Signal65

@Signal_65

We are here to ensure our partners become the signal of innovation in the noise of the technology markets.

signal65.com

Joined December 2023

Replies Reposts Media Articles

Pinned
Signal65
@Signal_65
Sep 18
Two platforms can hold the same number of agents and still give those agents very different response speed. Headcount is the number an enterprise sizes a node by, and the per-agent decode rate is what its users feel, which is why @Signal_65 PINNACLE publishes the whole
4
Signal65
@Signal_65
Sep 19
Enterprises that want to utilize the largest open-weight models in an agentic fleet have a narrower choice of nodes and infrastructure that can support that level of quality. Kimi-K3 at 2.78T parameters is the largest model in the @Signal_65 PINNACLE GPU results, served from its
Signal65
@Signal_65
Sep 18
Enterprises running agents on AMD Instinct today, or sizing a new Instinct fleet, now have an independent number for what the MI355X buys per node. @Signal_65 PINNACLE measured both generations on the same fully loaded 8-GPU node, the same agentic load and the same service level,
Signal65
@Signal_65
Sep 15
Local AI on @Windows is not confined to one class of machine. The same platform, containment, and identity model now runs from a mainstream laptop up to a deskside AI supercomputer. From our report: ➡️ Mainstream, the Copilot+ PC install base at roughly $800 to $2,500 ➡️ High
Signal65
@Signal_65
Sep 15
🚨🚨NVIDIA Vera CPU Agentic Performance🚨🚨 If you are planning capacity for agents (and aren't we all??), the CPU decides how many sessions a rack can keep busy, and the GPU spends much of its day waiting on it. Over the past few weeks Signal65 worked with @nvidia to measure
2