I’ve been following the progress on the Ampere Altra and Altra Max platforms for a while now, and I’m finally looking to pull the trigger on a dedicated build. My current home lab is a bit of a relic—an old dual-processor enterprise rack that I’ve been using for heavy virtualization and some large-scale compile tasks. Honestly, the noise and the heat are starting to get to me, especially during the summer months.
One specific point I’m curious about is the transition in actual power efficiency. My current machine is backed by a redundant 2200-Watt server power supply setup. While that’s fantastic for handling the massive transient spikes from older, power-hungry x86 chips, it feels like a total waste of electricity for my day-to-day workloads now. I’ve seen the official TDP specs for the Altra line, but I’m much more interested in the “at the wall” idle numbers for a typical enthusiast setup—something like an ASRock Rack motherboard with 128GB of RAM and a couple of NVMe drives.
In my experience, even the best high-efficiency PSUs lose a lot of their effectiveness when they’re running at a tiny fraction of their rated capacity. I’m hoping to downsize my entire infrastructure footprint. Has anyone here successfully moved from a high-power x86 workstation to an Ampere-based system? I’d love to know if the platform allows for scaling back the power overhead significantly, or if the I/O and motherboard components still keep the baseline draw relatively high compared to consumer-grade gear.
Do you think we’re finally at the point where ARM home labs can ditch the industrial-grade power solutions in favor of something more “prosumer” without sacrificing the high core counts we need?
I can draw on better figures when I’m off travel but I have a Q64-30 with 128GB 3200 speed and an M.2 drive that draws ~66 watts at idle. Runs with a modded cheap water cooler. PSU is an old ~400W ATX PSU. Not new enough to be platinum so will be fairly inefficient at the lower end of it’s range.
There may be a slight misunderstanding: “power efficiency” means performing more work with less power comsumtion compared to traditional x86_64 systems. Your view makes sense, but Ampere Altra is not designed for a home-build use case that leaves the server running in idle.
For real-world numbers, this is numbers from my Gigabyte R272-P30 with Q80-30 loading at 0%
root@lucky:~# ipmitool sensor
Watchdog | 0x0 | discrete | 0x0080| na | na | na | na | na | na
SEL | 0x10 | discrete | 0x1080| na | na | na | na | na | na
PSU1_PWR_OUT | 100.000 | Watts | ok | na | na | na | na | na | na
PSU1_PWR_IN | 116.000 | Watts | ok | na | na | na | na | na | na
PSU2_PWR_OUT | na | Watts | na | na | na | na | na | na | na
PSU2_PWR_IN | na | Watts | na | na | na | na | na | na | na
CPU_PWR | 32.000 | Watts | ok | na | na | na | na | 300.000 | na
BPB_FAN_1 | 4800.000 | RPM | ok | na | 1200.000 | 1500.000 | na | na | na
BPB_FAN_2 | 4800.000 | RPM | ok | na | 1200.000 | 1500.000 | na | na | na
BPB_FAN_3 | 4800.000 | RPM | ok | na | 1200.000 | 1500.000 | na | na | na
BPB_FAN_4 | 4800.000 | RPM | ok | na | 1200.000 | 1500.000 | na | na | na
CPUDIMMG0_PWR | 6.000 | Watts | ok | na | na | na | na | na | na
