Fractal Design Define 7 Compact review
Lian Li Galahad 360 liquid cooler review
Guru3D Rig of the Month - July 2020
Patriot Viper Steel DDR4 3600 MHz (64GB) review
Red Dead Redemption 2: PC graphics benchmark review (revisited)
Kioxia Exceria 960 GB SATA3 SSD review
HP Portable P500 1TB Portable USB3 SSD review
EK-Quantum Magnitude WaterBlock Review
Death Stranding: PC graphics performance benchmark review
Deepcool DQ850-M-V2L 850 W PSU review
Photo Shows The New Intel MCM based CPU with AMD-GPU
Intel this week announced its mobile MCM processor based on a 14 nm Coffee Lake-H CPU cores and a Vega based graphics solution. This is called a multi-chip module, and the first photo just surfaced. And it's totally weird to observe alright.
Intel plans this laptop-computer chip combining an Intel processor and an AMD graphics unit. The new product, which will be part of the 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, but based on several dies.
A photo surfaced online showing the massively huge MCM design, as shared by tech website Bits' n Chips on Twitter. On the left-hand side of the photo, you can see the Intel processor and to the left, the larger AMD GPU with the HBM2 stack.
A 14nm Vega architecture was the choice for the GPU, Intel mentions the usage of HBM2. Earlier this year Intel already announced their Embedded Multi-Die Interconnect Bridge (EMIB) making these things (a second die) possible, for modular builds. The Intel and the AMD part communicate over what basically is a PCIe Gen 3.0 bridge wired in-between each other. The first products based on MCMs will likely be embedded products and laptops, they should launch by Q1 2018.
aKiss
Member
Posts: 33
Joined: 2017-01-26
Member
Posts: 33
Joined: 2017-01-26
#5491226 Posted on: 11/10/2017 10:35 PM
The hell, men, am i Nostradamus or something?
https://forums.guru3d.com/threads/why-doesnt-amd-do-this.415241/
The hell, men, am i Nostradamus or something?
https://forums.guru3d.com/threads/why-doesnt-amd-do-this.415241/
KissSh0t
Senior Member
Posts: 7866
Joined: 2011-10-22
Senior Member
Posts: 7866
Joined: 2011-10-22
#5491234 Posted on: 11/10/2017 11:02 PM
I guess something like this would be well received by Apple.
I guess something like this would be well received by Apple.
meran
Senior Member
Posts: 281
Joined: 2008-07-17
Senior Member
Posts: 281
Joined: 2008-07-17
#5491301 Posted on: 11/11/2017 08:42 AM
This is Vega 11 the mainstream Vega that is rumored with 2048 shadder cores and single stack of HBM2 they are not sharing anything they are selling gpus to Intel that are still made in globalfoundaries fabs and the memory is from Samsung or hynix...
That is like saying all GPU's must use DDR5 etc. We have seen for a long time GPU's can use different types of memory, some would come with DDR3 back in the day with the same lower cost GPU featuring DDR2. AMD could have easily released Vega with GDDR5 if they wanted. GPU architecture doesn't dictate what memory it has to use. The part number in the OpenCL information for this new CPU/GPU combination is gfx804, a polaris based chip. I really can't see AMD sharing Vega with Intel just yet. Adored makes this clear in his video
This is Vega 11 the mainstream Vega that is rumored with 2048 shadder cores and single stack of HBM2 they are not sharing anything they are selling gpus to Intel that are still made in globalfoundaries fabs and the memory is from Samsung or hynix...
Aura89
Senior Member
Posts: 7873
Joined: 2008-07-31
Senior Member
Posts: 7873
Joined: 2008-07-31
#5491304 Posted on: 11/11/2017 08:52 AM
Lol I see my guess was on spot
, because the die size looked ~50% of Vega 56, nice.
now look how much it sa\ves comparing for 1060:
its also the thickness that is saved for not needing to pace the GPU PCB on top of the MB PCB.
It's really only saved if they don't have to use the same amount of fans/heatsinks, of which it being in the same package means that:
The heat all builds up in one area, but, the heat still builds up the same amount of heat, requiring the same cooling requirements
And
Both the CPU and GPU will heat eachother, not entirely certain that part is so great.
Lol I see my guess was on spot
, because the die size looked ~50% of Vega 56, nice.now look how much it sa\ves comparing for 1060:
its also the thickness that is saved for not needing to pace the GPU PCB on top of the MB PCB.
It's really only saved if they don't have to use the same amount of fans/heatsinks, of which it being in the same package means that:
The heat all builds up in one area, but, the heat still builds up the same amount of heat, requiring the same cooling requirements
And
Both the CPU and GPU will heat eachother, not entirely certain that part is so great.
Click here to post a comment for this news story on the message forum.
Senior Member
Posts: 1941
Joined: 2013-03-10
If they could have easily done it, they would have done it in some form, instead of releasing Vega so terribly late. No doubt the memory controlled in the GPU has a thing or two to say about it. Of course nothing prevents them from partially remaking Polaris to support HBM2, but would they do it when they already have a design using HBM2?