Page Overview: Complete operational chronicle, physical slot mapping, Docker configuration, and upgrade roadmap for server
smallpieces.
Server smallpieces originated as a budget multi-GPU prebuilt mining rig frame designed to process Folding@Home Work Units across cost-effective mid-tier graphics cards. Over time, it underwent a complete transformation from a memory-constrained Haswell system (smallpieces v1) to a high-density, 12-slot custom chassis built on Kaby Lake hardware (smallpieces v2).
smallpieces v1 Era (Feb 2026 – Jun 29, 2026) |
➔ | smallpieces v2 Era (Jun 30, 2026 – Present) |
|---|---|---|
| • GPNE Mining Rig Frame Case | • Octominer X12 Ultra 12-Slot Chassis | |
| • AX-B85M-ETH Motherboard | • B250 Custom Board | |
| • Intel Core i7-4790 (Upgraded from Celeron) | • Intel Core i7-7700 (Kaby Lake 4C/8T) | |
| • 8GB DDR3 SODIMM (Upgraded from 4GB) | • 32GB DDR3L SODIMM (2x 16GB, Upgraded) | |
| • Noctua NH-L9x65 Cooler | • Thermalright AXP90-X36 Cooler | |
| • 2400W Mining Power Supply | • 4x Quad Octominer HPE 750W PSUs | |
| • Shared 110V / 15A Circuit | • Dedicated 240V / 30A Electrical Circuit | |
| • Mixed GPUs (GTX 1070 / 2060 / 3060 Ti) | • 1x RTX 5070 + 1x RTX 4070 + 4x RTX 5060 Ti |
smallpieces v1 Era (Feb 2026 – June 2026)zswap Hack/opt to it.zswap Compression Hack:
zswap in-RAM compressed swap allocation.500fc1e1.., bb7dfb34..).38366633..).neumon and fahnalysis against AMD ROCm / OpenCL GPU architectures.smallpieces v2 Migration & Octominer Discovery (June 27 – June 30, 2026)| Component Category | Hardware Specification & Details |
|---|---|
| Chassis | Octominer X12 Ultra Custom 12-Slot Server Rig |
| Electrical Circuit | Dedicated 240V / 30A AC Circuit |
| Processor (CPU) | Intel Core i7-7700 (Kaby Lake 4C/8T, 3.6 GHz) |
| CPU Cooler | Noctua NH-L9i Low-Profile Cooler |
| System Memory (RAM) | 32GB (2x 16GB) DDR3L-1600MHz PC3-12800 SODIMM |
| Power Delivery | 4x (Quad) Octominer HPE 750W 200V-277V Platinum Plus PSUs |
| Cables & Power | C13/C14 14AWG 15A/250V Power Extension Cords |
The physical slot designations are silk-screened on the Octominer X12 Ultra motherboard.
Slot 1 is the CPU-adjacent physical/electrical PCIe Gen3 x16 slot.Slots 2 through 12 are physical x16 slots that are electrically PCIe Gen3 x1.| Motherboard Slot | GPU Model | PCI Bus ID | NVIDIA Index | GPU UUID | Target Container | Port | Notes |
|---|---|---|---|---|---|---|---|
| Slot 1 (x16) | MSI VENTUS RTX 5070 OC Dual | 0000:01:00.0 | 0 | GPU-d87d48a0-49cf-c9ec-f567-9f46ad84851e | foldingathome | 7396 | Active (12GB GDDR7, CPU-tied slot) |
| Slot 2 (x1) | Empty | — | — | — | — | — | Empty |
| Slot 3 (x1) | Empty | — | — | — | — | — | Empty |
| Slot 4 (x1) | MSI RTX 5060 Ti 8G SHADOW #2 | 0000:07:00.0 | 5 | GPU-8854baa9-8a40-a186-cef9-38b66e867190 | foldingathome | 7396 | Active |
| Slot 5 (x1) | Empty | — | — | — | — | — | Empty |
| Slot 6 (x1) | Zotac Blower RTX 4070 | 0000:04:00.0 | 4 | GPU-d65f3411-caf4-41a8-0248-0dcf200dee35 | foldingathome | 7396 | Active |
| Slot 7 (x1) | Empty | — | — | — | — | — | Empty |
| Slot 8 (x1) | MSI RTX 5060 Ti 8G VENTUS | 0000:06:00.0 | 3 | GPU-56d10701-aced-b504-06b0-0d0fc856686d | foldingathome | 7396 | Active |
| Slot 9 (x1) | Empty | — | — | — | — | — | Empty |
| Slot 10 (x1) | MSI RTX 5060 Ti 8G SHADOW #1 | 0000:03:00.0 | 2 | GPU-a83e55f9-8d3c-4770-8faf-cc15567610cd | foldingathome | 7396 | Active |
| Slot 11 (x1) | Empty | — | — | — | — | — | Empty |
| Slot 12 (x1) | PNY RTX 5060 Ti 16GB VERTO | 0000:02:00.0 | 1 | GPU-2e4a1d96-f5b4-0b5a-c5e2-91c467224a00 | foldingathome | 7396 | Active (16GB Dual Fan) |
The smallpieces host runs a consolidated single-service Docker Compose deployment located at /opt/docker-configs/fah-compose.yaml:
services:
foldingathome:
image: lscr.io/linuxserver/foldingathome:latest
pull_policy: always
container_name: foldingathome
restart: no
stop_grace_period: 90s
ports:
- "7396:7396"
volumes:
- /opt/docker-configs/foldingathome/config:/config
- /etc/localtime:/etc/localtime:ro
