Cache Coherent Interconnect for Accelerators (CCIX)

Dark Kaeser · 23 de Maio de 2016

Accelerators Unite ARM, IBM, X86
Seven vendors forge alternative to Intel, Nvidia

Advanced Micro Devices, ARM, Huawei, IBM, Mellanox, Qualcomm and Xilinx will define the Cache Coherent Interconnect for Accelerators (CCIX, pronounced C6).
...
Beyond machine learning, the CCIX partners said their interface will also help accelerators in applications ranging from big data analytics to network processing. Many see a growing role for function-specific accelerators to assist general-purpose processors as it becomes increasingly expensive to wring more benefits of Moore’s law.
Clicar para expandir...

http://www.eetimes.com/document.asp?doc_id=1329734&page_number=1

Connecting Processor Architecture and Accelerators Seamlessly

CCIX will allow components to access and process data irrespective of where it resides, without the need for complex programming environments. This ability to access data coherently dramatically improves performance and usability by enabling a driver-less and interrupt-less usage model. Additional capabilities include:

Both off-load and bump-in-the-wire inline application acceleration while leveraging existing server ecosystems and form factors thereby improving total cost of ownership (TCO)

Higher bandwidth compared to the existing interface

Orders of magnitude improvement in application latency

Seamless processor and accelerator hardware cache coherency support

Clicar para expandir...

http://www.ccixconsortium.com/

uma alternativa ao NVLink, mais abrangente e que tal como diz a introdução vai ser angóstico à ISA do CPU.

Nemesis11 · 23 de Maio de 2016

Sem a Intel e a nVidia na lista das marcas que o suportam, não vai a lado nenhum.

Dark Kaeser · 24 de Maio de 2016

Talvez sim, talvez não.

Mas também tendo em conta que a Intel tem o seu próprio Co-processador (Phi/Knights), o seu próprio interconnect (QPI e Omni-Path) e tendo em conta o valor dispendido na aquisição da Altera, qual é que era o interesse deles mesmo?

O NVlink tem o problema de para já apenas ter o apoio da IBM, embora possa certamente ser usado com cores ARM - eles tinham se bem me recordo apresentado algo para o sector automóvel - portanto em teoria até teria interesse.

Aliás algumas das empresas deste consórcio já tinham as suas próprias soluções.

Chip Upstarts Get Coherent With Hybrid Compute

AMD, ARM Holdings, Huawei Technologies, IBM, Mellanox Technologies, Qualcomm, and Xilinx are taking on, and if they succeed in creating a more open specification to link processors and accelerators (including functions on network cards or non-volatile storage) together with coherent memory, building hybrid systems will be significantly easier than it is today. The question we have, and that the CCIX has not yet answered, is precisely how this might be done.
Clicar para expandir...

http://www.nextplatform.com/2016/05/23/chip-upstarts-get-coherent-hybrid-compute/

AMD, ARM, Huawei, IBM, Mellanox, Qualcomm, Xilinx Form CCIX Accelerator Consortium

This announcement shows that the companies involved recognize the need to work closely together to present a seamless system architecture to the their customers, as that is exactly what Intel is widely expected to do with their acquisition of the FPGA vendor Altera, reaching Accelerator Nirvana with an all-Intel solution that likely includes Intel’s coherent bus called QPI. IBM OpenPOWER, Advanced Micro Devices x86, ARM partners AMD, Qualcomm and Huawei could each go it alone, or they can join forces. To not collaborate would cede a substantial advantage to Intel and risk fragmentation that the industry would not accept.
Clicar para expandir...

http://www.forbes.com/sites/moorins...erators-ccix-fantasy-or-nirvana/#6ccd43ed6afa

^ricardo^ · 26 de Maio de 2016

Nemesis11 disse: ↑

Sem a Intel e a nVidia na lista das marcas que o suportam, não vai a lado nenhum.
Clicar para expandir...

