Archive

Posts Tagged ‘intel’

Intel NUC Roadmap 2018 – 2019 – Gemini Lake, Coffee Lake, and Kaby Lake H

September 20th, 2017 13 comments

Intel’s new generation of Gemini Lake and Coffee Lake processors is expected to launch at the end of this year, beginning of next, and this morning I received Intel’s NUC roadmap that gives a good idea of what’s coming in 2018 and 2019.

Click to Enlarge

Intel plans a whole new generation of NUCs from “Hades Canyon VR” NUCxi7HVK model with a 100W Intel Core i7-xxxxK Kaby Lake-H processor to “June Canyon Celeron” NUC7CJYH model with a 10W Intel Celeron J4005 dual core “Gemini Lake” processor. All in all a total of 7 new NUCs should be launching in 2018.

Let’s have a closer look at the cheaper and lower power Gemini Lake models, starting with “June Canyon Celeron” NUC7CJYH specifications:

  • SoC – Intel Celeron J4005 dual core GLK processor (10W TDP)
  • System Memory – 2x DDR4-2400 slots for up to 8GB RAM
  • Storage – M.2 SSD slot, SDXC slot
  • Display – 2x HDMI 2.0a ports supporting up to two independent displays
  • Audio – front stereo headset, rear stereo out / TOSLINK
  • Connectivity – Gigabit Ethernet (TBC), 802.11ac 1×1 WiFi and Bluetooth 5

Intel “June Canyon Pentium” NUC7PJYH model has exactly the same specifications, except for the Intel Pentium J5005 quad core processor with 10W TDP. Both models are expected to launch in Q1 2018.

An interesting aspect of the more powerful NUCs based on Coffee Lake/Kaby Lake H processors is that they all come with one or two ThunderBolt 3 interfaces, and support Optane memory.

Intel Compute Card Dock DK132EPJ Specifications and User Manual Published

September 14th, 2017 4 comments

Intel unveiled the Compute Card at the very beginning of the this year, without that many details, except it would included a 7th Gen Intel Core, memory, storage and wireless connectivity, and connect compliant dock with a new standard connector featuring USB-C and extra I/Os. Later this year, we learned more details about some Apollo Lake and Kaby Lake Compute Cards including specifications and block diagram. However those cards won’t be of any use without docks, and while NexDock promised a laptop dock for the cards, I have not seen any other announcements, but we now have some info about Intel’s own Compute Card dock that looks like a mini PC as the company released technical specifications and user manuals for DK132EPJ dock, and three Compute Card SKUs.

 

Click to Enlarge

Intel Compute Card Dock DK132EPJ specifications:

  • CPU, Memory, Storage, Wireless – Via slot supporting certified Intel Compute Cards
  • Video Output – HDMI and mini DisplayPort
  • USB – 3x USB 3.0 ports
  • Connectivity – Gigabit Ethernet (via Intel I211-AT); Built-in compute card: 802.11ac WiFi and Bluetooth 4.2
  • Misc – Lock indicator; eject button+indicator; power button; security lock
  • Power Supply – 19V via power barrel jack
  • Dimensions – 151.76 mm x 145 mm x 20.5 mm

Click to Enlarge

The enclosure also supports 75 x 75 and 100 x 100 VESA mount so it can be mounted on the back of compatible monitors or televisions. The dock comes with a 19 power adapter with plug adapter for various countries and a 2-meter power cord. The operating system is not pre-installed in the Compute Card, but Windows 10 Home, Windows 10 Pro, Windows 10 Enterprise, Windows 10 Education, and Windows 10 IoT Enterprise are supported, and some Linux operating systems may be supported. The cards requires software support at least for authentication and the eject function.

Compute Card Dock Block Diagram – Click to Enlarge

The compute cards & dock should be available for purchase now, but they do not seem broadly available online, as I could just find the dock listed for $111.19 on Provantage (with the wrong product photo), and CD1C64GK Compute Card with Celeron N3350/4GB/64GB configuration for 641 AED (~$175 US) on “Gear Up Me” website with stock expected in 9 days. Alzashop has all four Compute Card SKUs with prices ranging from 143 to 527 Euros depending on model.

