If you receive high traffic loads, work with AI models or workloads or have lots of data that needs power and performance to back it, then a GPU Server might be just the ticket. In this blog, we’ll break down all the ins and outs, tech details and specs of our GPU Servers, as well as advise you on whether this could be the right server for you.

Let’s dive in.

Server model specs

External name

CPU

RAM

Storage

Price

NVIDIA Ampere A10 AMD EPYC 7313P 16 cores x 3.0 GHz NVIDIA A10 GPU 960GB NVMe (2x 960GB) Hardware RAID 1 £0.8194 per hour
Intel DC GPU Flex 170 Intel Xeon Gold 5412U 24 cores x 2.1 GHz Intel Data Centre GPU Flex 170 1.92TB NVMe (2x 1.92TB) Hardware RAID 1 £0.8611 per hour

NVIDIA A10

  • CUDA cores: 3840
  • Memory: 24 GB GDDR6
  • Memory bus: 384-bit
  • Memory bandwidth: 600 GB/s
  • Single-precision performance: up to 9.5 TFLOPS
  • Double-precision performance: up to 4.7 TFLOPS
  • Tensor performance: up to 75 TFLOPS (with sparsity)
  • GPU clock speed: 825 MHz (boost), 745 MHz (base)
  • Architecture: Ampere
  • TDP: 150W
  • Launch date: 2021
  • Targeted applications: AI, HPC, cloud gaming, virtualisation
  • Xe cores: 128
  • Memory: 16 GB of GDDR6 or 16 GB of HBM2
  • Memory bus: 256-bit (GDDR6), 1024-bit (HBM2)
  • Memory bandwidth: 320 GB/s (GDDR6), 1024 GB/s (HBM2)
  • Single-precision performance: up to 3.5 TFLOPS
  • Double-precision performance: up to 1.75 TFLOPS
  • Tensor performance: up to 26.6 TFLOPS
  • GPU clock speed: 1.5 GHz (max)
  • Architecture: Xe-HPC
  • TDP: 75W (single-precision), 125W (HBM2)
  • Launch date: 2022 (first production silicon shipped to select customers in 2022)
  • Targeted applications: cloud gaming, media transcoding, AI, high-performance computing

Why should you choose a GPU Server?

GPU servers come with loads of benefits and, depending on what you need it for, this type of server is perfect for powerful projects. Here’s a few other reasons why a GPU server could be right for you:

  • AI training – Our GPU servers are optimised for parallel processing and open-source LLMs. This lets businesses achieve precise data analysis while reducing errors and risks, allowing them to focus on developing their core business.
  • High-performance computing – These servers are also designed for high-performance computing, allowing businesses to quickly process large data volumes and make informed decisions. This is particularly useful for real-time input validation, task automation, and integrating accurate forecasts into workflows.
  • Deep learning – These capabilities are enhanced with our GPU servers, which can efficiently evaluate large, complex data sets. This improves accuracy and speed for image, text, and data analysis.
  • Big data analysis – Streamline large analysis tasks, and process large data sets centrally and efficiently. This enables businesses to gain deeper insights, standardise data handling, and evaluate data faster, driving informed decision-making and business success.
  • Hit play or pause – start or stop your plan at the end of the month, so you always have control. 

What’s the difference?

We’ve broken down the specs of each server, but what actually is the difference between the two? 

The NVIDIA A10 has more CUDA cores, which generally means higher raw processing performance, especially in single-precision workloads. However, the Intel Flex 170 has a more efficient architecture. This results in lower TDP and potentially lower power consumption. 

Another thing the A10 offers is a higher memory bandwidth and more memory, so certain workloads can benefit considerably from this. 

Looking more specifically at what the Intel 170 can support, if you run AI or HPC workloads, this kind of server is more suited to you.

And finally, the NVIDIA A10 supports PCIe 4.0, while the Intel 170 supports PCIe 4.0 as well as CXL (Computer Express Link).

Use cases

Another difference between NVIDIA A10 and Intel 170 is what they can be used for. While most of their use cases are very similar, there are some small variations.

The NVIDIA A10 is primarily targeted at cloud gaming, AI, HPC and virtualisation applications, with a strong focus on data centre applications. 

The Intel 170 can also be used for cloud gaming, AI and HPC applications, as well as media transcoding – with a stronger emphasis on efficiency, power consumption and flexibility.

Get in touch

If you’d like to learn more about our GPU Servers, or have any specific questions, please get in touch. Our team is available 24/7 to help with anything you need. Just give us a call on 0333 271 8682 or message us via live chat.