A Intel tambem inicialmente não apoiou as extensões 64bit da AMD ao x86, e agora todos os processadores da intel de 64bit usam as extensões da AMD.

muddymind · 28 de Maio de 2016

A intel na altura estava a levar forte e feio da AMD pois os PIII coppermine e os PIV Willamette levavam forte e feio dos Athlon 64 o que levou a um grande aumento de cota de mercado da AMD. Para além disse a arquitectura de 64-bit da intel (IA-64) foi um tremendo flop no mercado de servidores.

Agora a situação é diferente dado que tanto a intel como a nVIDIA têm uma força tremenda no mercado de servidores.

Dark Kaeser · 12 de Setembro de 2017

TMSC, ARM, Xilinx, Cadence Partner on 7-nm Process

The chip will be the first demonstration in silicon of Cache Coherent Interconnect for Accelerators (CCIX) enabling multi-core high-performance ARM CPUs working via a coherent fabric with off-chip FPGA accelerators,
...
The test chip, implemented on TSMC’s 7-nm process, will be based on the latest ARM DynamIQ technology, CMN-600 coherent on-chip bus, and foundation IP.
Clicar para expandir...

The test chip provides connectivity to Xilinx’s 16-nm Virtex UltraScale+ FPGAs over CCIX chip-to-chip coherent interconnect protocol.

“Our Virtex UltraScale+ HBM family is built using TSMC’s third-generation CoWoS technology, which is now the industry standard assembly for HBM integration and cache-coherent acceleration with CCIX," said Victor Peng, chief operating officer at Xilinx.

The test chip will tape out early in the first quarter of 2018, with silicon availability expected in the second half of 2018.
Clicar para expandir...

http://www.eetimes.com/document.asp?doc_id=1332267&

Dark Kaeser · 30 de Outubro de 2017

ARM talks about CCIX details

Essentially you take PCIe, throw out the heavy parts, basically the transaction layer and above, and keep the good bits. Starting with the PCIe Data Link Layer and Physical Layer, you put a CHI based CCIX Transaction Layer on top, plus more. It looks like this.
Clicar para expandir...

https://semiaccurate.com/2017/10/26/arm-talks-ccix-details/

Dark Kaeser · 5 de Outubro de 2018

A Xilinx apresentou o Versal que designa como ACAP (adaptive compute acceleration platform) que será baseado num FPGA a 7nm (TSMC) e que deverá ser lançado durante o próximo ano

https://www.nextplatform.com/2018/10/02/inferring-the-future-of-the-fpga-and-then-making-it/

https://www.top500.org/news/xilinx-unveils-its-most-ambitious-accelerator-platform/

vandaminator · 6 de Outubro de 2018

muddymind disse: ↑

A intel na altura estava a levar forte e feio da AMD pois os PIII coppermine e os PIV Willamette levavam forte e feio dos Athlon 64 o que levou a um grande aumento de cota de mercado da AMD. Para além disse a arquitectura de 64-bit da intel (IA-64) foi um tremendo flop no mercado de servidores.

Agora a situação é diferente dado que tanto a intel como a nVIDIA têm uma força tremenda no mercado de servidores.
Clicar para expandir...

Esse PIII ia até 1GHz, o Athlon XP e 64 sairam uns anos depois

Cache Coherent Interconnect for Accelerators (CCIX)

Dark Kaeser Power Member

Nemesis11 Power Member

Dark Kaeser Power Member

^ricardo^ Power Member

muddymind 1st Folding then Sex

Dark Kaeser Power Member

Dark Kaeser Power Member

Dark Kaeser Power Member

vandaminator Power Member

Partilhar esta Página

Pesquisas Úteis

Cache Coherent Interconnect for Accelerators (CCIX)

Dark Kaeser Power Member

Nemesis11 Power Member

Dark Kaeser Power Member

^ricardo^ Power Member

muddymind 1st Folding then Sex

Dark Kaeser Power Member

Dark Kaeser Power Member

Dark Kaeser Power Member

vandaminator Power Member

Partilhar esta Página