Reply to thread

Message: <blockquote data-quote="ibnanv" data-source="post: 22455264" data-attributes="member: 218596">                     <img src="https://images.anandtech.com/doci/12003/multichip_678x452.png" alt="" class="fr-fic fr-dii fr-draggable " style="" />                                                                                                          Today we have an announcement out of left field. Intel  has formally revealed it has been working on a new series of processors  that combine its high-performance x86 cores with AMD Radeon Graphics  into the same processor package using Intel’s own EMIB multi-die  technology. If that wasn’t enough, Intel also announced that it is  bundling the design with the latest high-bandwidth memory, HBM2.  Intel announced its EMIB technology over the last twelve months, with  the core theme being the ability to put multiple and different silicon  dies onto the same package at a much higher bandwidth than a standard  multi-chip package but at a much lower cost than using a silicon  interposer. At Intel’s Manufacturing Day earlier this year, they even  produced a slide (above) showcasing what might be possible: a processor  package with the x86 cores made on one technology, the graphics made in  another, perhaps different IO and memory or wireless technologies too.  With EMIB, processor design can become a large game of Lego.  <a href="https://images.anandtech.com/doci/12003/16.jpg" target="_blank"><img src="https://images.anandtech.com/doci/12003/16_575px.jpg" alt="" class="fr-fic fr-dii fr-draggable " style="" /></a>  EMIB came to market with the latest Intel Altera FPGAs. By embedding  the EMIB required silicon design into the main FPGA and each of the  chipsets, the goal was to add multiple memory blocks as well as data  transfer blocks in a mix and match scenario, allowing large customers to  have the design tailored to what they require. The benefits of EMIB  were clear, without the drawbacks of standard MCP design or the cost of  interposers: it would also allow a design to go beyond the monolithic  reticle limit of standard lithography processes. It was always expected  that EMIB would have to find its way into the general processor market,  as we start to see high-end server offerings approaching 900 mm2 over  multiple silicon dies in a single package.  <a href="https://images.anandtech.com/doci/12003/hc29.22.523-hetro-mod-platform-shumanrayev-intel-final-page-007.jpg" target="_blank"><img src="https://images.anandtech.com/doci/12003/hc29.22.523-hetro-mod-platform-shumanrayev-intel-final-page-007_575px.jpg" alt="" class="fr-fic fr-dii fr-draggable " style="" /></a>  Since the EMIB announcements, Intel’s Manufacturing Day, and Hot  Chips, word has been circulating about how Intel is going to approach  this from a consumer stand point. As part of the requirements of Intel’s  own integrated graphics solutions, <a href="https://www.anandtech.com/show/4122/intel-settles-with-nvidia-more-money-fewer-problems-no-x86" target="_blank">a 2011 cross-licensing deal with NVIDIA was in place</a> – this deal was set to expire from April 1st 2017, and no mention of extending that deal was ever made public. A <a href="http://fudzilla.com/news/graphics/42322-future-intel-cpus-might-come-with-amd-graphics-chip" target="_blank">couple of rumors floated</a>  around that Intel were set to make a deal with AMD instead, as despite  their x86 rivalry they were a preferred partner in these matters.  Numerous <a href="http://www.fudzilla.com/news/graphics/42815-intel-s-nvidia-gpu-license-expires-in-march" target="_blank">outlets with connections</a>  in both AMD and Intel had difficulty prizing any information out.  Historically Intel refuses to comment on such matters in advance. Other  potential leaks include published benchmarks over at SiSoft, although  nothing has been made concrete until today.  Intel’s official statements on the announcement offer a few details worth diving into.  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.   Intel interestingly uses a singular word for ‘product’, although this  does not indicate if it is a family or literally a single SKU in the  works. On Intel’s Core-H series processors, these are currently Kaby  Lake based running at 45W, with Intel’s integrated GT2 graphics. It  would be interesting to see if the graphics of the Core-H are then  stripped out as a new silicon design, or if they are re-spinning the  full Core-H silicon as a result and just displaying the integrated  cores, or are able to run both graphics segments independently (it is  likely a new spin of silicon, if I were a betting man). The use of HBM2  is not unsuprirising – Intel has successfully integrated HBM2 into its  Altera EMIB-based products so we would suspect that this is not going to  be overly difficult.  