Via Ian Morrison in Mini PCs and TV Boxes G+ Community

Intel Pentium Silver N5000 “Gemini Lake” Notebook Shows Up in Benchmarks

September 1st, 2017 4 comments

We all know that Gemini Lake processors will succeed Apollo Lake family, we have a good idea of the supported features. and even know that Intel is expecting 10 to 15% CPU integer performance improvement between the two generations. I’ve now been informed that an Intel Gemini Lake notebook powered by Intel Pentium “Silver” N5000 quad core processor clocked at 1.09 GHz was spotted in SiSoftware Sandra benchmark database.

Click to Enlarge

If we compare those results to the ones for Intel Pentium N4200 Apollo Lake processor with the same 1.1 GHz base frequency, we can indeed see some progress, and the results, albeit early, are quite encouraging, except for the GPU part with similar results.

Intel Pentium N5000 Intel Pentium N4200 Delta
Processor Arithmetic 31.11 GOPS 21.42 GOPS 1.45x
Processor Multi-Media 44.65 MPix/s 41.38 MPix/s 1.08x
.NET Arithmetic 11.5 GOPS 7.31 GOPS 1.57x
Processor Crypto 2.8 GB/s 1.41 GB/s 1.99x

It’s quite possible to turbo frequency is higher in the Gemini Lake processor, but I could not find any details about this yet.

Intel Introduces Movidius Myriad X Vision Processing Unit with Dedicated Neural Compute Engine

August 29th, 2017 No comments

Intel has just announced the third generation of Movidius Video Processing Units (VPU) with Myriad X VPU, which the company claims is the world’s first SoC shipping with a dedicated Neural Compute Engine for accelerating deep learning inferences at the edge, and giving devices the ability to see, understand and react to their environments in real time.

Movidius Myraid X VPU key features:

  • Neural Compute Engine – Dedicated on-chip accelerator for deep neural networks delivering over 1 trillion operations per second of DNN inferencing performance (based on peak floating-point computational throughput).
  • 16x programmable 128-bit VLIW Vector Processors (SHAVE cores) optimized for computer vision workloads.
  • 16x configurable MIPI Lanes – Connect up to 8 HD resolution RGB cameras for up to 700 million pixels per second of image signal processing throughput.
  • 20x vision hardware accelerators to perform tasks such as optical flow and stereo depth.
  • On-chip Memory – 2.5 MB homogeneous memory with up to 450 GB per second of internal bandwidth
  • Interfaces – PCIe Gen 3, USB 3.1
  • Packages
    • MA2085: No memory in-package; interfaces to external memory
    • MA2485: 4 Gbit LPDDR4 memory in-package

The hardware accelerators allows to offload the neural compute engine, for example, the stereo depth accelerator can simultaneously process 6 camera inputs (3 stereo pairs) each running 720p resolution at 60 Hz frame rate. The slide below also indicates Myriad X to have 10x higher DNN performance compared to Myriad 2 VPU found in Movidius Neural Compute Stick.

Click to Enlarge

The VPU ships with an SDK that contains software development frameworks, tools, drivers and libraries to implement artificial intelligence applications, such as a specialized “FLIC framework with a plug-in approach to developing application pipelines including image processing, computer vision, and deep learning”, and a neural network compiler to port neural networks from Caffe, Tensorflow, and others.

Myriad SDK Architecture

More details can be found on Movidius’ MyriadX product page.

Intel Wireless-AC 9560 CRF Module Adds 802.11ac WiFi and Bluetooth 5 to Gemini Lake/Cannon Lake Processors

August 23rd, 2017 3 comments

Last week, we saw that the upcoming Intel Gemini Lake processors integrated a CNVi (Connectivity Integration) block with a WiFi MAC, and Bluetooth MAC & Baseband Modem connected over a CNVio interface to a separate CRF (Companion RF) module handling the RF part. The design change is shown in the block diagram below with the old design on the left using wireless modules connected over PCIe and USB, and the new design on the right used in Cannon Lake/Gemini Lake processors which aims at saving power, cost, and size.

