Intel Kaby Lake + AMD Radeon product *spin-off*

  • Thread starter Deleted member 13524
  • Start date
It looks like PlayReady 3.0 is already implemented in Vega M's driver, since in AT's review they were able to do Netflix 4K HDR playback.


Maybe they're more concerned with it playing rips from UHD BluRays? Though that all depends on what re-encoding is done with the video/audio payloads.
PAVP is no longer a concern once they're dealing with ripped content... and Vega handles HEVC decoding just fine. The only problem is the digital content protection system that's in the UHD blu-ray itself.. which the NUC can't read anyways because there's no UHD blu-ray optical drive in there.

How many Hades Canyon buyers were planning on purchasing $200 USB UHD blu-ray readers? Honestly, it really doesn't look like a large number to me.
Especially considering stand-alone readers cost practically the same and you don't have to go through the hassle of setting up a special software within windows to read UHD blu-rays.

Wouldn't they be better off using Intel's GPU for playback anyways? That should be the more power efficient solution for video playback and avoid firing up the dGPU. Vega is efficient, but Intel's solution should still be better in most cases. Vega only makes sense if some app is using SSG style behavior.
Yes, but as @Malo said, the outputs are all hardwired to the Vega M.
I'm guessing laptop implementations will have all outputs going through Intel's IGP, but in that case they won't be able to get those whopping 6 video outputs (2xHDMI + 2xDP + 2x DP-through-thunderbolt), which is one of the factors that make Hades Canyon more interesting.
 
Why does a 6 display output wired only to the Vega make more sense than 2+4?
 
Why does a 6 display output wired only to the Vega make more sense than 2+4?
So that all display outputs can access the high-performance GPU and to avoid messing with the DRMs that require a direct connection between video decoding block and display.
 
Why does a 6 display output wired only to the Vega make more sense than 2+4?
In the case of a NUC, saving the ~10W idle power of the Vega or making use of the IGP would be less of a concern as battery life won't be an issue. For a laptop or tablet that is a significant savings and I could see the split returning.

I'd rather have seen the split remain and ability to toggle/remap the outputs with a button. Build in some basic KVM functionality. Having separate GPUs should allow for GPU passthrough with a hardware VM which would be a rather nice feature to have for security. Take a page from ChromeBook/SecureBoot and even have a switch/button to write-lock a SD card with the primary OS. Be able to run linux, chrome, etc and fire up a windows game inside a VM securely with most of the performance. Far more useful if playing older or mobile focused games as they are less demanding relative to current hardware and sandboxing everything at the same time.
 
pcworld apparently thinks „
Intel's Kaby Lake-G mystery: Why Radeon Vega M may be more Polaris than Vega
https://www.pcworld.com/article/326...eon-vega-m-may-be-more-polaris-than-vega.html


If there's no RPM in Vega M, Intel's response to PCWorld might get them in trouble...?

Intel said:
It is similar to the desktop Radeon RX Vega solution with a high bandwidth memory cache controller and enhanced compute units with additional ROPs (Render Output Units).


To be truthful, Vega M GH/GL's compute throughput ratio with everything else (fillrate, bandwidth) is quite anemic compared to all other Polaris or Vega solutions. If they also cut down RPM then the GPU's performance is going to hurt even on optimized titles.

Another thing to consider is that this is most probably the exact same chip as the 28 CU discrete Vega M for laptops, and AMD probably just put in there an interface for both an interposer and Intel's EMIB.

Tomshardware confirmed the featureset to be the same as Vega 10 and Raven Ridge's Vega. Which makes even less sense because:

- If RPM ALUs were power-inefficient, AMD wouldn't have them inside Raven Ridge which goes down to 12W cTDP.
- If this was actually a Polaris chip (which it could), then the featureset wouldn't be DX12_1.
- RPM could make up for the scarce compute resources in some scenarios, so looking at this GPU's resources with 64 ROPs and 205GB/s it would make even less sense to disable it.
- Intel's own gen9 supports 2*FP16 throughput, so it's not like disabling RPM would be a "custom order" from Intel to keep feature parity with their iGPU or something similar.
 
