MacBook Neo's A18 Pro Chip Hits a Wall in Blender Testing

The MacBook Neo arrived in indigo with a phone charger in the box—the same 20-watt brick that comes with an iPhone. This detail matters because the A18 Pro chip inside is exactly what Adam from Adam Doing Tech calls it: "a phone chip by the end of the day." What happens when you push phone-grade silicon to render complex 3D scenes tells you something about where Apple's hardware strategy succeeds and where it doesn't.

Adam tested the base model MacBook Neo—8GB of unified memory, five-core GPU—against Blender's benchmark scenes. The results split into two categories: what works and what absolutely doesn't.

The GPU Memory Problem

The Junk Shop project, one of Blender's standard benchmark scenes, immediately exposed the A18 Pro's fundamental limitation. When Adam configured Blender to use only the GPU for rendering, the system returned an error: "Out of GPU memory."

He tried again with the machine plugged into power, thinking maybe additional thermal headroom would help. Same error. The scene simply wouldn't render on the GPU alone. "As it seems so far, you cannot render a basic scene using the GPU that is included with the A18 Pro," Adam noted during testing.

This isn't a software bug or a settings issue. The A18 Pro chip shares 8GB of memory between system RAM and GPU VRAM. Complex Blender scenes can demand more dedicated graphics memory than that architecture provides. Desktop GPUs in Adam's comparison—an RTX 3090 Ti with 24GB of VRAM, even a mobile RTX 3070 Ti with 8GB of dedicated memory—have resources the A18 Pro fundamentally can't match.

Switching to CPU-only rendering worked. The Junk Shop scene completed in about six minutes after the initial kernel build. The machine stayed cool enough that Adam checked by touch: "Nope, it's not warm compared to Windows machine. On Windows machine, by now it will be flex."

The Unexpected Battery Advantage

What happened next challenges conventional wisdom about laptop performance. Adam unplugged the MacBook Neo and ran the same CPU-only render on battery power.

Plugged in: 6 minutes. On battery: 5 minutes 59 seconds.

He ran it again to confirm. Battery power consistently matched or slightly beat wall power performance. This contradicts the usual expectation that laptops throttle on battery to preserve charge. Adam made a point of showing that low power mode was disabled—he'd clearly dealt with that particular comment before.

The explanation probably lies in Apple's thermal management strategy. The A18 Pro is designed for a phone chassis with no active cooling. In a laptop with marginally better thermals, the chip may actually run into power delivery limits when plugged in that don't exist on battery. Or the power adapter itself—that 20-watt phone charger—can't supply enough juice to prevent throttling under sustained load. Either way, the machine performed identically whether tethered or mobile.

Hybrid Rendering: Finding the Sweet Spot

Adam then enabled both CPU and GPU rendering simultaneously. This configuration—using Apple's Metal ray tracing to coordinate between compute resources—completed the Junk Shop scene in 5 minutes 23 seconds. That's notably faster than CPU-only rendering and works around the GPU memory limitation by offloading overflow to the CPU.

The Fox Splash scene, less demanding than Junk Shop, told a different story. The GPU handled it alone without memory errors, completing renders in 24-25 seconds using the Eevee engine. CPU-only rendering took about the same time, though Adam noted this was within margin of error.

For viewport rendering—the real-time preview mode where you're actually working—GPU acceleration made a decisive difference. Adam tested the Classroom scene, switching between CPU and GPU rendering in the preview window. "If you want to preview everything, 100% use the GPU because with the CPU, it's definitely slower," he concluded. The viewport remained responsive in both modes, but the GPU rendered samples noticeably faster.

What This Means for Actual Work

The testing reveals a machine with clear boundaries. If your Blender projects fit within 8GB of shared memory, the MacBook Neo handles viewport work smoothly and renders scenes in reasonable timeframes. Step beyond that memory limit with complex projects, and the GPU becomes unusable for final rendering—you'll fall back to slower CPU-only mode or need to simplify your scene.

Adam's benchmark comparison provides context. The M1 Max with 64GB rendered Junk Shop in 23 seconds. The M4 with 16GB took 20 seconds. The M5 with 16GB: 16 seconds. The A18 Pro, limited to CPU rendering, took 6 minutes. The performance gap between phone silicon and Apple's pro chips is exactly what you'd expect.

But here's what's interesting: the machine actually works. It doesn't thermal throttle into uselessness. It maintains performance on battery. For less intensive scenes, the GPU provides smooth viewport rendering. The limitations are real and measurable, but they're not catastrophic.

Adam plans follow-up testing with video editing and After Effects, promising to "edit a full video on this machine and not just like have footage and then scrub through it." That distinction matters. Synthetic benchmarks and controlled tests tell part of the story. The other part emerges when you try to actually work on the machine for extended periods.

The MacBook Neo with an A18 Pro chip costs less than Apple's pro models. That price difference exists for reasons that real-world testing makes visible. Whether those limitations matter depends entirely on what you're rendering.

Rachel "Rach" Kovacs is Buzzrag's cybersecurity and privacy correspondent.