AMD may scrap the high-end options of its next generation of Radeon gaming GPUs to divert scarce resources into building GPUs for AI and high-performance computing (HPC) instead – a segment that’s undergoing something of a boom.
When AMD launches its RDNA 4 family of GPUs, possibly next year, there won’t be an AMD Radeon RX 8800 or 8900, according to TechSpot. This will give its rival Nvidia a clear run at manufacturing the best GPUs to meet the high-end gaming market, but could also serve to constrain supply and spike prices.
The line-up will resemble the RDNA 1 family of AMD GPUs, according to sources speaking to the publication, where the most powerful entry was the RX 5700 XT GPU. Subsequent generations included higher-end units such as 6800, 6900, and 6950 in RDNA 2, and 7800 and 7900 in last year’s RDNA 3 series.
AMD wants in on the AI boom
This rationale is simple. There’s a rush for hardware and components to service generative AI workloads – alongside a limited supply of resources and manufacturing capacity – and AMD wants to get in on the action.
Indeed, this is a segment in which there’s currently a shortage, with chipmaking giant TSMC lacking the capacity to ramp up production from vendors like Nvidia to meet industry demand.
Nvidia’s A100 and H100 chips, incidentally, currently lead the way in an AI servers market that’s reportedly set to surge to $150 billion by 2027, with AMD hoping to be a part of it. The main reason is Nvidia is enjoying profit margins of 823%, according to sister site Tom’s Hardware, on its H100 GPUs.
Rather than diverting semiconductors into its high-end consumer GPUs, the firm will focus on field programmable gate arrays (FPGAs) and general-purpose graphics processing units (GPGPUs). This is according to Brits and Chips - Eng.
The circuitry of the former is highly suited to machine learning and deep learning, while the latter are GPUs that also handle computational workloads normally undertaken by the best CPUs. They’re both ideal to meet the rising demand for GPUs for AI.
With constrained supply, however, does does mean we may see a return to the shortages and spiking prices for GPUs that we last saw in 2020. With fewer options, gamers may find themselves paying above the odds when building PCs, for example.