Problems? See this thread at archive.org.
First photos of the new Intel-AMD multi-chip module
The first images of the new Intel-AMD multi-chip module (MCM) in the wild have surfaced and going by the looks of it, appears to be exactly what ultrabooks and mini-PCs have been longing for all this while.
https://www.notebookcheck.net/First...l-AMD-multi-chip-module-surface.263224.0.html
Source(s)
Bits and Chips on Twitter
Guru3D
ChipHell
Rivals with benefits — Intel and AMD collaborate to create a new entrant to the 8th generation CPU family
In an interesting development, CPU arch rivals Intel and AMD have collaborated together to create a unique new entrant to Intel's 8th generation CPU family — a combination of an Intel H-series chip coupled to a...wait for it...a custom AMD Radeon GPU with HBM2 memory!
https://www.notebookcheck.net/Rival...o-the-8th-generation-CPU-family.262554.0.html
"Even the worst of rivalries can end in friendships. Intel's latest announcement just proves this point. After a number of rumors and subsequent denials, Intel's Vice President of Client Computing Group and General Manager of the Mobile Computing Platform, Christopher Walker, finally announced that Intel and AMD have collaborated together to create a new die design that combines an Intel H-series CPU and a custom-to-Intel AMD Radeon GPU along with High Bandwidth Memory 2 in a single package. This new design is facilitated by Intel's Embedded Multi-Die Interconnect Bridge (EMIB) technology that allows heterogeneous silicon to exchange data in extremely close proximity.
EMIB has been a traditionally server class FPGA technology and ever since the benefits of it were clear, there was always speculation as to how it can make its way to the consumer side of things. Intel's integrated graphics solutions, though good, are clearly not enough for handling escalating modern workloads combined with the necessity for a thin form factor. AMD's new Vega architecture has been successfully implemented in the new Ryzen Mobile CPUs so this unique combination could be a win-win situation for both the companies and customers."
"Walker said the following about the new die —
"The new product, which will be part of our 8th Gen Intel Core family, brings together our high-performing Intel Core H-series processor, second generation High Bandwidth Memory (HBM2) and a custom-to-Intel third-party discrete graphics chip from AMD’s Radeon Technologies Group* – all in a single processor package.
... Now, we’re opening the door for thinner, lighter devices across notebooks, 2 in 1s and mini desktops, while delivering incredible performance and graphics for enthusiasts."
Vice President and general manger for AMD Radeon Technologies Group, Scott Herkelman said,
"Our collaboration with Intel expands the installed base for AMD Radeon GPUs and brings to market a differentiated solution for high-performance graphics. Together we are offering gamers and content creators the opportunity to have a thinner-and-lighter PC capable of delivering discrete performance-tier graphics experiences in AAA games and content creation applications."
Supposed Intel-AMD Core i7-8705G benchmarks rival the GTX 1050 in performance
The Intel-AMD Kaby Lake-R and Polaris/Vega combo could potentially outperform the current console generation and the GeForce MX150 based on unconfirmed reports.
https://www.notebookcheck.net/Suppo...val-the-GTX-1050-in-performance.262790.0.html
"With the initial launch supposedly just months away, more concrete leaks are now hinting at what we can expect to come. The latest is a set of unidentified 3DMark 11 scores that are purportedly the upcoming Core i7-8705G and i7-8809G with the unnamed AMD Radeon Polaris/Vega graphics. The 3DMark 11 Graphics Performance scores sit at 3777 points and 3960 points for the i7-8705G and i7-7709G, respectively, to be a step below the average GTX 950M at 4357 points but well above the average GeForce 940MX at 2560 points. These scores, of course, are lower than expected given that the integrated Iris Plus Graphics 640 is already comparable to the GTX 850M and 950M when gaming. Nonetheless, the scores show we can expect just a minor GPU performance dip of only about 5 percent between the higher-end i7-7709G and lower-end i7-8705G.
The next set of unconfirmed benchmarks is potentially more interesting as it involves a repeatable scene in a shipped game. As leaked by Reddit and reported by Hot Hardware, the i7-8705G returns an average of 33.5 FPS in the built-in Ashes of the Singularity benchmark at 1080p High settings. When compared to our own benchmark tests, the result is nearly identical to the GTX 1050 (29.5 FPS) and, by extension, the GTX 965M. The more recent MX150 runs significantly worse at 17.1 FPS in the same benchmark test. Ashes of the Singularity is quite CPU intensive as well since it is a RTS title that benefits directly from higher processor core counts. Nonetheless, a wider variety of benchmarks will be necessary before assuming a bigger picture."
-
Thank you for the non-video reporting!
What this will surely bring is the death of 1366x720 displays. Finally.
This may be what my next 'digital' notebook will be based on.
The author is dreaming and/or out of touch with where Intel wants to take this. This isn't the end of their own igpu's - not by a long shot.
He miscalled this this by a mile.
From the link in the first post;
Last edited by a moderator: Nov 13, 2017 -
lol... now you're dreaming too.
That 'parasitic' igpu is what made mobile computers, mobile for me and many, many others.
A dGPU is very much unneeded - along with it's battery taxing and throttle inducing antics - in the majority of mobile use cases.
And yeah; gaming isn't even close to being 'some', let alone 'most'.
-
Ha! Take the best CPU you can (yeah; go ahead, pick the TR or EPYC ones...) and put it in a mobile system.
UNUSABLE.
Even Intel's silly little igpu would make it a real computer.
