Same 128GB but cheaper
Credibility score: 56/100 — Mixed Credibility. Several questionable claims detected. Watch with healthy skepticism.
BSmeter analyzed "Same 128GB but cheaper" and rated it 56/100 for credibility (a BS score of 44/100 — mixed credibility), on 2026-06-20. Its weakest claim — "Mac Mini faster than Spark, but Spark's prompt eval rate is 'ridiculous'. Cherry-picked 🍒" — scored 20/100 and was flagged as cherry-picked. 17 claims were checked against the video transcript. Scores are produced by BSmeter's AI analysis of the transcript, not independent human verification.
Of 17 claims analyzed: 1 scored under 40, 9 between 40 and 69, and 7 at 70 or above.
Claims analyzed
Comparing Thor and Spark RAM, highlighting Thor's lower price point. — No Frame (75/100)
At 0:00
Setting up the core comparison: same RAM, big price difference. Straightforward setup for the video's premise. 💰
Why this score: The speaker is laying out the main question of the video: why buy the more expensive Spark if the Thor has the same amount of RAM and is significantly cheaper? This is a direct, factual comparison to frame the upcoming tests.
Original quote: “The Jetson Thor is Nvidia's new developer kit, and the DJX Spark is for everybody, but what's not often talked about is the Thor has 128 gigs of RAM, just like the Spark. That's a lot of RAM, and it's more than $1,000 cheaper than the Spark. So, why don't we just buy the Thor instead of the Spark?”
Justifying machine selection based on shared memory bandwidth. — No Frame (75/100)
At 0:20
Picking machines based on a specific, shared spec. That's a solid basis for comparison. 🤓
Why this score: The speaker explicitly states the commonality (273 GB/s memory bandwidth) that links the three devices being compared. This provides a clear, technical rationale for their selection, which is good for a benchmark video.
Original quote: “The reason I picked these three is because the M4 Pro, the Thor platform, and the Spark all have 273 GB per second memory bandwidth.”
Highlighting shared memory bandwidth, then dropping the Nvidia sponsorship. Classic switcheroo. 🤡 — Volume Game (45/100)
At 0:30
Talks up a key spec, then casually mentions Nvidia sent the product for testing. The 'by the way' is doing heavy lifting. 💸
Why this score: The speaker first establishes a technical point (shared memory bandwidth) as a key reason for comparison, then immediately follows with a disclosure that Nvidia provided one of the devices for testing. The casual 'by the way' minimizes the impact of the sponsorship, which could influence the perceived objectivity of the tests, even if the speaker states they bought the others. It's a common tactic to present information that might affect perception after a seemingly objective statement.
Original quote: “three is because the M4 Pro, the Thor platform, and the Spark all have 273 GB per second memory bandwidth. And I mentioned before many times that memory bandwidth is really important for machine learning, but that doesn't tell the whole story. By the way, Nvidia sent me this Thor to do some…”
Comparing Thor and Spark with a price difference, setting up a test. Straightforward setup. — No Frame (75/100)
At 0:30
Setting up the premise for the video, comparing two products and their price difference. No tricks here, just the setup. 💰
Why this score: The speaker is introducing the main comparison of the video, highlighting a key difference (price) and a similarity (RAM) between the Nvidia Jetson Thor and DJX Spark. This is a clear, direct setup for the upcoming tests and discussion, not using any manipulative framing.
Original quote: “The Jetson Thor is Nvidia's new developer kit, and the DJX Spark is for everybody, but what's not often talked about is the Thor has 128 gigs of RAM, just like the Spark. That's a lot of RAM, and it's more than $1,000 cheaper than the Spark. So, why don't we just buy the Thor instead of the Spark?…”
Stating power usage for Mac Mini M4 Pro (8W), Thor (31W), and Spark (44W). Specific data points. — No Frame (75/100)
At 2:17
Giving specific power consumption numbers for each device. These are testable metrics, not just vibes. 📊
Why this score: The speaker presents specific power usage figures for each of the three devices (Mac Mini M4 Pro at 8 watts, Thor at 31 watts, Spark at 44 watts). These are quantifiable measurements that can be verified through testing, and the speaker explicitly states this is 'just to kick things off' for further benchmarking. This is a direct presentation of data, not a framing trick.