Intel has now uploaded the product brief for their first wireless CRF module: Intel Wireless-AC 9560 with the following key features:

  • Connectivity
    • WiFi
      • Dual band 802.11 a/b/g/n/ac wave 2 2×2 WiFi up to 1.73 Gbps using 160 MHz channels
      • Standards – IEEE 802.11a/b/g/n/ac, 802.11d, 802.11e, 802.11h, 802.11i, 802.11w, 802.11r, 802.11k, 802.11v pending OS support
      • Security
        • WPA and WPA2, 802.1X (EAP-TLS, TTLS, PEAP,EAP-SIM, EAP-AKA, EAP-AKA’) authentication methods
        • PAP, CHAP, TLS, MS-CHAP, MS-CHAPv2 protocols
        • 64-bit and 128-bit WEP, TKIP, 128-bit AES-CCMP encryption
      • Supports roaming between access points
    • Bluetooth 5
    • Antenna diversity and Radio ON/OFF control supported
  • Connector Interface – M.2: CNVio
  • Dimensions
    • 9560NGW model: 30 x 22 x 2.4 mm (M.2 2230 form factor)
    • 9560D2W model: 16 x 12 x 1.57 mm (M.2 1216 form factor)
  • Weight – 9560NGW: 2.8 grams, 9560D2W: 0.7 gram
  • Certifications – FIPS, FISMA, UL, C-UL, CB (IEC 60950-1), and other regulatory certifications depending on country

Intel has been providing M.2 2230 and M.2 1216 wireless modules for a while, so there’s no direct space saving here with CNVi when using those modules, but they should still come with less components than previous design.

We’ll have to wait until systems based on Gemini Lake or Cannon Lake start selling before being able to purchase Wireless-AC 9560 module, or see it bundled in computers or laptops.

Intel Gemini Lake Block Diagram and Yet More Info

August 14th, 2017 6 comments

So yesterday, I wrote about some of the new features of Intel Gemini Lake processors like native HDMI 2.0, 4-wide pipeline, 10-bit VP9, and possible built-in 802.11ac wireless controller. I went to bed, and somehow this morning I woke up with something that looks like Gemini Lake (GLK) block diagram, and a few more details.

So we indeed have HDMI 2.0 output, as well as DP 1.2a and eDP 1.4, and an embedded wireless controller via the CNVi (Connectivity Integration Architecture) block for WiFi’s MAC and Bluetooth’s MAC + Baseband modem. We’ll have plenty of USB 3.0 host interfaces, and the usual PCIe and SATA 3 interfaces. Still no UFS support, but eMMC 5.1 is supported, as well as x128 DDR4, LPDDR3 and LPDDR4 memory up to 2400 MHz (No ECC support). Cache size is confirmed to be 4MB for up to four GoldMont Plus (GLM+) cores, which combined with the wider pipeline (4 vs 3) will contribute to 10% to 15% better CPU performance compared to Apollo Lake. The Gen9LP GPU in GLK processors will come with up to 18 execution unit.

Another difference will be the update of GMM speech acceleration engine to the GNA version of the SIP with support for DNN (Dynamic Neural Networks) algorithms that could be used for applications such as Microsoft Cortana.

New Features of Intel Gemini Lake Processors – HDMI 2.0, 10-bit VP9 Codec, 4-Wide Pipeline, and More

August 13th, 2017 5 comments

Most recent low power Intel mini PCs are now based on Apollo Lake family with SoC such as Celeron N3450 or Pentium N4200, but we’ve known for a while that Gemini Lake processors will succeed those starting in Q4 2017, and we can expect some Celeron/Pentium SKUs like Intel Pentium J5005 or Intel Celeron N4000, but so far I had not seen that many details. However, an anonymous tip pointed me to some interesting publicly available information.

First, a kernel patch reveals a little about the CPU pipeline:

Add perf core PMU support for Intel Goldmont Plus CPU cores:
– The init code is based on Goldmont.
– There is a new cache event list, based on the Goldmont cache event list.
– All four general-purpose performance counters support PEBS.
– The first general-purpose performance counter is for reduced skid PEBS mechanism. Using :ppp to indicate the event which want to do reduced skid PEBS.
– Goldmont Plus has 4-wide pipeline for Topdown

Goldmont Plus is the microarchitecture  used in Gemini Lake processor. Goldmont found in Apollo Lake processors only uses a 3-wide pipeline, so there should be some performance benefits here.