But you'll tell me you'd pick something better than an igpu anyway. Good luck powering it for more than a few minutes.
There isn't one right 'path', nor one right platform.
Give credit where credit is due.
-
At least you got that part right
There were and would have been many other solutions for mobile GPU, more innovation in different directions - better one's would prevail over time.
Intel stifled innovation in mobile GPU by taking it on themselves and blocking integration with higher performance GPU's with the CPU, and it's clear they now know that they made a big mistake, and as hard as it is to admit it and turn that big lumbering mistake around, Intel is doing it.
Let's stop talking about it, and see what happens, ok?
Last edited: Nov 13, 2017 -
But we're here to talk about stuff.
That silly little gpu that Intel has been making a fool of themselves with has probably sold more units and made more money for Intel than what other pure gpu makers have made. You can laugh and so is Intel, all the way to the bank.
I sincerely hope Intel keeps stifling innovation like you seem to think. We're all the better off for it.
It truly is amazing how people can look at the same thing and see completely opposite things, huh?
Like I already said; the better one has already prevailed. Best CPU with just enough graphics hp to give humans the video output they need. They predicted how slow it would be for the start, but they're going to sling shot past anything in the solar system as they pick up exponentially more speed.
How are they doing this? Because they have a plan. Their plans don't span timelines like consumers may like or appreciate, but they do follow through where and when they can.
They plan long and they plan cohesively. Sure; not all work out, but when they do, it isn't a fluke. It is their hard work and dedication paying off.
Dedication to their investors and their customers - even when those seem to be at opposite ends sometimes.
I can easily see how the power of a dGPU is needed for certain workflows/playloads - but I can also see it being implemented much better than what anyone up to now has offered us.
I trust Intel to be going into this venture with Raja with not only open eyes; but also a clear goal of what needs to be accomplished by him and the team they let him build.
Just like FPU's were once optional on a platform, gpu 'compute' power is too right now for many, many workloads. Where Intel is headed is to change that. Disruptively.
Combined with all the other parts of the platform Intel based systems offer for consumers; this will be an order of magnitude different computing world for mere humans - once the programming and the O/S subsystems have been put in place.
You can look at Intel's slow, methodical journey with igpu's as a mistake. But you'll only be right if Intel stops producing and innovating with gpu's, today.
Given the environment right now over at Intel; that isn't happening soon.
-
You posted an hour of video when I'm heading out the door soon...
Along with the video, can you at least post your reasons for the 'interesting' part(s) in a few short sentences?
btw, even when I get back to my computer - I still won't have time to see 1Hr of fluff for 10 seconds worth of 'ahhh!).
hmscott likes this. -
hmscott, best part of the post above: " Rumor: Hades Canyon NUC with AMD Graphics Spotted".
Didn't even need to see the video!
-
Sure, since you asked nicely.
Links first,
Upcoming NUCs: https://www.pcper.com/image/view/86083?return=node/68457
Leaked performance: https://www.pcper.com/category/tags/kaby-lake-g
Device Name: https://cdn.wccftech.com/wp-content...-Kaby-Lake-G-Series-Integrated-Radeon-GPU.jpg
Early leak: https://tech4gamers.com/wp-content/uploads/2017/04/intel-kaby-lake-g.jpg
GPU leak: http://ranker.sisoftware.net/show_r...d3eed6e5d6e5d7f183be8ea8cda895a583f0cdf5&l=en
More leaks: https://hothardware.com/news/intel-...ntegrated-amd-radeon-graphics-benchmarks-leak
Based on that, let's apply logic to accurately speculate.
Performance:
First, we know the package size is 58.5mm x 31mm. Based on that we can calculate the die size of the GPU portion to be ~200mm2. That is rather large, and quite close to Polaris 10's 232mm2 size.
Applying logic, a score of a mere 4000 in 3DMark11 P would be an absolute waste of effort, and cost, and power. Because a small modification to an existing solution would be better, since Iris Pro P580 can do 3500 at 45W TDP. This would be the 2nd only implementation of a new EMIB packaging technology, the first ever cost-effective HBM2 use.
The NUC roadmap also calls the higher-end, 100W version "Hades Canyon VR". Based on TDP, and the significant effort to talk to AMD, and integrate a not too cheap, HBM2/EMIB solution, you'd need 1050 Ti+ performance.
Number of SPs:
The sisoftware results show a 694E:C0 model with 24CUs and 1536SP. According to the Hothardware leak, it shows what I'd call some inconsistency because the 694E:C0 is the lower end model, and 694C:C0 is the high end one. It's possible that we'll see multiple models using the same SP GPU, just differing in clocks.
HBM2 clocks:
The same leaks show 700MHz/800MHz for the HBM2 clock. Since it uses DDR, the final clock turns out to be 1.4/1.6GHz. Because they are likely having parametric(performance related) yield problems, no HBM2 implementations exist that runs at the full 2GHz. Highest one is in Vega at 1.89GHz. I think 1.4/1.6GHz is quite realistic.
Polaris or Vega?
Geekbench result shows Device Name of gfx804. If you search for it that seems to be bunch of AMD GPUs that are in the Pro and the Embedded section. They are all Polaris. Right now the gfx804 series are very low end parts with 512SPs, but gfx804 with 694C:C0 doesn't exist. A yet nonexistent part destined for Intel.hmscott likes this.
Intel AMD Multi-Chip Module (MCM) Laptops and Desktops
Discussion in 'Hardware Components and Aftermarket Upgrades' started by hmscott, Nov 11, 2017.