environment:
- TZ=America/Los_Angeles
- PUID=2005
- PGID=2000
- ACCOUNT_TOKEN=
- MACHINE_NAME=smallpieces
- NVIDIA_VISIBLE_DEVICES=all
- NVIDIA_DRIVER_CAPABILITIES=all
- CUDA_DEVICE_ORDER=PCI_BUS_ID
- OpenCL_DEVICE_ORDER=PCI_BUS_ID
deploy:
resources:
reservations:
devices:
- driver: nvidia
capabilities: [gpu, utility, compute] NVIDIA_VISIBLE_DEVICES=all): Rather than hardcoding static device_ids: ['0', '1', '...'] arrays (which fail if a GPU drops or shifts indices), the compose spec delegates all GPU pass-through via modern NVIDIA Container Toolkit resource reservations (driver: nvidia, capabilities: [gpu, utility, compute]).CUDA_DEVICE_ORDER=PCI_BUS_ID and OpenCL_DEVICE_ORDER=PCI_BUS_ID ensure that CUDA and OpenCL device numbering within the container maps strictly to physical PCIe bus hierarchy rather than arbitrary driver compute heuristics.stop_grace_period: 90s): Provides a generous 90-second shutdown buffer, allowing active OpenMM simulation steps and checkpoints to flush cleanly before SIGKILL.restart: no) & Host Reboot Trade-Off: Prevents Docker daemon auto-restart loops during hardware faults, permitting host systemd watchdog units (fah-runaway.service) to manage recovery and clean container restarts. The trade-off is that if the host workstation is rebooted during active folding hours, folding does not automatically start on boot and must be manually initiated. To make this convenient, the companion helper scripts /opt/bin/fah-start and /opt/bin/fah-stop were created. Because fleet scheduling relies on systemd oneshot services triggered by timers (fah-start.timer / fah-stop.timer), manually invoking systemctl start directly can leave a oneshot service stuck in an active state where subsequent timer alarms fail to trigger it; fah-start and fah-stop cleanly trigger the startup/shutdown sequence and immediately reset the service back to an inactive/clean state so future automated timer triggers fire reliably.To prevent orphaned slot bindings or phantom "unsupported" devices in F@H v8 client.db when replacing physical GPU cards:
docker compose down), execute the physical card swaps on the Octominer board.docker compose up -d). F@H dynamically re-enumerates the fresh hardware complement without phantom entries. Select all live GPUs in the UI and, save changes, and click the Play icon.smallpieces array to 6 GPUs.
gpu:01:00:00): Overnight telemetry across 12 completed Work Units establishes the steady-state performance ceiling for the RTX 5070 VENTUS in Slot 1 x16 (gpu:01:00:00 | UUID d87d48a0..). On benchmark Project 18246, TPF converged down to 1m43.3s (103.3s), yielding a peak production rate of 9.01M PPD and a 95th percentile rate of 8.86M PPD (+9.3% net PPD surge and -6.2s/frame speedup over the RTX 4070 N-1 baseline in the same slot). High-frequency NVML profiling confirms peak PCIe bus throughput of 1.79 GB/s without link throttling.gpu:04:00:00): RTX 4070 (gpu:04:00:00 in x1 riser | UUID d65f3411..) yields 4.66M PPD (@ 145W cap / 83.2W real load). High-frequency NVML sampler records peak burst transfers of 1.21 GB/s during CUDA frame step exchanges.smallpieces host output currently streams at 27.60M PPD, bringing total live fleet production to ~66.5M PPD (projected ceiling ~69.2M PPD when itztli is not running).616.0W on 240V power with zero thermal throttling or watchdog faults.fah-runaway) were traced to Ada Lovelace (RTX 4070) driver behavior (cuEventSynchronize busy-wait spin-locking), which pegged CPU cores at 93.2%. Blackwell cards (cudaDeviceScheduleYield) yield CPU threads cleanly to kernel idle (intel_idle), dropping host CPU utilization to 71.5% with 28.5% idle headroom under 5 Blackwell GPUs.~11.5% host CPU demand (total 5-GPU host load 65.8%).~13.2% host CPU demand (total 5-GPU host load 68.4%).~15.5% host CPU demand (total 6-GPU host load 93.2%).smallpieces to 7 Blackwell GPUs (1x 5070 Dual + 6x 5060 Ti) is physically possible on the Core i7-7700 host, projecting 36.95M PPD at 41.60 WUs / day (~88-92% CPU load).fah-runaway).