Another patch indicates the processor will natively support HDMI 2.0 output:

Geminilake has a native HDMI 2.0 controller, which is capable of driving clocks up to 594Mhz. This patch updates the max tmds clock limit for the same.

Apollo Lake processors only support HDMI 1.4 natively, and while HDMI 2.0 is possible, it requires an external DP to HDMI 2.0 converter, which won’t be needed in Gemini Lake processors.

The last link to Intel 2017Q2 Graphics stack page lists the supported codecs and post-processing support in Gemini Lake processors via the VAAPI driver:

Add support for Gemini Lake (aka. GLK)
– Decoding: H.264/MPEG-2/VC-1/JPEG/VP8/HEVC/HEVC 10-bit/VP9/VP9 10-bit
– Encoding: H.264/MPEG-2/JPEG/VP8/VP9/HEVC/HEVC 10-bit/AVC low power CQP mode
– VPP: CSC/scaling/NoiseReduction/Deinterlacing{Bob, MotionAdaptive, MotionCompensated}/ColorBalance/STD

Finally, as I searched more about the Goldmont Plus microarchitecture, I found Wikichip page that also claims the processor will integrate an 802.11ac wireless controller, so no external module is needed. I could not find any other reference to this last claim, except for a FanlessTech tweet also claiming DDR4, Bluetooth, and 4MB L2 cache. Gemini Lake processors will be manufactured with 14-nm process like their Apollo Lake predecessors.

Movidius Neural Compute Stick Shown to Boost Deep Learning Performance by about 3 Times on Raspberry Pi 3 Board

August 9th, 2017 14 comments

Intel recently launched Movidius Neural Compute Stick (MvNCS)for low power USB based deep learning applications such as object recognition, and after some initial confusions, we could confirm the Neural stick could also be used on ARM based platforms such as the Raspberry Pi 3. Kochi Nakamura, who wrote the code for GPU accelerated object recognition on the Raspberry Pi 3 board, got hold of one sample in order to compare the performance between GPU and MvNCS acceleration.

His first attempt was quite confusing as with GoogLeNet, Raspberry Pi 3 + MvNCS achieved an average inference time of about 560ms, against 320 ms while using VideoCore IV GPU in RPi3 board. But then it was discovered that the “stream_infer.py” demo would only use one core out of the 12 VLIW 128-bit vector SHAVE processors in Intel’s Movidius Myriad 2 VPU, and after enabling all those 12 cores instead of just one, performance increased to around 108 ms average time per inference. That’s almost 3 times faster compare to using the GPU in RPi3 for this specific demo, and it may vary for other demos / applications.

That’s the description in YouTube:

Comparison of deep learning inference acceleration by Movidius’ Neural Compute Stick (MvNCS) and by Idein’s software which uses Raspberry Pi’s GPU (VideoCore IV) without any extra computing resources.

Movidius’ demo runs GoogLeNet with 16-bit floating point precision.Average inference time is 108ms.
We used MvNC SDK 1.07.07 and their official demo script without any changes. (ncapi/py_examples/stream_infer/stream_infer.py)
It seems something is wrong with the inference results.
We recompiled graph file with -s12 option to use 12 SHAVE vector processor simultaneously.

Idein’s demo also runs GoogLeNet with 32-bit floating point precision. Average inference time is 320ms.

It’s interesting to note the GPU demo used 32-bit floating point precision, against 16-bit floating point precision on the Neural Compute Stick, although it’s unclear to me how that may affect performance of such algorithms. Intel recommends a USB 3.0 interface for MvNCS, and the Raspberry Pi 3 only comes with a USB 2.0 interface that shares the bandwidth for the USB webcam and the MvNCS, so it’s possible an ARM board with a USB 3.0 interface for the stick, and a separate USB interface for the webcam could perform better. Has anybody tested it? A USB 3.0 interface and hub would also allow to cascade several Neural Compute Sticks.