v1.9.222026-11-13Climbing the Folding@Home leaderboard has become a pet hobby of mine since a pal turned me back on to distributed science projects in the Spring of 2025. Right about that time, and unrelated, my father passed away and left me a small amount of money. I thought to myself, what would Dad totally not understand? Why building a multi-GPU rig to climb the F@H leaderboard, of course! And so I got right to it.
This site gives some details of my folders, their roadmaps if there is one, and how they are performing. There's also some info about the "AI"-rolled set of python scripts that have been "developed" to collect these stats and automate some aspects of the folding operations. Feel free to drop me a line with any questions.
itztli: v1.5. Flagship Node. AMD Ryzen 9 9950X3D + 1x NVIDIA RTX 4090. Folder, primary telemetry aggregator and my CachyOS desktop.foldimate: v1.0. Partner's CachyOS Gaming Rig & Overnight Folder. AMD Ryzen 9 7950X + 64GB DDR5 + 1x ASUS NVIDIA RTX 4070 Ti (12GB). Side-by-side Valheim & Elden Ring SCO co-op and Lonely Mountains: Downhill daily ride crusher by day; RTC-woken 195W efficiency-capped folder by night (00:01–06:15).origami: v2. High-Density Server Node. AMD EPYC 7543 (32C/64T) on ASRock Rack ROMED8-2T with 256GB 8-channel DDR4-3200 ECC RAM in a Phanteks Enthoo Pro 2 Server Edition chassis. 2x NVIDIA RTX PRO 4000 Blackwell (24GB) + 1x Zotac GeForce RTX 5070 (12GB). 128 PCIe 4.0 lanes for high-density multi-GPU folding.smallpieces: v3. 5-GPU All-Blackwell Node (Completed Oct 29, 2026; Alphacool ES Guardian PWM Hubs Nov 13, 2026). AMD Ryzen 9 9900X (12C/24T, 65W Eco Mode) + ASUS TUF Gaming X670E-PLUS WiFi 6E + 32GB DDR5-6000 + ADT-Link R34A x4/x4/x4/x4 bifurcation (4x PCIe Gen 5.0 x4 risers) + 1x PCIe Gen 4.0 x4 direct on motherboard + 2x Alphacool ES Guardian 4-Port Powerfan Hubs (CHA_FAN2 Intake & CHA_FAN1 Exhaust) in the GPNE 8-fan wind-tunnel chassis. GPUs: 1x RTX 5070 (12GB) + 4x RTX 5060 Ti (8GB/16GB).scraps: v2. 5-GPU Ada + Ampere Octominer X12 Ultra Node (Expanded Nov 1, 2026). Intel Core i7-7700 (4C/8T @ 3.00 GHz thermal sweet spot with FahCore thread affinity isolation) + 32GB DDR3L RAM + 4x HPE 750W Platinum PSUs. Current GPUs: 1x ZOTAC RTX 4070 Blower (12GB, Slot 1 PCIe 3.0 x16) + 1x ZOTAC RTX 3070 (8GB, Slot 2 PCIe 3.0 x1) + 3x RTX 3060 Ti (8GB, Slots 4, 7, & 10).Comprehensive fleet hardware timeline (Feb 2026 – Present), EOC monthly/daily production stats, cross-referenced average PPD speeds per hardware era, and detailed node specifications.
View Inventory & Production Timeline →Archive of daily 24-hour fleet performance audits (08:00 PDT to 08:00 PDT), peak PPD speed benchmarks, 24-hour Work Unit completion counts, health alerts, and full 24-hour PPD timeline graphs.
View Morning Reports Gallery →588,289-sample neumon study of watts, heat (BTU/hr), PPD/W across 13 eras, CPU frequency/Vcore tuning (Ryzen 9 9900X 65W Eco & i7-7700 2.40–3.00 GHz), PCIe scaling on smallpieces & origami, datacloset HVAC/Audimute balance, and controlled RAM speed A/B benchmarks.