Last edited by a moderator:
This is the Tom's article investigating Kaby Lake-G and looking into features such as FP16
http://www.tomshardware.com/news/kaby-lake-g-vega-polaris-graphics,36844.html

Their results comparing FP16 and FP32 with several GPUs;

aHR0cDovL21lZGlhLmJlc3RvZm1pY3JvLmNvbS9ML1kvNzYyODM4L29yaWdpbmFsL0ltYWdlMS5qcGc=


Of course that may just mean currently disabled, or that it does not exist *shrug*.
And fair to say for now
Tom's Hardware said:
Intel points to the HBM controller and additional ROPs as evidence that Kaby Lake-G wields a Vega-class Radeon graphics chip, also citing "enhanced compute units," implying the NCU. Given the amount of customization that went into this design, it's possible that Kaby Lake-G's Radeon graphics are a hybrid with enough Vega functionality to justify Vega marketing.
Intel says similar to Vega solution, which is the stand-out; but where it leaves it with optimised Vega branded games is debatable.
 
Last edited:
Does this fly in the face of earlier articles about GPP considering Dell,HP and maybe a fewother OEMs are not meant to have signed up to GPP according to same source articles reporting GPP in the 1st place stating they are fighting GPP?
Point being not even the big OEMs seem to have a broad range/commitment flooding the market or more importantly even reviewed apart from Intel pushing it the most in NUC form.
Could be possible GPP involved but also other factors may contribute and wouldn't it be more likely the smaller OEMs would brand it in a different section if under GPP pressure rather than actually stall making one if they feel it is worth their time.

What is the cost to the OEM for these components and where would OEM's position it or even availability to OEMs, especially when Intel has its own NUC product (appreciate not mobile but makes that segment very competitive) launched and position relative to performance tier of Intel+Nvidia dGPU gaming laptop.
For reference as not sure, what would be an equivalent OEM Intel+Nvidia dGPU be to Kaby Lake G laptop?
I must say the HP Spectre Kaby Lake-G seems rather expensive relative to their other models out there.
Edit:
Better clarification.
 
Last edited:
Point being not even the big OEMs seem to have a broad range/commitment flooding the market or more importantly even reviewed apart from Intel pushing it the most.
Could it be possible but other factors could also contribute *shrug*, but time will tell as any of the smaller OEMs would create the model within a new brand (such as Arez if not possible with ROG), although I think the latest AMD motherboard is in ROG but then it does raise questions that this is not technically an APU and so under pressure from the GPP.
Being an ideally embedded design Intel had a head start with the NUCs. The design is likely still being tested with OEMs providing input and looking towards future iterations. OEMs could have entirely different form factors to differentiate their products. Saw an article earlier about AMD graphics changes in ChromeOS. Higher performance tier designs are likely coming as Google, Apple, Amazon, Roku, telecoms, etc could jump on the bandwagon with the smaller devices. Days of the ATX standard may be limited.
 
Being an ideally embedded design Intel had a head start with the NUCs. The design is likely still being tested with OEMs providing input and looking towards future iterations. OEMs could have entirely different form factors to differentiate their products. Saw an article earlier about AMD graphics changes in ChromeOS. Higher performance tier designs are likely coming as Google, Apple, Amazon, Roku, telecoms, etc could jump on the bandwagon with the smaller devices. Days of the ATX standard may be limited.
Yeah mostly agree (although I think ATX still has legs), which would tie into why it is still too early to expect a broad range of OEM (smaller companies such as Asus,etc) Kaby Lake-G products and especially when it comes to laptops.
How many of the Dell/HP Radeon Kaby Lake-G laptops have been reviewed-launched and at what price; comes back to what you say IMO.
That said cost I feel has some aspect it to this, which does not stand out when one looks at the Intel NUC as it seems it has a great price, but look at the cost of the HP Spectre Radeon Kaby Lake-G relative to the other Spectre models and it is not quite so enticing and would have an impact on how/when the smaller gaming OEM's launch a product IMO.

I guess we need a laptop review to tell where such a product sits relative to a model with Intel CPU+Nvidia dGPU, which may also be a consideration for OEM's focusing on gaming (price-performance-value), along with component availability.
 
Last edited:
Back
Top