Gráfica AMD Vega II 7nm (Radeon VII, Vega II Pro Duo)

[Dark Kaeser]

@hefguerra no 1º post está lá essa informação (menos o preço), é só ir às specs, a configuração base começa nos Xeon 8 cores, 32GB Ram e Radeon 580x, o topo que foi o que postei a seguir é que vai até aos Xeon 28 cores, 1.5TB Ram e 2x Radeon Pro Vega Duo
https://www.apple.com/mac-pro/specs/

EDIT: apenas coloquei a "novidade" por assim dizer, a Vega VII duo, poi a Vega VII já se conhece e há de haver um tópico dedicado ao Mac Pro no tópico da Apple.

Não deixa de ser curioso terem como opção um FPGA para processamento de vídeo

02:26PM EDT - New Apple PCIe card: Afterburner. FPGA card for ProRes/ProRes RAW video decoding acceleration
02:26PM EDT - Up to 3 streams of 8K ProRes Raw

https://www.anandtech.com/show/14479/the-apple-wwdc-2019-keynote-live-blog

 

[Dark Kaeser]

Novo update ao ROCm, que entre as várias melhorias específicas para as Radeon Instinct MI50 e MI60, sendo de destacar

New features and enhancements in ROCm 2.5

AMD Infinity Fabric™ Link enablement
The ability to connect four Radeon Instinct MI60 or Radeon Instinct MI50 boards in one hive via AMD Infinity Fabric™ Link GPU interconnect technology has been added.

https://github.com/RadeonOpenCompute/ROCm#new-features-and-enhancements-in-rocm-25

 

[Dark Kaeser]

Só mesmo pela curiosidade...

Mentor and AMD verify massive Radeon Instinct Vega20 IC design on AMD EPYC in ~10 hours with ecosystem partners Microsoft Azure and TSMC

AMD engineers executed a physical verification pass of the Radeon Instinct™ Vega20 -- its largest 7nm chip design -- in ~10 hours using the TSMC-certified Calibre™ nmDRC software platform from Mentor, a Siemens business, running on the Microsoft Azure cloud platform using HB-series virtual machines, powered by AMD EPYC™ processors.

Using TSMC 7nm Calibre design kits running in Azure, AMD successfully completed two verification passes in ~19 hours – a dramatic reduction in total physical verification turnaround time, despite the massive 13.2B transistors on the AMD design. In addition, AMD scaled Calibre nmDRC out to 4,140 cores across 69 HB virtual machines, enabling its engineers to balance tight deadlines against demanding resource requirements and other costs.

https://semiwiki.com/forum/index.ph...stem-partners-microsoft-azure-and-tsmc.11599/

 

[Dark Kaeser]

Penguin Computing Upgrades Corona with Latest AMD Radeon Instinct GPU Technology for Enhanced ML and AI Capabilities

As previously released, the cluster consists of 170 two-socket nodes with 24-core AMD EPYCTM 7401 processors and a PCIe 1.6 Terabyte (TB) nonvolatile (solid-state) memory device. Each Corona compute node is GPU-ready with half of those nodes today utilizing four AMD Radeon Instinct MI25 accelerators per node, delivering 4.2 petaFLOPS of FP32 peak performance. With the MI60 upgrade, the cluster increases its potential PFLOPS peak performance to 9.45 petaFLOPS of FP32 peak performance. This brings significantly greater performance and AI capabilities to the research communities.

AMD’s Radeon Instinct MI60 accelerators bring many new features that improve performance, including the Vega 7nm GPU architecture and the AMD Infinity FabricTM Link technology, a peer-to-peer GPU communications technology that delivers up to 184 GB/s transfer speeds between GPUs – which is 5.75X faster than PCIe Gen 3, and full-chip Error-correcting code (ECC)11 and Reliability, Accessibility and Serviceability (RAS) 12 technologies. The new accelerators also utilize the latest ROCm open source software stack, which is now integrated into leading frameworks like TensorFlow and PyTorch and maps workloads to the heterogeneous compute resources of the underlying hardware.

https://www.penguincomputing.com/co...chnology-for-enhanced-ml-and-ai-capabilities/

Tudo isto vai aparecendo na SC19 (Supercomputing '19) e coincide com o anuncio do ROCm 3.0

AMD released version 3.0 of its ROCm Open Software Platform. New features include:

• Introduction of ROCm 3.0 with new innovations to support HIP-clang – a compiler built upon LLVM, improved CUDA conversion capability with hipify-clang, library optimizations for both HPC and ML.
• ROCm upstream integration into leading TensorFlow and PyTorch machine learning frameworks for applications like reinforcement learning, autonomous driving, and image and video detection.
• Expanded acceleration support for HPC programing models and applications like OpenMP programing, LAMMPS, and NAMD.
• New support for system and workload deployment tools like Kubernetes, Singularity, SLURM, TAU and others.

 

[Dark Kaeser]

Bom lá se decidiram...

AMD Radeon Pro VII is a PCIe Gen 4.0 graphics card...

The card has a typical board power of 250W, which is slightly less than RTX 5000 (260W). It requires a single 8-pin and a single 6-pin power connector, which is another difference compared to the RX model, as the card was spec’ed at 295W TDB with dual 8-pin.

AMD is also bringing back a new multi-GPU bridge for the Radeon Pro VII. The bridge is called ‘Infinity Fabric Link’ and it has a bandwidth up to 168 GB/s.

https://videocardz.com/newz/amd-ann...ing-16gb-hbm2-memory-and-infinity-fabric-link

As únicas novidades face a Radeon VII é o TDP ter descido para os 250W, ter PCIe-4 e agora passou a ter um bridge para multi-GPU.