NVIDIA Tried to Tell me How to Benchmark RTX Spark

Wouldn't a M5 Max Macbook Pro be a more apt comparison given they cost similar amounts to the RTX Sparks with the same specs?

0:27 im i the only one who remembered the 50series presentation with numbers without meaning??😊

Would have been much more fair to compare M5 Max 128gb if we consider the spark laptop being tested were around 7000 dollars…

8:30 I love how this graph just looks like a massive middle finger and appropriately for AI, it has 6 fingers 😂

So Jensen made a laptop for Jensen.

I don't know about the other laptops but it really doesn't seem fair to us a Macbook M5 Pro with only 24 gb as comparison, it seems like a M5 Max competitor with that price.

Spark would've been a compelling option if it were 4000$ cheaper, just like the rest of the nVidia product line

These are shareholder devices. There is no target audience, but it looks great to shareholders because AI

Classic example of: "I ship what I make best to corporations, let's see how can we make profit out of rest of the junks to average consumer." Nvidia is very consistent on this bar.

"We sold all 5090s to data centers for 10k a pop. Now let's sell the 5070 to consumers at 5090 prices." Nvidia - we are passionate... about making money.

Why not comparing with M5 Max with 40 cores? That is a direct competitor to AMD 395+.

at 300GB/s memory bandwidth, you are gonna get like 7 tok/s on most large models. so unless you are running inference on a MoE model, you are much better off buying a 5070 and running an open source project like Strata to fit a large model on it.

Funny thing this chip was my last job at Mediatek before I left 3 years ago