Generation on the GPU, search on the VDPU

The company's stated concept divides labor inside an AI server into two lanes. Under the "generation on GPU, search on VDPU" formulation, GPUs take charge of the generative stage, while VDPU hardware is dedicated to search workloads. The framing positions the two processor types as complements rather than substitutes, with each handling the task it is assigned in the company's server architecture.

Four-card configuration shown in the US

The performance disclosure centered on a server populated with four cards — a four-card configuration — and was made in the United States, where the results were presented publicly. The verified claim establishes that the performance of this four-card setup was revealed; specific benchmark figures such as throughput or latency were not part of the verified claim and are not reported here.

What the verification found

The claim was examined against eight primary sources, and the review supported the account as stated: that Dinotisia presented the four-card server's performance in the US under its "generation on GPU, search on VDPU" banner. On that basis, the final determination on the claim is True.