Intel announces HPC Knights Landing Xeon Phi accelerator
The initial systems with Knights Landing Xeon Phi accelerator have been shipped to customers, says Intel. Knights Landing is the a manycore x86 architecture that Intel has developed over the years.
The previous iteration of the hardware, Knights Corner, used a modified Intel Pentium core (P54C). The Knights Landing-based Xeon Phi is a liquid-cooled system with six DIMM slots, dual gigabit Ethernet ports , 16GB of MCDRAM, AVX-512, and 72 CPU cores.
With the introduction of the new Xeon Phi accelerators it is possible to boot from the cpu and the system memory is integrated on the CPU for additional performance. It is possible to create a complete cluster with only Xeon Phi withouth any other Xeon systems.
The Knights Landing Xeon Phi card has been tested and validated ever since late last year by various partners , including Cray. The new Xeon Phi card is produced at 14nm and contains up to 72 cores, it features 16GB of multi-channel dram which can do a transfer rate of 400GB/s. Knights Landing provides support for 36 PCI-E 3.0 lanes and uses what is referred to as the Omni-Path architecture .
Omni-Path is part of Intel's Scalable System Framework for High Performance Computing. Intel says that the accelerator can be used for applications in the field of artificial intelligence, an emphasis that is inspired by Google who started their own artificial intelligence project.
Rumor: Intel and AMD to present Kaby lake and Zen in January 2017 - 06/08/2016 11:46 AM
At least that info comes to us through Digitimes. Intel would plan an introduction at CES with that processir series called Kaby. During that same time-frame the same time AMD would reveal processors...
Intel announces leadership and organizational changes - 07/03/2015 07:30 AM
Intel Corporation's CEO, Brian Krzanich, announced a series of leadership and organizational changes in a message to employees. Intel President Renée James informed the board that she will be steppi...
Intel Announces Availability of 5th Gen. Core vPro Processor Family - 01/30/2015 10:09 AM
Intel Corporation today announced the availability of its 5th generation Intel Core vPro processor family that provides cutting-edge features to enable a new and rapidly shifting workplace. To meet th...
Intel Announces Broadwell the 5th gen Intel Core Processor family - 01/05/2015 10:37 PM
Built on Intel’s cutting edge 14nm manufacturing process, the 5th Gen Intel Core processors deliver premium performance, stunning visuals and enables improved battery life to take computing ...
Intel Announces Core M entry level mobile processors - 09/05/2014 08:09 PM
Intel launched the new Intel Core M processor, which will power new 2 in 1 devices from a variety of manufacturers including: Acer, ASUS, Dell, HP, Lenovo and Toshiba. Delivering the optimal blend of ...
Senior Member
Posts: 11721
Joined: 2012-07-20
Did I really read P54c as last iteration? I had that so many years ago. So, project was canned for pretty long time.
72Cores (Or maybe threads) is not that impressive in 2016. And in 2017 it will not be impressive at all.
And those slides showing actual workloads... Compared to what exactly?
AMD is pretty bad in their presentations, and approximating real world performance from that can end up anywhere within 30% above or below.
But this presentation. It tops that since it is compared to unspecified product.
- - - -
I guess it is going to be 20% faster than grandma's wheelchair.
Senior Member
Posts: 13287
Joined: 2004-05-16
Did I really read P54c as last iteration? I had that so many years ago. So, project was canned for pretty long time.
72Cores (Or maybe threads) is not that impressive in 2016. And in 2017 it will not be impressive at all.
And those slides showing actual workloads... Compared to what exactly?
AMD is pretty bad in their presentations, and approximating real world performance from that can end up anywhere within 30% above or below.
But this presentation. It tops that since it is compared to unspecified product.
- - - -
I guess it is going to be 20% faster than grandma's wheelchair.
It's four threads per core and IIRC it's 3Tflops of double precision and 6 single.
I guess the advantage is the check boxes on the second slide.
Senior Member
Posts: 1779
Joined: 2005-08-12
Most Widely Adopted
for Machine Learning 2
2 Source: Intel Estimates
You're kidding, right?
Yet another marketing bs. Besides, you can find more rendering apps supporting GPUs than Intel Xeon Phi.
Senior Member
Posts: 2467
Joined: 2014-01-21
Did I read that correctly $6,000 Grand?! As in six thousand us dollars? Does that come with an Intel office (female compainion?) haha.
Senior Member
Posts: 2684
Joined: 2011-05-17
Be on the lookout for Miles Bennett Dyson.