Ran the ~260GB EM-only bench on the x99 box, mostly cause I was curious if I even could with 256GB RAM and the core count I have:
| time taken | hardware | operating system | notes |
|---|---|---|---|
| 1h 1min 6s | Xeon 2630L (10c/20t 65W TDP), 4x64GB DDR4-2133 ECC LRDIMMs | Win10 Pro | CPU steady state is 2.0 Ghz at the 45W power draw noted in the earlier results. This run only needed ~211 GB physical and ~223 GB virtual memory per the output window. |
| 1h 3m 5s | Xeon 2630L (10c/20t 65W TDP), 4x64GB DDR4-2133 ECC LRDIMMs | Pop!_OS 22.04 | This run needed ~233 GB physical and ~245 GB virtual memory per the output window. |
And ran the 60GB CFD-EM one on the hardware that still could (i.e. no sense swapping the ram back into the 7200U 
). Both solvers just in case something wacky popped out:
| time taken | hardware | operating system | notes |
|---|---|---|---|
| 21h 49m 47s | Thinkpad T480 - i5-8350U (4c/8t), 64GB DDR4-2400 | Pop!_OS 22.04 | PARDISO run. Had Vitals set up this time, CPU floats between 2.3-3.2 Ghz during the run - rather than leveling out at the 1.7 base clock as previously thought. |
| 20h 48m 26s | Thinkpad T480 - i5-8350U (4c/8t), 64GB DDR4-2400 | Pop!_OS 22.04 | FGMRES run |
| 16h 31m 50s | Lenovo Legion 5 - Ryzen 5600H (6c/12t), 64GB DDR4-3200 | Win11 Home | PARDISO run. High Perf. power plan. |
| 16h 18m 53s | Lenovo Legion 5 - Ryzen 5600H (6c/12t), 64GB DDR4-3200 | Win11 Home | FGMRES run. High Perf. power plan. |
| 11h 27m 40s | Xeon 2630L (10c/20t 65W TDP), 4x64GB DDR4-2133 ECC LRDIMMs | Pop!_OS 22.04 | PARDISO run |
| 12h 1m 1s | Xeon 2630L (10c/20t 65W TDP), 4x64GB DDR4-2133 ECC LRDIMMs | Pop!_OS 22.04 | FGMRES run |
| 13h 21m 2s | Xeon 2630L (10c/20t 65W TDP), 4x64GB DDR4-2133 ECC LRDIMMs | Win10 Pro | PARDISO run |
| 13h 27m 49s | Xeon 2630L (10c/20t 65W TDP), 4x64GB DDR4-2133 ECC LRDIMMs | Win10 Pro | FGMRES run |
| 7h 2m 8s | Ryzen 7800X3D (8c/16t), 64GB DDR5-6000 | Pop!_OS 22.04 | PARDISO run. Did not require -3drend sw for this one. |
| 7h 0m 18s | Ryzen 7800X3D (8c/16t), 64GB DDR5-6000 | Pop!_OS 22.04 | FGMRES run. Did not require -3drend sw for this one. |
I’m not sure what to make of the CFD-EM results, honestly. The Xeon and the 5600H seem to have spread out more on this one than the CFD-only, and in the other direction. Additionally, which solver is faster seems to flip flop, too. But likewise, for this particular workload the 7800X3D is getting smoked by the M1 Max that was posted upthread. 