CPUDIMMG1_PWR | 3.000 | Watts | ok | na | na | na | na | na | na
CPU0_Status | 0x80 | discrete | 0x8080| na | na | na | na | na | na
SOC_TEMP | 45.000 | degrees C | ok | -128.000 | -10.000 | -5.000 | 105.000 | 108.000 | 127.000
CORE_VRD_TEMP | 41.000 | degrees C | ok | -128.000 | -10.000 | -5.000 | 88.000 | 93.000 | 127.000
CPU_DIMMG0_TEMP | 37.000 | degrees C | ok | -128.000 | 0.000 | 5.000 | 84.000 | 87.000 | 127.000
CPU_DIMMG1_TEMP | 41.000 | degrees C | ok | -128.000 | 0.000 | 5.000 | 84.000 | 87.000 | 127.000
M2_G0_AMB_TEMP | 41.000 | degrees C | ok | -128.000 | -10.000 | -5.000 | 100.000 | 105.000 | 127.000
HDD_TEMP_0 | 28.000 | degrees C | ok | -128.000 | -10.000 | -5.000 | 55.000 | 60.000 | 127.000
HDD_TEMP_1 | 33.000 | degrees C | ok | -128.000 | -10.000 | -5.000 | 55.000 | 60.000 | 127.000
MB_CPU_OUT_TEMP | 37.000 | degrees C | ok | -128.000 | -10.000 | -5.000 | 85.000 | 90.000 | 127.000
MB_CPU_IN_TEMP | 34.000 | degrees C | ok | -128.000 | -10.000 | -5.000 | 85.000 | 90.000 | 127.000
MB_ETH_TEMP | 37.000 | degrees C | ok | -128.000 | -10.000 | -5.000 | 85.000 | 90.000 | 127.000
FRONT_PANEL_TEMP | 30.000 | degrees C | ok | -128.000 | -10.000 | -5.000 | 85.000 | 90.000 | 127.000
PS1_Status | 0x1 | discrete | 0x0180| na | na | na | na | na | na
PS2_Status | 0x0 | discrete | 0x0080| na | na | na | na | na | na
PSU1_HOTSPOT | 41.000 | degrees C | ok | -128.000 | -10.000 | -5.000 | 102.000 | 105.000 | 127.000
PSU2_HOTSPOT | na | degrees C | na | -128.000 | -10.000 | -5.000 | 105.000 | 110.000 | 127.000
P_0V6_0123 | 0.588 | Volts | ok | na | 0.511 | 0.539 | 0.665 | 0.693 | na
P_0V6_4567 | 0.588 | Volts | ok | na | 0.511 | 0.539 | 0.665 | 0.693 | na
P_0V85_VDDC | 0.889 | Volts | ok | 0.630 | na | na | na | 1.078 | na
P_12V | 11.895 | Volts | ok | na | 10.270 | 10.790 | 13.130 | 13.650 | na
P_1V5_VDDH | 1.544 | Volts | ok | na | 1.176 | 1.308 | 1.852 | 1.896 | na
P_1V8_PCP | 1.814 | Volts | ok | na | 1.426 | 1.541 | 2.059 | 2.174 | na
P_1V8_VDDH | 1.814 | Volts | ok | na | 1.426 | 1.541 | 2.059 | 2.174 | na
P_3V3 | 3.270 | Volts | ok | na | 2.820 | 2.958 | 3.616 | 3.754 | na
P_3V3_SOC | 3.255 | Volts | ok | na | 2.730 | 2.898 | 3.696 | 3.864 | na
P_5V | 5.114 | Volts | ok | na | 4.292 | 4.498 | 5.474 | 5.680 | na
P_5V_STBY | 5.089 | Volts | ok | na | 4.292 | 4.498 | 5.474 | 5.680 | na
P_VBAT | 2.904 | Volts | ok | na | 2.311 | 2.502 | na | na | na
SOC_VDDCR | 0.756 | Volts | ok | na | 0.399 | 0.448 | 1.400 | 1.456 | na
P_VDDCR_CPU | 1.001 | Volts | ok | na | 0.399 | 0.448 | 1.400 | 1.456 | na
SYS_POWER | 100.000 | Watts | ok | na | na | na | na | na | na
MB_DIMMG0_O_TEMP | 34.000 | degrees C | ok | -128.000 | -10.000 | -5.000 | 115.000 | 120.000 | 127.000
MB_DIMMG1_O_TEMP | 36.000 | degrees C | ok | -128.000 | -10.000 | -5.000 | 115.000 | 120.000 | 127.000
MB_DIMMG1_I_TEMP | 33.000 | degrees C | ok | -128.000 | -10.000 | -5.000 | 115.000 | 120.000 | 127.000
MB_DIMMG0_I_TEMP | 33.000 | degrees C | ok | -128.000 | -10.000 | -5.000 | 115.000 | 120.000 | 127.000
In the idle state, PSU1_PWR_OUT/PSU1_PWR_IN * 100 implied an 86.2% power conversion efficiency.
At 100% CPU load, power conversion efficiency of the PSU1 is 92.98%
