AIME A8004 - Deep Learning Server

If you are looking for a server specialized in maximum deep learning training, inference performance and for the highest demands in HPC computing, the AIME A8004 multi-GPU 4U rack server takes on the task of delivering.

The AIME A8004 is based on the new ASUS ESC8000A-E12 barebone which is powered by two AMD EPYC™ Genoa processors, each with up to 96 cores. Totaling a CPU performance of up to 384 parallel computable CPU threads.

Its GPU-optimized design with high air flow cooling allows the use of eight high-end double-slot GPUs like the latest NVIDIA H100, NVIDIA A100, NVIDIA L40S, RTX 6000 Ada, RTX 5000 Ada GPU models.

Definable GPU Configuration

Choose the desired configuration among the most powerfull NVIDIA GPUs for Deep Learning and Rendering:

Up to 8x NVIDIA H100

The NVIDIA H100 is the flagship of the NVIDIA Hopper processor generation and the successor to the common NVIDIA A100 accelerator cards. The NVIDIA H100 is based on the GH-100 processor in TSMC 4N manufacturing with 14592 CUDA cores, 456 fourth-generation Tensor cores and 80 GB HBM2 memory. A single NVIDIA H100 GPU already breaks the peta-fp16 performance barrier. Eight accelerators of this type add up to more than 6.000 teraFLOPS tf32 and up to 72.000 teraTOPS fp8 performance. The NVIDIA H100 is currently the most efficient and fastest deep learning accelerator card available. The H100 is also the first accelerator card that benefits from up to 64 GB/s transfer rates through the PCIe 5.0 bus of the AIME A8004.

Up to 8x NVIDIA A100

The NVIDIA A100 is the flagship of the NVIDIA Ampere processor generation and the successor to the legendary NVIDIA Tesla V100 accelerator cards. The NVIDIA A100 is based on the GA-100 processor in 7nm manufacturing with 6912 CUDA cores, 432 third-generation Tensor cores and 40 or 80 GB HBM2 memory with the highest data transfer rates. A single NVIDIA A100 GPU already breaks the peta-TOPS performance barrier. Eight accelerators of this type add up to more than 2.000 teraFLOPS fp32 performance.

Up to 8x NVIDIA RTX 6000 Ada

The RTX ™ 6000 Ada is built on the latest NVIDIA GPU architecture: Ada Lovelace. It is the direct succesor of the RTX A6000 and the Quadro RTX 8000. The RTX 6000 Ada combines 568 fourth-generation Tensor Cores, and 18.176 next-gen CUDA® cores with 48GB of graphics memory for unprecedented rendering, AI, graphics, and compute performance.

Up to 8x NVIDIA RTX 5000 Ada

The RTX ™ 5000 Ada is built on the latest NVIDIA GPU architecture: Ada Lovelace. It is the direct succesor of the RTX A5000/A5500 and the Quadro RTX 6000. The RTX 5000 Ada combines 400 fourth-generation Tensor Cores, and 12.800 next-gen CUDA® cores with 32GB of graphics memory for a convincing rendering, AI, graphics, and compute performance.

Up to 8x NVIDIA RTX A5000

With its 8,192 CUDA and 256 Tensor cores of the Ampere generation, the NVIDIA RTX A5000 has about the performance of a RTX 3090. However, with its 230 watts of power consumption and 24 GB of memory, it is a very efficient accelerator card and especially for inference tasks a very interesting option.

All NVIDIA GPUs are supported by NVIDIA’s CUDA-X AI SDK, including cuDNN, TensorRT which power nearly all popular deep learning frameworks.

Dual EPYC CPU Power

The latest AMD EPYC server CPU with support for DDR5 and PCIe 5.0 delivers up to 96 cores with a total of 192 threads per CPU with an unbeaten price performance ratio.

The available 2x 128 PCI 5.0 CPU lanes of the AMD EPYC CPU allow highest interconnect and data transfer rates between the CPU and the GPUs and ensures that all GPUs are connected with full x16 PCI 5.0 bandwidth.

A large amount of available CPU cores can improve the performance dramatically in case the CPU is used for preprocessing and delivering of data to optimaly feed the GPUs with workloads.

Up to 30 TB direct NVMe SSD Storage

Deep Learning is most often linked to high amount of data to be processed and stored. A high throughput and fast access times to the data are essential for fast turn around times.

The AIME A8004 can be configured with up to two exchangeable U.2 NVMe triple level cell (TLC) SSDs with a capacity of up to 15.36 TB each, which adds up to a total capacity of 30 TB of fastest NVMe SSD storage.

Since each of the SSDs is directly connected to the CPU and the main memory via PCI 4.0 lanes, they achieve consistently high read and write rates of more than 4000 MB/s.

Optional RAID: Up to 60 TB NVMe SSD Storage

For further storage of large datasets and training checkpoints often additional storage capacity is required. The A8004 offers the option to use its six additional drive bays with a hardware reliable RAID configuration. Up to 60 TB fastest NVME SSD storage with RAID levels 0 / 1 / 5 / 10 and 50.

As usual in the server sector, the SSDs have an MTBF of 2,000,000 hours with 1 DWPD and a 5-year manufacturer's guarantee.

High Connectivity and Management Interface

The A8004 can be fitted with standard 2x 10 Gbit/s RJ45/SFP+ LAN ports or with up to 2x 100 Gbit/s (GBE) network adapters for highest interconnect to NAS resources and big data collections. Also for data interchange in a distributed computing cluster the highest available LAN connectivity is a must have.

The AIME A8004 is completely remote manageable through ASMB11-iKVM IPMI/BMC, powered by AST2600, which makes a successful integration of the AIME A8004 into larger server clusters possible.

Optimized for Multi GPU Server Applications

The AIME A8004 offers energy efficiency with redundant Titanium grade power supplies, which enable long time fail-safe operation.

Its thermal control technology provides more efficient power consumption for large-scale environments.

All setup, configured and tuned for perfect Multi GPU and deep learning performance by AIME.

The A8004 comes with preinstalled Linux OS configured with latest drivers and frameworks like Tensorflow, Keras and PyTorch. Ready after boot up to start right away to accelerate your deep learning applications.