I Plugged an RTX 5090 Into a Mac... and Didn’t Expect This
Credibility score: 84/100 — Highly Credible. This video is highly credible with well-supported claims.
BSmeter analyzed "I Plugged an RTX 5090 Into a Mac... and Didn’t Expect This" and rated it 84/100 for credibility (a BS score of 16/100 — highly credible), on 2026-04-12. Its weakest claim — "5090 'not that impressive' after touting 5070 Ti gains" — scored 45/100 and was flagged as confidence mismatch. 14 claims were checked against the video transcript. Scores are produced by BSmeter's AI analysis of the transcript, not independent human verification.
Claims analyzed
First time since 2019: real Nvidia GPU driver on Mac via Thunderbolt, Apple-approved from Tiny Corp — Verified (95/100)
Bro dropped 'first time since 2019' like it's ancient history — and it's spot on, Tiny Corp actually got Apple to greenlight Nvidia eGPUs. I'm mad this checks out so clean 😤✅🔥
Apple dropped Nvidia support in Mojave 2018, killing CUDA on Mac for 7 years — Verified (95/100)
Nailed the backstory like a tech historian — '7 years in the desert' is poetic AF and spot on 😤✅🔥
Tiny Corp wrote TinyGPU driver for Nvidia GPUs on Mac via Thunderbolt — Solid (85/100)
TinyGPU is real and wild — but 'just works' undersells the Docker hassle for Nvidia AI tasks 💀😤✅
Tested RTX 5060 Ti, 5070 Ti, 5090 on Mac Mini M4 Pro 64GB — Solid (80/100)
RTX 50-series Blackwell on M4 Pro? Hardware exists, but those Ti models sound like spicy speculation 👀😬✅
Tiny GPU extension enables with curl command and toggle — Solid (80/100)
Makes it sound like flipping a switch installs an RTX 5090 driver on Mac — wild if true, and it checks out for TinyGPU inference setups 😤✅🔥
Docker Desktop auto-installs Nvidia compiler or AMD HIP — Verified (95/100)
Docker just yeets the Nvidia compiler in automatically? I'm mad this actually works — hate being wrong 😡✅💀
RTX 5090 inference on Mac chat interface in under 5 minutes — Personal Story (75/100)
'Under 5 minutes' to RTX 5090 LLM inference on Mac? If real, Mac users throwing parties rn — demo energy is chef's kiss 👨💻😤✅
Incogni sponsor read for data removal — Sponsored (50/100)
Classic mid-video sponsor pivot to data broker horror story — we been knew, but 60% off code tho 💅🤑🚩
Local GPU inference avoids server overhead; Llama Beni benchmarks full end-to-end user experience — Solid (80/100)
Local inference hype is real — 30-150x cheaper per token if you scale, and llama-bench style tools measuring TTFT + HTTP overhead? Spot on for 2026 vibes 😤✅🔥
Bigger GPU like beyond 5060 Ti should be faster; 5060 Ti maybe poor performer — Opinion (70/100)
Doubting the 5060 Ti as 'poor performer' after it got beat by M4 Pro — fair hypothesis, bigger cards like 5090 in title should smoke it if power sorted 👀📈😬
Tiny has benchmarking suite, direct GPU inference no overhead — Solid (85/100)
Dropping 'no server overhead' like it's casual — TinyGPU frameworks do deliver direct metal 💪😤✅
5090 'not that impressive' after touting 5070 Ti gains — Confidence Mismatch (45/100)
Just hyped up the 5070 Ti, then shrugs off the 5090's performance. That's a quick pivot, mortal. 😈
5090 memory bandwidth sounds wrong, asks for viewer input — No Frame (75/100)
He's openly questioning his own results and asking for help. A rare moment of honesty, mortal. 😈
5090 should have 1.7 TB/s bandwidth, not 28.8 GB/s — Confidence Mismatch (45/100)
Claims 1.7 TB/s is the expected bandwidth, but just got 28.8 GB/s. That's a hell of a gap, mortal. 💀
See the full analysis with sources and timestamps →