The next bit is the interesting one: ‘custom-to-Intel … discrete  graphics chip’ from AMD RTG. This means that none of AMD’s current  product stack has silicon dedicated to EMIB, but AMD is going to  leverage its semi-custom design to provide graphics chiplets for Intel  to add to its silicon.  ‘In close collaboration, we designed a new semi-custom graphics  chip, which means this is also a great example of how we can compete and  work together, ultimately delivering innovation that is good for  consumers… Similarly, the power sharing framework is a new connection  tailor-made by Intel among the processor, discrete graphics chip and  dedicated graphics memory. We’ve added unique software drivers and  interfaces to this semi-custom discrete GPU that coordinate information  among all three elements of the platform.’   One of the questions about running multiple chips in a single package  is how to manage all the bandwidth and power. AMD has recently solved  that issue in its server processors and inside its APUs by using their  Infinity Fabric, which if I were to guess would not be under the purview  of this collaboration. It states that with collaboration that the chip  shares a power framework, which will be an interesting deep dive when we  get information as to whether Intel offering separate power rails for  the CPU and GPU segments, using an integrated voltage regulator (like  Broadwell did), or doing something similar to AMD by using a unified  power rail sharing mechanism with digital LDOs as was <a href="https://www.anandtech.com/show/11964/ryzen-mobile-is-launched-amd-apus-for-laptops-with-vega-and-updated-zen/4" target="_blank">announced with Ryzen Mobile only a couple of weeks ago</a>.  ‘Look for more to come in the first quarter of 2018, including systems from major OEMs based on this exciting new technology.’  It looks like Intel is ready to make some announcements over the next  few months on this project, and CES is just around the corner in  January.  Though taking a step back, we have to consider what this means and  what market Intel is aiming for. AMD recently launched (with products  coming soon) their Ryzen Mobile platform, designed with quad-core Zen  and up to 10 CUs of Vega graphics. The announcements from Intel and AMD  do not state what graphics core they are using (they could be one  generation behind for competitive reasons?) however it does state that  they are using Core-H series processors, which are typically in the 45W  range. AMD currently hasn’t announced anything in that segment, and  deciding to focus Ryzen Mobile at the thin and ultralight notebook  categories first. If AMD does bring Ryzen Mobile up, then this new  product will be in direct competition.  <a href="https://images.anandtech.com/doci/12003/intel-8th-gen-cpu-discrete-graphics-2.jpg" target="_blank"><img src="https://images.anandtech.com/doci/12003/intel-8th-gen-cpu-discrete-graphics-2_575px.jpg" alt="" class="fr-fic fr-dii fr-draggable " style="" /></a>  Looking at the image provided by Intel on the new product arrangement  actually adds a new question or two to the bucket list. Here we have an  Intel chip on the right, the AMD custom graphics in the middle, and the  HBM2 chip next to it. The Intel chip is a long way away from the AMD  chip, which would suggest that these two are not connected via  EMIB if the mockup was accurate. The close proximity of the big chip in  the middle to what looks like a HBM2 stack does suggest that it is  connected via EMIB, as given by how close the chips in the Altera  products are:  <a href="https://images.anandtech.com/doci/12003/emib_comp.jpg" target="_blank"><img src="https://images.anandtech.com/doci/12003/emib_comp_575px.jpg" alt="" class="fr-fic fr-dii fr-draggable " style="" /></a>  EMIB is being used, but it does not look like it is being used for  all the chips together. It’s worth noting that neither Intel nor AMD  offered pre-briefings on this announcement, so there are a lot of  unanswered questions hanging around as a result.  A final thought. Apple uses a lot of Intel's 45W processors for  iMacs; offering AMD graphics (Apple's preferred pro-graphics partner)  into the segment that previously Intel's Crystalwell/eDRAM based  products exist might be the next step on that product cycle evolution.</blockquote>

Verification: Awruddata maasa keeyada?

Top Bottom