A breakdown of NVIDIA, AMD and Apple desktop AI machines says software fit matters more than speed. Memory and price differ a lot.
Three big chipmakers now sell desktop computers built to run AI at your desk, and a new breakdown in Entrepreneur says the right pick depends less on speed than on the software your team already uses. The piece is a contributor's opinion, but its numbers help sort out NVIDIA, AMD and Apple.
The article says a local AI box does not replace the cloud for large training jobs. It makes sense for three things: keeping sensitive data off other companies' servers, avoiding per-use fees when you test an AI tool hundreds of times a day, and owning a fixed machine that is not shared or rate limited.
The NVIDIA DGX Spark has 128GB of memory and runs on CUDA, NVIDIA's long-established AI software. Existing code runs without changes. The catch is data speed. At 273 GB/s, it is the slowest of the three at moving model data. Reviewers find it generates text more slowly than Apple's Mac Studio.
AMD's Ryzen AI Max+ 395 chip comes in two kinds of products. Third-party mini PCs, such as the GMKtec EVO-X2 and the Framework Desktop, cost a fraction of AMD's official Ryzen AI Halo Developer Platform. LTT Labs says the extra money on the official kit buys curated software setups, called AI Playbooks, and AMD support, not better hardware.
Speed is the weak spot. LTT Labs found Apple's Mac Studios outperform these AMD machines, which produced text two to three times slower. AMD's software for this work, ROCm, is also more mature on Linux than on Windows. One bright spot: the chip ran OpenAI's gpt-oss 120B model at up to 30 tokens per second, which are small chunks of text, on the 128GB version.
The Mac Studio offers the M5 Max with up to 128GB of memory, or the M5 Ultra with up to 512GB. Data speed is 614 GB/s on the Max and 1.2 TB/s on the Ultra. That is the highest of the three, and it can load larger models. The cost is that code must run through Apple's tools instead of CUDA or ROCm. Common apps such as Ollama and LM Studio work out of the box.
Even 512GB is not enough for the largest open models, the article notes. Choose the ecosystem your work already depends on first, then compare hardware.
Pick the machine that matches the software you already use first, then compare memory, speed and price.
Microsoft reportedly cut some AI usage limits by 90 percent, and Meta's use of Claude Code reportedly halved. Here is what that could mean for your tools.
Read more: Microsoft cuts some AI usage limits 90%; Meta coding tool use halvesApple's new M6 Mac mini costs $900 and handles everyday creative work well, but Engadget says NVIDIA graphics cards are far stronger for gaming.
Read more: Apple's new M6 chip trails NVIDIA graphics cards but costs far lessLawmakers in both parties want Nvidia's CEO to answer questions about a reported tie to an alleged Chinese intelligence asset, and chip rules could shift as a result.
Read more: Lawmakers from both parties want Nvidia's CEO to answer questions
