Processador ARM for server

Apenas 1 nó do Cray CS500 com o Fujitsu A64FX.

g3BvePs.jpg


https://www.anandtech.com/show/15169/a-success-on-arm-for-hpc-we-found-a-fujitsu-a64fx-wafer
 
Comparar com um dual 24 Core não é muito justo. É ok para comparar a performance com o mesmo número de cores, mas não muito mais. A AMD tem os 7H12 com 64 cores por CPU e a Intel, 56 nos xeons 9XXX.

Também não sei se 32 GB por nó não será algo limitado. A Intel nos Phi colocou ddr4, a seguir a hmc para poder ter mais RAM disponível.
 
giphy.gif



Website.png

eMAG%208180_575px.png

The workstation is only offered with a single CPU SKU, the eMAG 8180. This isn’t to be confused with Intel’s 8180: this one has more cores!
:lol:
Avantek offers the system with three optional graphics cards: AMD FirePro W2100, a Radeon Pro WX 5100, and the NVIDIA Quadro GV100. OS options are variants of Linux: Ubuntu, CentOS, SUSE SLES, and openSUSE.
https://www.anandtech.com/show/15165/arm-server-cpus-you-can-now-buy-amperes-emag-in-a-workstation

E ainda...

Kunpeng Desktop Board

csm_kqjFVC4VtCxWQC3Tr7xwbZ_650_80_63e73d4095.jpg


Board Model D920S10
Processors 1 Kunpeng 920 processor with 4/8 cores & 2.6 GHz
Internal Storage 6 SATA 3.0 hard drive interfaces, 2 M.2 SSD slots
Memory 4 DDR4-2400 UDIMM slots, maximum capacity 64 GB
PCIe Expansion 1 PCIe 3.0 x16, 1 PCIe 3.0 x4, and 1 PCIe 3.0 x1 slots
LOM Network Ports 2 LOM NIC, supporting GE network ports or optical ports
USB 4 USB 3.0 and 4 USB 2.0
https://e.huawei.com/en/products/servers/kunpeng/kunpeng-desktop-board


Kunpeng Server Board

csm_srbxYHjSZWXZQx6czXPaTZ_650_80_3cb6203744.jpg


Board Model S920X00
Processors 2 Kunpeng 920 processors
Internal Storage Up to 16 x 3.5" SAS/SATA HDDs or SSDs, or 16 x 2.5" NVMe SSDs
Memory Up to 32 DDR4-2933 DIMMs
PCIe Expansion 8 PCIe 4.0 x8, or 3 PCIe 4.0 x16 + 2 PCIe 4.0 x8
LOM Network Ports 2 LOM NICs, supporting GE/10GE/25GE
Management Module 1 Hi1710 intelligent management chip
https://e.huawei.com/en/products/servers/kunpeng/kunpeng-server-board
 
The workstation is only offered with a single CPU SKU, the eMAG 8180. This isn’t to be confused with Intel’s 8180: this one has more cores!
Estes a copiar o nome dos cpu's Intel e os outros a chamar A64FX ao cpu ARM deles que mais parece um qualquer Athlon da AMD. Acho que estão a tentar vender gato por lebre, ou seja ARM como sendo cpu's x86.. :-D
 
No caso da Fujitsu, eles não precisam que aquele processador se confunda com Athlons. :) Está num mercado muito diferente. Já o eMAG sim, o nome 8180 é estúpido. Infelizmente não é a única empresa a praticar nomes para enganar o mais distraído. Pelo menos no mercado servidor, não há grandes desculpas para se fazer confusão.
 
Amping Up The Arm Server Roadmap

Ampere-Arm-Server-Chip-roadmap.jpg


The next-generation Ampere chip will scale up to 80 cores on a monolithic die, and will be etched in the 7 nanometer processes created by fab partner Taiwan Semiconductor Manufacturing Corp.
The Ampere Quicksilver chip will have eight memory channels, just like the eMAG 1 did, and Wittich says it will have as more memory bandwidth as the eMAG 1 provided, and that further it is getting its DRAM memory controllers from a third party as many Arm server chip makers do. The Quicksilver chip will be supporting the CCIX interface for linking to accelerators like GPUs, and will support two-socket NUMA configurations as well as single-socket implementations. CCIX will be the transport for these NUMA links.
https://www.nextplatform.com/2019/12/13/amping-up-the-arm-server-roadmap/


Looking Ahead To Marvell’s Future ThunderX Processors

marvell-thunderx-roadmap.jpg

https://www.nextplatform.com/2019/12/10/looking-ahead-to-marvells-future-thunderx-processors/
 
Interessante que nos dois casos, vão ser chips monolíticos e a Marvell não parece muito entusiasmada em relação a chiplets.

