordable platfrom than AMD's consumer SKUs for workstation workloads, and Epyc is super affordable compared to both Intel's consumer and Xeon platform choices.

AMD is the Anti-Product-Segementer compered to Intel's Pro-Product-Segementer offerings with Intel charging extra for more PICe lanes, Need RAID functionality(RAID Key Purchase Required to unlock RAID features), etc. AMD's Consumer and professional platforms offer more of the same features acoss all their respective platfrom offerings compared to Intel with its charge extra and above an already overpriced product stack that cost so much already top to bottom for Intel's product lines.

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

By submitting this form, you accept the Mollom privacy policy.