Intel is stacking memory on top of the new Xeons

Subject: General Tech | November 22, 2013 - 12:29 PM |
Tagged: Supercomputing Conference, Intel, SoC, Near Memory, knights landing

Intel spilled more beans about the new Near Memory architecture that will be accompanying their new Xeon release.  The memory will be stacked directly onto the CPU giving much quicker access than you would normally see from DDR3 which has to travel over the motherboard.  They have not disclosed expected speeds, which could be up to what we see in current CPU caches only in much larger sizes.  This is not quite a Xeon SoC but in the presentation The Register heard of Intel's plans to incorporate optical fabrics and switches onto the CPUs as well with size being the only limit.  Perhaps they do have a leg to stand on when they claim the return to power of homogeneous computing.

View Full Size

"According to an EE Times report, Intel's Rajeeb Hazra, a VP and general manager of its data centre group, said Intel would customise high-end Xeon processors and Xeon Phi co-processors by closely integrating memory, both by adding memory dies to a processor package and, at a later date, integrating layers of memory dies into the processor along with optical fabrics and switches."

Here is some more Tech News from around the web:

Tech Talk

Source: The Register

November 22, 2013 | 06:05 PM - Posted by HelloFriends (not verified)

How Much RAM are we talking about? 64GB, 128GB or more?

Could we supplement total RAM after buying the CPU, or is it more like enormous CPU Cache?

Ultimately, I assume our fan would be cooling both CPU and system RAM at the same time- any thermal issues?

Also, would this arrangement facilitate RRAM use?

Interesting to see this come to fruition, but I assume this would necessitate a new Motherboard and chipsdet to implement...

November 22, 2013 | 07:50 PM - Posted by Anonymous (not verified)

It sounds stopgap, until Intel can come up with some heterogeneous solutions of its own, as even in the HPC and server market, heterogeneous CPU/GPUs, (AMD/ARM future HSA) based, and others' heterogeneous solutions have better resource utilization, and lower enegry usage! There will always be a market for some of Intel's x86 SKUs, on the very high end, and as fabric contollers/switches but everywhere else, in servers HSA is making inroads, and GPGPU, AMD and NVIDIA, will not go away! And do not forget that IBM is also heavely invested in Optical research/products, and Intel is not the only company with plans for stacked memory CPUs! The Idea behind HSA and being able to utilize CPUs and GPUs of any different architecture for an all resources combined computation, while still retaining their graphics or genegral purpose compute effiencies, should they still be needed for specific workloads, is too good to pass up! Intel, with HSA, as it has been with moblie, appears to be in a state of denial, and Intel's presence in the mobile devices market proves that it is not good to not innovate! Intel, over the years, has spent most of its R&D money on maintaining is Chip Fab process node lead, and very little on designing a low power CPU architecture, until recently, and what are they spending the new resarch money on, still x86 and not a new architecture. In that time waisted, the market has shifted to the companys that utilize ARM IP for mobile, and servers, and Nvidia/AMD GPGPU for HPC!

November 23, 2013 | 08:09 AM - Posted by Klimax (not verified)

This read like marketing gibberish and unstructured to boot.

Also I think I spotted several unsupportable assertion in there...

November 23, 2013 | 01:29 PM - Posted by Anonymous (not verified)

Were it only was mentioning/promoting one company you, remotely, could say it reads like marketing, but the lack of flowery diction, and excess use of superlatives in the post, and the mentioning of Intel's competition: AMD, Nvidia, The companies that license and build CPUs based on the ARM instruction set, and Arm Holdings refrence designs, and IBM (power pc, optical interconnects, mainframes, servers, OS), pretty much proves that it is not marketing quackspeak and Intel is in fact not the sole innovator of CPUs on the planet! Intel is not the sole creator of FINFET, and Intel's process lead in chip fabrication, is more a function of Intel haveing more money to spend on the chip fabrication equipment and IP, that is available to any company with the money to, pay the Non-Intel companies, that supply the world with the chip fabrication machinery and knowhow! Sure it hurts when an Intel fanboy's Godhead is put in a true light, but read up on those other companies, and find something different to worship, other than a supplier of CPU parts(Intel), or any other CPU/device/software parts supplier, for that matter! HSA is not the sole property of one company, but a design Hardware/software framework that enables heterogeneous computing devices CPUs and GPUs of whatever make, model, or instruction set, be used for general purpose compute, while still retaining their native uses, if that is the best for the job at hand. Intel is more than welcome to join the HSA foundation.

November 22, 2013 | 10:23 PM - Posted by wow&wow (not verified)

Why noy just fabricating memory chips for Samsung, Hynix, Micron, ...