Graphic cards

Graphic cards - essential in almost every application area

Whether you work on complex 3D designs with your workstation systems or keep your finger on the pulse of the latest trends in artificial intelligence, machine learning and virtualisation - you always need high-performance graphics or computing accelerators for your workstations and servers. Because in the field of professional industrial applications, nothing runs without powerful graphics cards.

NVIDIA or AMD - certified graphic cards for your applications

To ensure that demanding software solutions such as CAD applications and video editing software function properly, their manufacturers recommend and certify various professional graphics cards for use (ISV certification). Simple graphics chips, as they are often integrated in processors, usually do not offer the necessary performance here and are therefore unsuitable for these areas of application. So there is no way around a dedicated graphics card. The US manufacturers NVIDIA and AMD are highly specialised in this segment and serve almost every application area and performance spectrum with their very broad product ranges.

For workstations, NVIDIA offers powerful options in various price and performance classes with the graphics card models of the Quadro and RTX series, while the Tesla models and their successors are intended for use in data centres. The newer models from NVIDIA do not use the names Quadro and Tesla. Instead, they are based on the architecture used in each case and bear a corresponding name, such as the models NVIDIA RTX A4000 and NVIDIA A100, which are based on the Ampere architecture.

AMD's product portfolio also includes the Radeon Pro and FirePro series of workstation graphics cards for many performance requirements and almost every budget. The FirePro S models and their successors, the AMD Instinct series, are available for server computing accelerators.

Graphics cards for workstations and servers - the differences

The structure of server and workstation as well as the tasks of the graphics cards in the respective areas differ fundamentally. Therefore, apart from the form factor and the connection via PCI Express (PCIe), there are only a few similarities.

The computing accelerators in the server area are primarily designed to improve computing performance and for virtualised environments. Thus, they usually do not have to output direct video signals and usually do not have video outputs for connecting monitors. In addition, cooling is often semi-passive. In this case, the existing strong air flow in the server housing is used via air conduction systems to flow to the heat sink of the graphics card, thus ensuring optimal cooling.

A workstation, on the other hand, is operated like a conventional desktop PC and graphics accelerators are predominantly installed. 3D models and graphics can be displayed on several high-resolution professional monitors connected via HDMI and Displayport ports. The older standards VGA and DVI are being used less and less due to technical limitations in resolution and image quality. Depending on the model, high-performance workstation graphics cards are equipped with one or more fans for active cooling to ensure reliable and rapid dissipation of the heat generated. To enable the quietest possible operation, the airflow in the enclosure is significantly weaker here and in the vast majority of cases insufficient for semi-passive cooling of a graphics card.

ServerShop24 - Refurbished graphic cards, available and reasonably priced

A new purchase is usually associated with high costs and long delivery times. You can avoid these factors by buying a refurbished graphics card from our range. If you need a quick and inexpensive replacement in case of a defect or if you want to upgrade your systems in a cost-conscious way, we are your reliable partner for used and tested server parts, components and accessories with more than 10 years of experience. You are also at the right address in our online shop if your application does not require the latest high-end accelerators. Because second-hand, older graphics card models from our stock can often score points with their excellent price-performance ratio and still fully meet your requirements. Our experienced team will advise you in a friendly and competent manner if you have any questions about your order. Please use our ticket system or one of the other contact options.

Graphics processor and video memory - top performance through advanced technology

The performance potential of graphic cards is primarily determined by the architecture of the graphics processor (GPU) and the type of graphics memory (VRAM) used. The number and structure of the available computing units affects how many data streams can be processed in parallel and thus the speed of the calculation.

Ray tracing cores, for example, accelerate rendering workloads and bring massive speed advantages in the areas of architecture, construction and in the creation of prototypes. Tensor cores, on the other hand, support processes with very high data volumes, such as in the field of artificial intelligence (AI) and machine learning (ML) with neural networks. Complex simulations in high-performance computing (HPC) also benefit from the performance increase provided by the tensor cores.

The graphics memory serves as a dedicated high-performance memory for the graphics card and is used independently of the system's main memory. The size of the video memory varies here depending on the area of application and performance class. High-end accelerators have up to 40 GB VRAM and sometimes even more. Depending on the model and generation, GDDR memory (Graphics Double Data Rate) is used here in various specifications, which is characterised by particularly high transfer rates for memory-intensive workloads. However, various graphics card models are also available with HBM memory (High Bandwidth Memory). This offers advantages in memory bandwidth and power consumption.

If your application is too demanding for a single GPU, or you want to speed up your processes further, some models still offer the option of connecting the processing power of several graphics cards together. NVIDIA has developed SLI and NVLink technology for this purpose. By integrating additional GPUs, you may be able to delay the purchase of a completely new generation of graphics cards. If your systems do not have enough free space for multiple graphics cards, you can also use variants with multiple graphics processors (multi-GPU) for high performance requirements. In this case, several GPUs are installed on one graphics card, each with its own video memory. The achievable performance increases significantly with supported applications.

Before buying a graphic card - What should be considered?

Before buying a graphics card, the requirements for trouble-free installation and operation should be checked. On the one hand, it is important to consider the form factor. Professional graphics cards are available in single- or dual-slot designs, depending on their performance. The available free space in your system therefore determines which variant you can install.

Furthermore, the power supply is an important factor. Under full load, a power requirement (TDP) of more than 200 watts is not unusual for a modern high-performance graphics card. Taking the rest of the system into account, it is therefore necessary to check whether the existing power supply unit offers sufficient power reserves for the use of the desired graphics card.

Depending on the applications you work with, the supported versions of the programming interfaces (APIs) also play a decisive role. Here it is important to check whether the selected graphics card can be used for the required versions of DirectX, OpenGL and Vulkan, for example.