Why you possibly can’t simply construct your personal RTX Spark PC
I’ve been placing the staggeringly costly $7,000 Asus ProArt P16 laptop computer by way of some uncommon exams for my AI-focused RTX Spark evaluate, together with operating a sprawling real-world analysis workload by way of OpenAI’s gigantic 120-billion-parameter gpt-oss mannequin. Fairly than asking an AI to jot down a poem about Jensen Huang or make me an image of a cat on a bike, I needed to see whether or not this specialised {hardware} may deal with the kind of stupidly bold native AI work I’d usually reserve for cloud fashions.
It may. Extra curiously, my monster desktop can too. Simply a lot, a lot slower.
For one frozen comparability, each programs chewed by way of the very same workload, producing 472,084 tokens alongside the best way. The ProArt P16 RTX Spark completed in 33 minutes and 13 seconds. My desktop wanted 52 minutes and 51 seconds.
That’s a 19-minute, 38-second ass-kicking delivered to a totally kitted, fire-breathing high-end desktop packing an RTX 5090.
The explanation will get to the guts of what makes these new high-memory AI programs so attention-grabbing — and why you possibly can’t merely construct your personal equal with typical fanatic PC components.
128GB isn’t all the time 128GB
The ProArt P16 RTX Spark packs 128GB of unified LPDDR5X reminiscence, with roughly 96GB obtainable to the GPU throughout my testing. My desktop pairs its RTX 5090’s screaming-fast 32GB of GDDR7 reminiscence with 96GB of typical system RAM.
