A $168 BC250 board provides an unusually direct route into testing hardware closely related to the PlayStation 5, but the experiment immediately comes with an important qualification: this is not simply a PS5 graphics card removed from a console. The board uses the same underlying APU according to the presentation, yet arrives restricted to 24 compute units rather than the PS5’s 36 and lacks the console’s surrounding hardware and software environment. Establishing those differences early gives the experiment a useful foundation and prevents the comparison from being reduced to misleading benchmark numbers.
Getting the BC250 operational becomes almost as interesting as benchmarking it. Storage and power are straightforward, while cooling requires a homemade solution that eventually involves 3D printing a fan shroud. From there, custom BIOS flashing, operating-system changes, configuration work and a failed SSD turn the project into a genuine hardware-modification exercise. The self-deprecating commentary keeps these complications entertaining, although the rapid summary of several technical steps means viewers interested in reproducing the setup receive more of an overview than a detailed guide.
Black Myth: Wukong provides the first meaningful gaming test, with the stock 24-compute-unit configuration averaging roughly 45–47 FPS at the described settings. Unlocking 36 compute units raises that figure to 55 FPS, while enabling all 40 reaches only 56 FPS. That limited scaling is one of the experiment’s more useful results because the presenter does not pretend that additional compute units automatically translate into proportional gaming performance. Possible clock, voltage and CPU limitations are offered as explanations rather than definitive conclusions, appropriately reflecting the uncertainty in the test.
Shadow of the Tomb Raider exposes that bottleneck more clearly. The 24-CU configuration averages 83 FPS at 1080p using the highest preset, while 36 and 40 compute units produce only about 85 FPS. Reported GPU-bound measurements support the argument that the processor is restricting further gains in this workload, making the test more informative than simply celebrating an unlocked frame rate. The comparison with the PS5’s 60 FPS mode is interesting, but it should not be interpreted as a clean measure of unused console performance because the BC250 is running under substantially different software and hardware conditions.
The 3DMark Steel Nomad Light results finally give the unlocked graphics hardware room to scale more dramatically. Scores rise from 6,633 at 24 compute units to 8,674 at 36 and 9,248 at 40, demonstrating a much clearer benefit when the earlier CPU limitation is reduced. The claimed estimate that a real PS5 would score somewhere around 8,500 to 9,000 cannot be directly verified through the experiment because, as the presenter acknowledges, the console itself cannot run the benchmark. Consequently, the 36-CU result is an intriguing approximation rather than proof that the modified board exactly reproduces PS5 graphics performance.
That distinction ultimately defines both the experiment’s value and its limitation. Matching compute-unit counts does not recreate the PS5’s clocks, memory behavior, operating environment, game-specific optimization or complete system architecture, and the presenter encounters enough stuttering and memory constraints to demonstrate why those differences matter. Even so, comparing 24, 36 and 40 compute units across actual games and a dedicated GPU benchmark produces a much more revealing investigation than a purely specification-based comparison would have.
Presentation is energetic and unusually candid about failed attempts, improvised solutions and uncertainty, which makes a technically dense project accessible without stripping away all of its complexity. The humor occasionally pulls attention away from the hardware analysis, and the subscription interruption arrives just as the benchmark progression is gaining momentum, but neither problem overwhelms the experiment. Most importantly, the testing generates a coherent conclusion: the graphics portion of this APU can deliver impressive results when given enough compute resources, while CPU limits, memory constraints and the absence of console-specific optimization complicate any attempt to translate those results directly into PS5 performance.
Pros
- The BC250 provides an unusually close and technically interesting basis for investigating PS5-related APU performance.
- Testing 24, 36 and 40 compute units reveals how differently performance scales across GPU-heavy and CPU-limited workloads.
- Black Myth: Wukong, Shadow of the Tomb Raider and 3DMark provide complementary tests rather than relying on a single benchmark.
- Bottlenecks, stuttering, memory limitations and uncertain explanations are acknowledged instead of being hidden behind favorable results.
- The difficult setup process and improvised cooling solution add practical hardware experimentation to the benchmark-focused premise.
Cons
- The BC250 cannot perfectly reproduce a PS5’s complete hardware and software environment, limiting direct performance comparisons.
- The estimated PS5 3DMark score cannot be tested directly, making that portion of the comparison inherently approximate.
- Clock and voltage tuning are not fully controlled before some results are interpreted, leaving uncertainty about the cause of limited scaling.
- Several complicated setup and modification steps are summarized too quickly to make the project a comprehensive reproduction guide.
- The mid-test subscription interruption briefly disrupts an otherwise well-paced benchmark progression.
A clever piece of hardware experimentation turns an inexpensive BC250 into a revealing look at how much performance a PS5-related APU can deliver once its restrictions are altered. The imperfect equivalence to an actual console prevents the benchmarks from becoming definitive PS5 measurements, but transparent limitations, varied testing and genuinely interesting scaling results make the investigation both entertaining and technically worthwhile.