Original quote: “using about 8 watts. The Thor is using about 31 watts and the Spark is using about 44 watts. That's just to kick things off. I'm going to run a couple LM here and then do some benchmarking so you get an idea of how these perform. There's a brand new model called Minist. Haven't tested that on the…”
Presenting initial power usage stats for Mac Mini, Thor, and Spark. Clear data. ✅ — No Frame (75/100)
At 2:18
Providing specific power consumption numbers for each device before starting benchmarks. This is just setting the stage with data. 📊
Why this score: The speaker presents specific power usage figures for each device (Mac Mini M4 Pro at 8 watts, Thor at 31 watts, Spark at 44 watts) and clearly states this is 'just to kick things off' before further benchmarking. This is a straightforward presentation of initial data, setting a baseline for the upcoming tests without any apparent rhetorical manipulation.
Original quote: “using about 8 watts. The Thor is using about 31 watts and the Spark is using about 44 watts. That's just to kick things off. I'm going to run a couple LM here and then do some benchmarking so you get an idea of how these perform. There's a brand new model called Minist. Haven't tested that on the…”
Speaker infers Thor's GPU loading/unloading from a 'bouncing' graph, suggesting inefficiency. — Confidence Mismatch (45/100)
At 2:30
Saw a wiggly line and immediately jumped to 'inefficient loading/unloading.' That's a big leap from a graph, chief. 📈🤷♂️
Why this score: The speaker observes a 'bouncing' characteristic in the Thor's GPU usage graph and immediately interprets this as 'loading and unloading at a rapid pace,' leading to a conclusion of potential inefficiency compared to the Spark. While the observation of the graph is valid, the leap to a specific internal mechanism ('loading and unloading') and a definitive conclusion about efficiency without further diagnostic data or deeper analysis is a significant assumption. They even admit 'I don't know what's going on on the inside there,' yet still draw a strong conclusion.
Original quote: “while it's happening, take a look at what's going on with uh the Spark here. There is the GPU usage. We're going at about 95% usage. And over here on the Thor about 94 to 96% usage as well. But the characteristic of the of the line is very different on these two. The Spark is very even. So it's…”
Speaker infers Thor's GPU 'bouncing around' means it's 'loading and unloading' and 'may not be as efficient' as Spark. — Confidence Mismatch (45/100)
At 2:30
Went from 'I don't know what's going on on the inside' to a confident diagnosis of 'loading and unloading' and 'less efficient.' That's a leap of faith, not data. 🤸♂️
Why this score: The speaker explicitly states they 'don't know what's going on on the inside' but then proceeds to interpret the visual characteristic of the graph (bouncing around) as definitive evidence of 'loading and unloading' and potential inefficiency. This is an inference based on limited information, presented with a degree of certainty that doesn't fully match the acknowledged lack of internal knowledge. It's a hypothesis, not a confirmed conclusion, despite the confident tone.
Original quote: “But the characteristic of the of the line is very different on these two. The Spark is very even. So it's constantly using 95% of that GPU. But the Thor is bouncing around a little bit, which tells me that just based on the graph, I don't know what's going on on the inside there, but it's telling…”
Declaring Mac Mini the winner and Thor the loser based on token generation. — Confidence Mismatch (45/100)
At 4:30
Calling a 'winner' and 'loser' when the numbers are pretty close. The confidence is doing more work than the data. 🏆🤡
Why this score: While the Mac Mini did have a slightly higher token per second rate (39 vs 34.97 and 30.82), declaring clear 'winners' and 'losers' when the differences are not massive can be an overstatement. It's a bit of a subjective interpretation of relatively close data points.
Original quote: “Uh, well, I guess the Mac Mini won this round and the Thor lost this round with the Spark coming in right in the middle there.”
Comparing Thor and Spark token generation speeds with specific numbers. — No Frame (75/100)
At 4:30
Giving specific numbers for comparison, that's just straight data. No tricks here. 📊
Why this score: The speaker is presenting direct benchmark results with specific numerical values for token generation speed, allowing for clear comparison without rhetorical manipulation.
Original quote: “We have 30.82 82 tokens per second on the Thor and 34.97 tokens per second on the Spark.”
Mac Mini faster than Spark, but Spark's prompt eval rate is 'ridiculous'. Cherry-picked 🍒 — Cherry-Picked (20/100)
At 6:30
Mac Mini is faster on tokens/sec, but the Spark gets a 'ridiculous' prompt eval rate. Conveniently highlights the Spark's win while downplaying its loss. 🙄
Why this score: The speaker presents the Mac Mini as having a higher tokens per second rate (38.95) than the Spark (32.92), but then immediately pivots to praise the Spark's 'ridiculous' prompt eval rate (2,817) without giving the Mac Mini's prompt eval rate for direct comparison. This cherry-picks the metric that makes the Spark look better.
