LATEST NEWS|HIGHRESO

GPU MAKES SMILE

GPU servers provide the IT infrastructure flexibility to handle core solutions required in new digital era. This includes AI and deep learning, high-quality rendering resolution for 8K broadcasting, auto-driving simulators, computational fluid dynamics etc. HIGHRESO Inc.’s in-house GPU specialized data center possesses and runs in-house developed servers.
We are the number one in Japan in terms of the number of GPU data center servers and the experience of operating GPU data centers.

On year 2019, we assemble our skill and experts at Shika machi, Ishikawa prefecture, and built a large scale “Shika machi first GPU data center” which is totally developed and operated by HIGHRESO. Starting from GPU cloud service “SOROBAN”, to the collaborative research with “Japan Advanced Institute of Science and Technology (JAIST)”, we share high spec environments with low prices not limited to private enterprises but also to academic research institutes. We believe that we can increase the number of smiles by providing these creative technologies to people that need them. Following is the introduction of HIGHRESO GPU business.

A bright future driven by GPU

CPU (Central Processing Unit) are the brains of PC and servers taking the commanding role. On the other hand, GPU (Graphics Processing Unit) manipulates computer graphics and image processing as if a factory handling mass operation.

GPU is useful in algorithms that process that process large blocks or data in parallel, for example 3D graphics etc. Recently GPU servers are being focused and utilized for its high computing performance.

Difference between CPU and GPU

Both CPU and GPU may perform computation to run PCs and servers, however they each have their own specifications. CPU plays the role of brain as it is excellent at summarizing and processing data sent to computer by HDD, memories, OS, programs, etc. High level of performance is achieved by supercomputers that contains CPU servers consolidating a number of CPUs with high computing capability. On the other hand, GPU servers combine a pile of GPUs that are experts of handling simple calculation simultaneously. For the same reason, AI and deep learning, as well as high resolution graphics rendering, are getting attention as computers that may compute calculations faster than CPU servers.

Processing the work in sequence order
Higher speed forecasting the next processing, in order to process in sequences
Specialized to parallel processing
Core may be simplified as same processes are perform at once

GPU market expansion in Japan is limited

When calculation is necessary for most services nowadays, GPU servers are important “calculation resources” but Japan falls behind as of now.

The following are some possible reasons:

  • High technical experience are necessary to assembly and configure AMD’s (Advanced Micro Devices) GPU to perform them capably
  • Not enough experiences setting up and operating data centers with a pile of GPU servers
  • Introductory barrier to NVIDIA’s licensed GPU servers due to the high prices
  • Disaster risk is high (for example earthquake)
  • Lack of know-how in operating GPU servers stably as GPU servers generates heat constantly

We think that not only establishing large-scale GPU data center in Japan, but it is also possible to provide high performance “Made in Japan calculation resources” safely once we solve the above obstacles.

Why do HIGHRESO use NVIDIA GPUs

NVIDIA’s products are mainly being used in GPU server market currently.
This is backed by the evolution of CUDA (Compute Unified Device Architecture) platform that is generally used for GPU computational performance. Programmers may operate GPU’s computational capability through customizing “C language” that is a commonly used programming language. AI Deep Learning framework “TensorFlow” etc. originated by Google are produced based on CUDA.

HIGHRESO is the first company in Japan to sign “Cloud Service Provider Compute Agreement” with NVIDIA – to participate in NVIDIA partner program for Cloud/IaaS (Infrastructure as a service) field.
Thanks to our close partnership with NVIDIA, we installed a number of revolutionized “NVIDIA A100” GPU installed in our data centers, contributing to a smooth and ultimate cloud service.

HIGHRESO’s expertise in operating GPUs is gained through the experience of operating more than 20,000pcs of GPU, building in-house GPU servers and managing them consistently. We are now able to provide GPU servers at reasonable prices.

Why GPU servers are necessary in current and the next generation

Scientist and researchers using supercomputers with CPUs realized the high computational abilities of GPU in mass calculation like weather and earthquake simulation. Formulating the technology using GPU to replace part of the calculations that had been carried out by CPU gradually, and this is being practically used in high resolution 3D rendering, AI Deep Learning and various fields.

In the near future, these contemporary technologies will be practically realized, and more users will be enjoying convenient and safe services. “Computing resources” will become important society infrastructures as enormous calculations are to be carried out through various networks.

Largest GPU data center in Japan

Shika machi First Data Center in Ishikawa prefecture is operating 9,000pcs of GPU, which is the largest data center in Japan. The computing capability of our servers equipped with the latest generation GPU NVIDIA A100 achieve a peak of 312TFLOPS.*
Using our servers equipped with NVIDIA A100, we provide GPU cloud service “SOROBAN” at an affordable price.
*FP32 Tensor core, with sparsity

TOP

LOADING ...