spdup.net

Tech news
English Español Français Deutsch Português Italiano Nederlands

Thunderbolt 5 eGPU Gaming Tested with Razer Core X V2 Delivers Massive Performance Gains

Thunderbolt 5 eGPU Gaming Tested with Razer Core X V2 Delivers Massive Performance Gains

Introduction

The latest leap in external graphics performance arrived with Razer’s Core X V2, the first eGPU enclosure built around Thunderbolt 5. Tech creator ETA Prime put the dock through its paces, pairing it with an RTX 5090 and a high‑end laptop to see how much bandwidth and power improvements translate into real‑world gaming results. Below is a detailed look at the hardware, the test methodology, and the performance outcomes that could reshape the eGPU market.

Razer Core X V2 – Design and Capabilities

Physical Dimensions and Compatibility

  • GPU Length: up to 362.7 mm (14.27 in)
  • Width: 82 mm
  • Height: 185.1 mm (fits four‑slot cards)
  • Supports virtually any modern desktop GPU, including the massive RTX 5090.

Power Delivery

  • External ATX power supply slot – the enclosure does not include a PSU, giving users freedom to choose wattage.
  • 140 W USB‑C Power Delivery on the Thunderbolt 5 port, sufficient to charge high‑end laptops while the eGPU runs.
  • Includes a 120 mm cooling fan for thermal management.

The modular design, combined with Velcro cable management, makes installation straightforward, even with the largest cards.

Thunderbolt 5 – What It Brings to eGPUs

Thunderbolt 5 doubles the raw bandwidth of its predecessor:

  • 120 Gbps unidirectional, 80 Gbps bidirectional.
  • Still limited by the PCIe 4.0 x4 interface used in current eGPU implementations, capping throughput at 64 Gbps.
  • This is still faster than Thunderbolt 4 (40 Gbps) and marginally ahead of Oculink, which tops out around 62‑63 Gbps.

In practice, the extra headroom means less bottleneck when the GPU can already saturate the PCIe 4.0 x4 lane, delivering higher data rates for demanding workloads.

Test Setup

Components

  • Enclosure: Razer Core X V2 (Thunderbolt 5)
  • GPU: Tough Gaming RTX 5090 (four‑slot, 650 W TGP)
  • Power Supply: ASUS ROG Strix 18 W, fully modular, providing ample headroom for the GPU and the 140 W PD requirement.
  • Host Laptop: ASUS ROG Strix 18 equipped with its own RTX 5090 (disabled for eGPU testing).
  • Secondary Test Device: MSI Claw A8 handheld PC with USB 4 ports (no Thunderbolt 5).

Configuration Notes

  • The internal laptop GPU was disabled to isolate the eGPU performance.
  • All benchmarks were run on an external monitor to ensure the GPU handled the full video pipeline.
  • For handheld testing, the RTX 5090 eGPU was connected via USB 4, the highest‑speed interface available on the device.

Performance Benchmarks

CUDA Transfer Speeds

Using a CUDA bandwidth test, the Thunderbolt 5 connection achieved:

  • Host‑to‑Device: 52‑57 Gbps
  • Device‑to‑Host: similar range These numbers surpass typical Thunderbolt 4/USB 4 eGPU performance (≈32‑36 Gbps) and approach the theoretical PCIe 4.0 x4 limit.

Gaming Tests

GameResolution & SettingsLaptop RTX 5090 (internal)Razer Core X V2 RTX 5090 (eGPU)
Cyberpunk 20774K Ultra, no DLSS52 FPS avg90 FPS avg
Horizon Zero Dawn4K Ultra, no DLSS83 FPS avg84 FPS avg
Borderlands 44K Ultra, no DLSS78 FPS avg80 FPS avg

The eGPU outperformed the internal laptop GPU in Cyberpunk 2077 by a large margin, while delivering comparable results in Horizon Zero Dawn and Borderlands 4. Power draw on the eGPU peaked around 500 W, reflecting the higher TGP of the desktop‑class RTX 5090.

Handheld PC Test (USB 4)

Connecting the same eGPU to the MSI Claw A8 via USB 4 yielded:

  • CUDA bandwidth: ~36 Gbps, matching the upper limits of USB 4.
  • Game: 1440p Ultra (no DLSS) – average 70 FPS.
  • Enabling DLSS 3 boosted the average to 276 FPS, demonstrating that AI upscaling can compensate for bandwidth constraints on handhelds.

While USB 4 cannot match Thunderbolt 5’s raw speed, the performance remained on par with high‑end Thunderbolt 4 eGPU setups.

Thunderbolt 5 vs. Thunderbolt 4 and USB 4 – Quick Comparison

  • Bandwidth: 120 Gbps (TB5) vs. 40 Gbps (TB4) vs. ~40 Gbps (USB 4).
  • Real‑world GPU throughput: TB5 achieved ~48 Gbps in CUDA tests, roughly 50 % higher than TB4/USB 4.
  • Power Delivery: 140 W PD on TB5 versus 100 W on most TB4/USB 4 implementations.
  • Compatibility: TB5 maintains backward compatibility with TB4 and USB 4 devices, but only TB5 hosts can unlock the full bandwidth.

Conclusion

The Razer Core X V2 demonstrates that Thunderbolt 5 can finally give eGPUs the bandwidth they need to approach desktop‑class performance. In synthetic bandwidth tests the dock reaches the limits of the PCIe 4.0 x4 interface, and in real‑world gaming it delivers a noticeable boost over internal laptop GPUs, especially in bandwidth‑heavy titles like Cyberpunk 2077.

Even when paired with a USB 4 handheld, the eGPU provides smooth 1440p gameplay, and AI‑driven upscaling (DLSS 3) can bridge any remaining gap. As more Thunderbolt 5 hosts enter the market, the Razer Core X V2 positions itself as a future‑proof solution for creators and gamers who need the flexibility of an external GPU without sacrificing power or performance.

Key takeaways:

  • Thunderbolt 5 offers a substantial bandwidth uplift that translates into higher GPU throughput.
  • The modular design of the Core X V2 lets users pair any ATX PSU with any modern GPU, making it adaptable for both entry‑level and enthusiast builds.
  • Expect a growing ecosystem of Thunderbolt 5 laptops and desktops, which will make high‑end eGPU setups like this increasingly practical.

Stay tuned for upcoming comparisons against Oculink and newer USB 4 docks, as well as tests with more modest GPUs that will further clarify the performance envelope of Thunderbolt 5 eGPUs.

For detailed specifications and purchase links, see the description below the original video.

Watch Original Video