Using the GPU is not expensive; it is often already there

In this post, David Stevenson, product manager for the Mako Core print software development kit, explains how using the graphics processing unit (GPU) alongside the central processing unit (CPU) to render images for print is not the costly route that is sometimes imagined.

When I talk to customers and prospects about Mako’s GPU-native renderer, Apex, their reaction is first excitement when they see how fast it handles complex PDF transparency, followed by concern that adding a GPU to their solution might not be cost effective. Many engineers still associate GPU acceleration with expensive discrete graphics cards, major hardware investments and proprietary software stacks but that’s an outdated view.

Most CPUs already include a GPU

Modern Intel and AMD processors typically include integrated graphics, such as Intel Core processors with UHD, Iris Xe or Arc integrated graphics, and AMD Ryzen APU models with Radeon graphics. This means no additional hardware purchase is required. The GPU is already part of the CPU you have installed, so enabling GPU acceleration often means using hardware you have already paid for.

Are discrete GPUs expensive?

When higher throughput is required, adding a discrete GPU is far less expensive than many expect. Two decades of gaming has meant that even modestly priced GPUs offer incredible graphics performance. At the time of writing, typical current market pricing indicates:

Entry level GPUs such as NVIDIA RTX 3050 or Intel Arc A580: approximately 200 to 300 USD
Mid-range GPUs such as RTX 4060 or AMD RX 7600: approximately 300 to 400 USD
Upper mid-tier GPUs such as RTX 4070: approximately 500 to 600 USD

These prices are comparable to many standard IT hardware upgrades and are small relative to the cost of additional servers. High-end data centre GPUs exist, but they are not required for Apex imaging and raster workloads.

Vulkan supports integrated and discrete GPUs

Apex uses Vulkan, a modern, cross-platform, low overhead API to drive the GPU. Vulkan runs on integrated and discrete GPUs from NVIDIA, AMD, Intel and Apple. Support across Windows and Linux is mature and vendor maintained. This means solution providers can choose a GPU that fits their desired price/performance.

The real cost is leaving performance idle

The real cost question is not whether a GPU costs a few hundred dollars, but how much it costs to process everything on the CPU. CPU-only processing can mean:

Higher sustained CPU utilization
Longer job completion times
More cores required to meet throughput targets
Additional servers to scale performance

If GPU acceleration reduces processing time significantly, it can increase throughput without increasing machine count. That is not an added expense, it’s a more efficient use of existing hardware.

GPU acceleration is an efficiency decision

GPU acceleration is not about luxury hardware; it’s about matching the workload to the right architecture. With Mako’s Apex renderer, many workloads involve repeated operations across large pixel buffers. These are highly parallel tasks and GPUs are designed specifically for this type of computation.

Most modern systems already include GPU capability. Vulkan provides broad support. Entry and mid-range discrete GPUs cost hundreds, not thousands, of dollars. The question is no longer whether GPUs are expensive, it’s whether it makes sense to ignore hardware that is already available.

About the author

David Stevenson, product manager for Mako Core SDK at Global Graphics Software — David Stevenson, Product Manager, Mako Core SDK

David Stevenson is the product manager for the Mako Core™ SDK and responsible for the performance component in SmartDFE™, the AI-accelerated digital front end platform for high-speed, single-pass inkjet presses. Throughout his career, he has specialized in electronic documents, starting with Xerox Corporation as a product manager for Venture Publisher, an early star of desktop publishing on PCs. That was followed by a 13-year career at Adobe, specializing in various aspects of PDF, including creative print workflows, electronic forms and accessibility. At Helix he continues to focus on PDF technology and solutions for page description languages (PDLs).