Marvell does not want to use cores and chips that are designed for laptops and desktops and then gang them up inside a single socket, chiplet style, to make a server processor.

A parte final do artigo do ThunderX também é bem interessante.

“Intel has a lot of legacy circuits that go back decades to support applications, but Arm has a clean slate architecture,” says Hegde. “We have a custom Arm core that is designed for server applications, and when we look at the performance per watt and the performance per area, we clearly see a big advantage. We have about a 20 percent die area advantage over Naples, and we have a similar power advantage. And when we move to 7 nanometers with ThunderX3, we see that our area and power advantage actually gets better. Our area compared to AMD Rome and Intel Ice Lake is better, and our power efficiency will be significantly better.

Na actual geração, o ThunderX2 é de longe a melhor proposta ARM em servidores.
 
Uma imagem do Gravitron 2 da Amazon:
vgvV86p.jpg


E um post interessante sobre ele:
The new part is Graviton2 and this is an exceptional server processor that will be a key part of the EC2 compute offering powering the M6g (general purpose), M6gd (general purpose with SSD block storage) the C6g (compute optimized), the R6g (memory optimized) and the R6gd (memory optimized with SSD block storage) instance families. This 7nm part is based upon customized 64-bit ARM Neoverse N1 cores and it is smoking fast. Rather than being offered as an alternative instance type that will run some workloads with better price/performance, it’s being offered as a better version of an existing, very highly-used EC2 instance type, the M5.

Here’s comparative data between M6g and M5, the previous generation instance type, from an excellent Forbes article by Patrick Moorhead:

  • >40% better integer performance on SPECint2017 rate (estimate)
  • >20% better floating-point performance on SPECfp2017 Rate (estimate)
  • >20% better web serving performance on NGINX
  • >40% better performance on Memcached with lower latency and higher throughput
  • >20% better media encoding performance for uncompressed 1080p to H.264 video
  • 25% better BERT ML inference
  • >50% better EDA performance on Cadence Xcelium EDA tool
This is a fast part and I believe there is a high probability we are now looking at what will become the first high volume ARM Server. More speeds and feeds:

  • >30B transistors in 7nm process
  • 64KB icache, 64KB dcache, and 1MB L2 cache
  • 2TB/s internal, full-mesh fabric
  • Each vCPU is a full non-shared core (not SMT)
  • Dual SIMD pipelines/core including ML optimized int8 and fp16
  • Fully cache coherent L1 cache
  • 100% encrypted DRAM
  • 8 DRAM channels at 3200 Mhz
The Anapurna team at AWS is doing amazing work. I wish I could show you all the work they currently have underway but only some of it is public. Even with multiple, difficult competing projects concurrently underway, they delivered Graviton2 on an unusually short schedule seldom seen in the semi-conductor world. It’s a great team to work with and Graviton2 is impressive work.

ARM Servers have been inevitable for a long time but it’s great to finally see them here and in customers hands in large numbers.

https://perspectives.mvdirona.com/2020/01/aws-graviton2/

Atenção que os M5 usam Intel Xeons. Não vejo uma comparação com o Epyc.
 
O novo Ampere Altra:
  • 80 cores ARM [email protected] Ghz
  • Sem SMT
  • 1 MB L2 por core e 32 MB L3 Totais
  • Não é soldado e pode ser usado em 1 ou 2 sockets
  • 128 Lanes Pci-Ex com 1 socket e 192 Lanes Pci-Ex com 2 sockets
  • Suporte CCIX
e1kQWZh.jpg


Gráficos:
wNsydW1.jpg


a1nrshM.jpg


9bqXfNU.jpg


Servidores 1P (Gigabyte) e 2P (Provavelmente Lenovo):
wu6EIXg.jpg


https://www.servethehome.com/ampere-altra-80-arm-cores-for-cloud/

Notar que os scores do Epyc e Xeon foram "diminuidos" artificialmente, por estarem a comparar usando o gcc. E os benchs foram feitos com uma versão a 3.3 Ghz, quando o topo a nível comercial parece ser a 3.0 Ghz.
 
Eu percebi bem? Eles estão a usar o SoC de um módulo COM e depois montado numa "daughter" board com o formato mini-itx, à lá Pentium II Style?
 
Sim claro, mas eles afinal produzem o quê?
A maioria das empresas embedded produz em formatos tradicionais estranhos como os COM, SBC, Qseven, entre outros.
Eles decidiram produzir uma série de daughter boards só porque sim?
 
Back
Topo