RTX 3090

Overview

RTX 3090 appears here mostly as a workstation or server GPU used in multi-GPU compute systems, often for CryoSPARC, Relion, deep learning, and other sustained high-load workloads. Tickets focus less on simple setup and more on GPU disappearance, bus errors, thermal or power sensitivity in dense chassis, fan wear, and the lifecycle complications of an end-of-life consumer card being used in professional systems.

Known Issues

• GPU Hardware Failure (~24 tickets) - cards falling off the bus, failing to initialize, disappearing from nvidia-smi, or crashing under load. Examples: #10302, #12942, #14702, #30015, #32956.
• CPU Cooler Failure (~2 tickets) - some multi-GPU instability investigations later involved broader chassis thermal issues rather than the GPU alone. Examples: #20230, #26585.
• System Boot Failure (~7 tickets) - no-POST or no-power symptoms initially blamed on 3090s but later traced to power, motherboard, or full-system faults. Examples: #10302, #11973, #31105, #33686, #42446.
• fan failure (~3 tickets) - noisy or failing 3090 fans, especially on older Zotac cards, with replacement support complicated by age and variant matching. Examples: #31442, #37080, #40658.
• power and thermal instability (~8 tickets) - dense 3090 systems crashing, throttling, or powering off under multi-GPU load due to airflow, paste, padding, power-delivery, or PCIe power-management issues. Examples: #20230, #26199, #26585, #27547, #29976.

Common Questions

• Why does nvidia-smi lose one RTX 3090 or show “Unable to determine the device handle”? Support usually tried slot swaps, single-GPU testing, PCI-bus mapping, and stress testing to see whether the failure followed the card. Examples: #11045, #14702, #15232, #30015, #34306.
• What does “GPU has fallen off the bus” usually mean on 3090 systems? In these tickets it often pointed to a marginal GPU, heat-sensitive card, airflow problem, or broader PCIe stability issue rather than a simple driver reinstall. Examples: #12942, #17232, #20230, #26199, #27547.
• Can Exxact still replace a failed RTX 3090? Frequently only with limits: many 3090 models were already end-of-life, so support had to work through manufacturer repair, offer credit, or substitute other GPUs when exact replacements were unavailable. Examples: #12942, #15575, #17232, #17557, #34306.
• Why does a replacement 3090 or substitute card not match the original exactly? Several tickets involved different cooler sizes, extra power-connector needs, or substitute A5000-class replacements because the original 3090 SKU was no longer stocked. Examples: #14702, #17232, #17557.
• Could the issue be power or cooling instead of the card itself? Yes. Some multi-3090 systems only failed under simultaneous load and were stabilized through chassis, thermal, or PCIe power-management changes rather than a single bad-GPU swap. Examples: #20230, #26199, #26585, #29976.
• Are fan or cosmetic cooler problems covered? On older 3090s, especially out-of-warranty Zotac variants, fan-noise or cooler issues often became paid quote requests rather than warranty RMAs. Examples: #31442, #37080, #40658.

Related Products

• RTX A5000 - common substitute or comparison point when failed 3090s could no longer be replaced directly (#17232, #17557).
• RTX 4090 - often used as a newer comparison card in troubleshooting or replacement discussions, especially around slot and cabling validation (#34939, #41891).
• Other Zotac RTX 3090 variants - Trinity OC, blower, and ASUS Turbo-style cards created confusion because cooler size, power connectors, and warranty path differed by exact model (#10334, #12942, #14702, #31442).
• 4x GPU Exxact workstations - many 3090 tickets are really about dense multi-GPU platform behavior rather than a single standalone card, so nearby platform pages may matter as much as the GPU page itself (#3001, #26199, #26585).

Referenced by

• Power Distribution Board Failure — issue affecting this product (×3)
• GPU Jobs Crash Node Poweroff — issue affecting this product (×4)
• GPU Hardware Failure — issue affecting this product (×31)
• No Trouble Found RMA — issue affecting this product (×3)
• Overheating — issue affecting this product (×2)
• RMA Workflow — issue affecting this product (×29)
• CryoSPARC Integration — issue affecting this product (×1)
• Software Installation — issue affecting this product (×3)
• Power Supply Failure — issue affecting this product (×2)
• Shipping Damage — issue affecting this product (×1)