Original quote: “Mech Mini is done. We got 38.95 tokens per second there. 32.92 tokens on the Spark.”
Claims Spark uses almost double the power due to powerful compute and more cores. — Confidence Mismatch (45/100)
At 6:30
Connects 'almost two times higher power' directly to 'more cores' like it's the only factor. A bit too confident on the causation. 🤔
Why this score: While the Spark does show higher power usage (141W vs. 70W for Mac Mini, 90W for Thor), the speaker attributes this directly and solely to 'that really powerful fast compute that's on there with more cores'. This is a confident assertion of a direct causal link without presenting specific evidence or acknowledging other potential contributing factors to power consumption in complex systems. It's a plausible explanation, but stated with more certainty than the presented data alone supports.
Original quote: “Almost two times higher power usage on the Spark. And that's all being used by that compute, that really powerful fast compute that's on there with more cores.”
Mac Mini using 56.87GB out of 64GB, but memory pressure is 'nice and clean'. — Confidence Mismatch (45/100)
At 8:30
Almost maxed out on memory, but 'nice and clean'? That's a bold take on being 90% full. 😬
Why this score: The speaker states that the Mac Mini is using 56.87 GB out of its 64 GB of memory, which is over 88% utilization. Despite this high usage, they confidently declare the memory pressure 'nice and clean' with 'no orange or red'. This feels like a confidence mismatch because such high memory usage typically indicates significant pressure, even if the system isn't yet showing critical warnings. It downplays the near-maxed-out state.
Original quote: “just by a little bit. So, it's using more of the GPU. Memory used on the Mac Mini, 56.87 87 GB out of the 64, but the memory pressure is nice and clean there. No orange or red, so we're good.”
Highlighting Spark's prompt eval speed as 'amazing' despite diminishing returns on token generation. — Volume Game (45/100)
At 10:30
Diminishing returns on one stat, then 'amazing' on another. The volume on 'amazing' is doing overtime 🔊
Why this score: The speaker acknowledges a 'diminishing return' for the Spark on one metric (token generation) but immediately pivots to emphasize its 'amazing' performance on another (prompt eval rate), using strong comparative language ('more than two times faster, almost three times faster'). This uses volume to downplay a negative and amplify a positive, creating a more favorable overall impression.
Original quote: “A little bit of a diminishing returns there on the Spark, even though the prompt eval rate still was amazing on the Spark. More than two times faster, almost three times faster than anything else. So 283 tokens per second on prompt eval which is basically prefill.”
Claims Llama Bench uses more GPU than Olama, then admits no concrete numbers 🤡 — Confidence Mismatch (45/100)
At 12:30
Says 'actually utilizing more GPU' then immediately 'I don't have any concrete numbers.' Just vibes, no data. 🤷♂️
Why this score: The speaker makes a definitive statement about GPU utilization ('actually utilizing more GPU') but then immediately undermines it by admitting they have no 'concrete numbers.' This is a classic confidence mismatch, presenting an observation as a fact without the data to back it up.
Original quote: “Now, take a look at this. On the Mac, Llama Bench is actually utilizing more GPU than when we were running Olama just by a tiny bit. You can kind of visually see it. I don't have any concrete numbers here, but maybe we'll see some when we actually see the final results here. Olama might have just a…”
Comparing token generation speeds, Mac Mini vs. Thor/Spark — No Frame (75/100)
At 14:30
Just straight up numbers and comparisons here, no funny business. Benchmarking as promised. 📊
Why this score: The speaker is presenting specific benchmark numbers for different hardware configurations (Thor, Mac Mini) and comparing their performance in token generation and prompt processing. This is a direct, data-driven comparison without apparent rhetorical tricks.
Original quote: “much the same as what Olama gave us. 4.56 tokens per second for TG128. TG128 just means token generation. Then we got about 283 tokens per second for PP512 which is prompt processing or prefill on the Thor. It's not done yet. So wait for that one. On the Mac Mini, 42 which is higher for prompt…”
Acknowledges not testing variability, then leans on 'theoretically' for performance claims. — Volume Game (45/100)
At 16:30
Says he didn't test it, then immediately pivots to 'theoretically' to still make the point. Classic volume game. 🤷♂️
Why this score: The speaker admits a lack of direct testing for variability but then uses 'theoretically' to assert the Thor's deterministic latency. It's a subtle way to make a claim without the direct evidence just presented, softening the blow of the missing data while still implying the benefit.
Original quote: “I didn't test the variability between different things. I just showed you examples, but theoretically, that's what it's supposed to be good at.”
See the full analysis with timestamps →