Standardized /opt/docker-configs/fah-compose.yaml specification, OPENMM_CPU_THREADS=4 host RAM relief, NVIDIA driver reservations, fah-start / fah-stop timer lifecycle, and graceful GPU hardware swap routine across all hosts.
smallpieces v3 (Ryzen 9 9900X All-Blackwell) & v2 HistoryCompleted Oct 29, 2026 5-GPU All-Blackwell Ryzen 9 9900X (65W Eco Mode, Thermalright AXP90-X53 Full Copper) + PCIe Gen 5.0/4.0 x4 architecture (~45.8M–46.8M PPD @ 795W cap / 754W draw), 6→1 GPU pre-cutover drain benchmark, and historical v2 Octominer slot mappings.
Viewsmallpieces v3 Architecture & History →
origami v2 System Architecture & RoadmapHardware summary (AMD EPYC 7543 32-Core + Noctua NH-U14S TR4-SP3 + 256GB 8-Channel DDR4-3200 ECC RAM + ASRock Rack ROMED8-2T + Phanteks Enthoo Pro 2 Server Edition with 13x Super Flower MEGACOOL fans), physical-to-logical PCIe slot mappings (2x RTX PRO 4000 Blackwell + 1x RTX 5070), and multi-GPU architecture.
Vieworigami Architecture & Roadmap →
foldimate v1 Hardware & Overnight Folding ArchitecturePartner's Ryzen 9 7950X (be quiet! Silent Loop 3 240) + RTX 4070 Ti CachyOS gaming rig (Valheim, Elden Ring SCO, Lonely Mountains: Downhill), RTC S3 wake automation, KDE sleep inhibition, and shift-scoped 195W/285W power profile.
Viewfoldimate Architecture →
scraps v2 (Octominer X12 Ultra) GPU Ordering Map & History5-GPU Ada + Ampere Octominer X12 Ultra array (1x RTX 4070 Blower + 1x RTX 3070 @ 52°C + 3x RTX 3060 Ti, ~20.6M–21.5M PPD @ 725W cap / 674.5W draw), GPNE-to-OX12U slot lineage, and MSR 0x1a0 unlock + 3.00 GHz FahCore thread affinity tuning.
scraps v2 GPU Map & Details →
Showcase of our 6 biggest fleet engineering wins—including our controlled 968-WU multi-GPU power-cap A/B experiment (OpenMM memory-bandwidth Amdahl's Law vs single-GPU gaming rules), doubling scraps PPD, and a 7-tip Field Guide.
| Component Name | Current Version | Operational Role & System Description |
|---|---|---|
| Global Suite | v1.9.22 |
Parent Orchestrator, Fleet Deployment Helpers & SSH Propagators |
| fahnalysis | v8.10.10 |
High-Precision F@H Telemetry, PRCG Footprint Tracking & Speed Aggregators |
| neumon | v6.16.3 |
Sub-Second TUI Terminal Dashboard & Fleet Overtemperature Logger |
| baseliner | v3.7.1 |
Kernel Bottleneck Diagnostic, Real-Time FIFO Monitor & Rolling Sample Retention |
| heatpanic | v2.1.3 |
Multi-Node Overtemperature Shutdown Safety Daemon |
| octoctl | v2.2.15 |
Octominer Motherboard I2C Watchdog Fan Regulator & AC Failsafe |
| fah-runaway | v1.2.8 |
F@H Infinite Loop / Bad Work Unit Watchdog & Preemption Daemon |
| nvpowerctl | v1.3.4 |
Multi-GPU Power Regulator, CPU Governor, MSR Turbo Enforcer & FahCore Thread Affinity Pinning |
| gpufanpin | v1.4.0 |
Per-GPU Positional NVML Fan Floor, Declarative lm-sensors Labels & Motherboard hwmon Chassis PWM Curve Regulator |