External GPUs Face Challenges Despite Thunderbolt 5 Launch

The launch of Thunderbolt 5 has reignited interest in external graphics processing units (eGPUs), but significant limitations remain. While eGPUs have found a niche market over the past decade, the slow adoption of Thunderbolt 5 technology has affected their overall effectiveness. For gaming enthusiasts and professionals requiring extra graphical power, understanding the realities of eGPUs is crucial before investing.

Thunderbolt 5: Progress and Limitations

Thunderbolt 5 is now available on several computer models, though it faces slow adoption as many manufacturers opt for USB4 instead. This trend does not spell doom for Thunderbolt eGPUs, as USB4 maintains compatibility and offers performance comparable to Thunderbolt 4. However, expectations should be moderated; even with Thunderbolt 5, eGPUs typically experience a performance hit compared to internal GPUs.

Historically, eGPUs have delivered between 70% and 90% of their desktop performance potential, a trend likely to continue. While Apple has transitioned to Thunderbolt 5 for all new Macs, compatibility issues arise as GPU drivers do not function on Apple Silicon, limiting their effectiveness on these devices. Windows users have reported compatibility concerns, including instances of CPU throttling when using Thunderbolt eGPUs.

Newer eGPU enclosures promise increased bandwidth, but a sole focus on bandwidth specifications may overlook broader challenges. Thunderbolt 5 offers up to 80 Gbps of bandwidth (or 120 Gbps with boost), significantly exceeding the 40 Gbps maximum of Thunderbolt 3 and 4. Yet, the Thunderbolt controller introduces latency, particularly when used with laptops or handheld devices that utilize their internal displays. While connecting to an external monitor may alleviate some latency, the performance reduction can still be significant.

The Future of eGPUs and Alternative Solutions

Despite the advancements of Thunderbolt 5, it still lacks the necessary bandwidth for high-end graphics cards, which are often the preferred choice for eGPU setups. As integrated graphics processing units (iGPUs) become more powerful, the need for eGPUs may diminish. Nevertheless, a core group of enthusiasts remains drawn to the concept, willing to navigate the trade-offs involved.

Midrange GPUs may present a more suitable option for eGPU users, as they do not demand the same high bandwidth as flagship models. Cards such as the AMD RX 9060 XT or Nvidia RTX 5060 Ti are likely to deliver closer to their full potential, providing a valuable boost in frames per second without overwhelming the available bandwidth. However, potential users must consider the high costs associated with eGPU enclosures and the necessary power supply units, which can make them a more expensive option compared to building a desktop PC.

While Thunderbolt remains a prominent standard for eGPU connections, alternatives like OCuLink exist. This open standard connects directly to the PCIe bus and offers cost advantages, although it has not gained widespread support among laptop and desktop manufacturers. Handheld consoles do support OCuLink, albeit with certain quirks, such as requiring a system reboot to enable it.

The future of eGPU technology may hinge on the development of new standards that provide more than four lanes of PCIe connectivity. However, given the niche market for external GPUs, substantial funding for such advancements remains unlikely. For now, eGPUs continue to serve a specific audience, providing additional computational power for tasks such as AI and data analysis, albeit with inherent limitations in latency and performance.