barren wadi
System Information
GPU: NVIDIA GeForce RTX 5090
Software Version: 1.9.12
OS: Windows 11 H25
Description of the Issue
The Salad distribution algorithm currently fails to automatically prioritize high-profit container workloads over low-yield bandwidth tasks.
Even when there is high market demand for high-end GPUs like the RTX 5090, the software remains "stuck" in low-earning bandwidth jobs (generating only micro-amounts) instead of pivoting to available container jobs.
The Evidence
Stagnation: My rig was running a bandwidth job with negligible returns for several hours, despite high demand for 5090-specific container jobs.
Manual Trigger: Immediately after a manual restart of the Salad application, the system assigned a container job paying $0.07/hour.
Conclusion: The software is technically capable of running the high-value job, but the automation logic fails to "preempt" or stop a low-value task to make room for a significantly more profitable one.
Expected Behavior
The algorithm should continuously monitor the marketplace. If a high-value container workload becomes available that significantly outperforms the current task's profitability, the system should automatically switch (pivot) to the higher-paying job without requiring a manual restart.
Actual Behavior
The system prioritizes the "continuity" of a low-paying job over "profitability," effectively locking out high-end hardware from lucrative workloads until the user intervenes manually.
Steps to Reproduce
Run Salad with an RTX 5090.
Observe the system picking up a low-yield bandwidth/micro-job.
Wait for high-demand periods (confirmed via market monitoring).
Notice that the system does not switch.
Restart Salad -> High-value job is assigned instantly.
Note: I have log files available upon request to verify these findings. No errors were found in the logs, confirming this is a logic/priority issue rather than a technical crash.
vernal epoch