clock idle load 1 load 2 load 3 load 4 2000MHz 283W 290W 297W 305W 311W 2333MHz 284W 296W 309W 317W 326W 2666MHz 285W 305W 324W 335W 347W 3000MHz 286W 313W 340W 354W 368WWith other programs producing the load we could get up to 412W with pure CPU load, 419W with CPU and memory load, and 423W by also accessing the hard disks.
load 0 1 2 3 4 5 6 7 8 2000MHz 283W 292W 302W 311W 321W 329W 336W 344W 353W 3000MHz 291W 310W 328W 345W 362W 377W 396W 413W 431WThe load was produced using
yes >/dev/null (IIRC that
load produced 412W at load 4 3000MHz before the CPU upgrade). By
replacing one yes process with a process that copies 1GB
of memory repeatedly, the power consumption at load 8 3000MHz went up
to 451W. Replacing more yes processes with
memory-intensive ones reduces power consumption.
load 0 1 2 900MHz 292W 321W 351WThe load was produced using
yes >/dev/null
clock voltage idle load 1 load 2 load 3 load 4 1000MHz 1100mV 121W 124W 127W 131W 134W 1800MHz 1350mV 161W 171W 181W 191W 202W 2000MHz 1350mV 167W 178W 190W 200W 212W ondemand 121W 154W 165W/189WLinux-2.6.14.3 seems to prefer to put the second process on the second chip, so we usually got the 189W consumption with two non-nice processes and the ondemand frequency governor; we got the 165W number by starting three processes, then killing the middle one, so that both processes run on the same chip, and the other chip idles.
clock idle load 1 load 2 2000MHz 100W 111W 121W 2333MHz 101W 115W 127W 2666MHz 102W 118W 133W 3000MHz 103W 123W 140W179W when running
glxgears in addition to two instances
of nice yes >/dev/null at 3000MHz. 150W-155W when
running UT2004 without active background jobs. 185W peak when running
Titanquest on Windows XP.
Variations on graphics cards:
Connect3D ASUS Sapphire Sapphire Gainward Sapphire
Radeon Nvidia Ultimate Radeon Radeon Ultimate
X850XT EN8600GT Radeon 3870 4650 Radeon
silent X1650Pro Ultimate 4670
98W 110W 94W 148W 95W 102W Linux idle (free drivers: radeon, nv), 2 HDs spinning
98W 103W 94W 104W 83W 87W Windows idle, usually 1 HD spinning
185W 152W 150W 202W 121W 142W Windows TitanQuest peak, usually 1 HD spinning
All but the X850XT and the 4650 are fanless; with the 3870 the power
consumption reached 230W in one of the 3DMark06 benchmarks.
yes >/dev/null
loads (and idle drives), I see:
clock idle load 1 load 2 1600MHz 103W 112W 122W 2133MHz 104W 120W 136W 2666MHz 104W 133W 156WWith the load
gforth -e ": foo begin again ; foo" used in
most other results here, I see:
clock idle load 1 load 2 2666MHz 104W 123W 132WThe difference between the loads was not as big on other machines where I tested both.
clock idle load 1 load 2 1600MHz 85W 94W 100W 2400MHz 86W 105W 115WMoreover, we tried a few different graphics cards on this machine, and measured the following (under Windows):
Card idle UT2004 Aquamark Radeon X550 86W 122W Radeon X800GTO 94W 163W Radeon X850XT 104W 181W 190W
~15W less with a Radeon 9600 (instead of the Gforce 4200). ~20W less when idle under Linux with cpufreq (Cool'n'Quiet support) @800MHz. With these changes, ~83W when idle, ~145W compiling, ~160W gaming.
|------ power ------|
clock voltage idle load 1 load 2
1000MHz 1200mV 83W 93W 102W
1800MHz 1250mV 86W 103W 121W
2000MHz 1300mV 88W 109W 130W
2200MHz 1350mV 92W 116W 143W
A very similar machine, but with an Athlon 64 X2 4600+ (cpu family 15
model 43) consumes as follows:
|------ power ------|
clock voltage idle load 1 load 2
1000MHz 1200mV 89W 98W 106W
1800MHz 1200mV 90W 105W 120W
2000MHz 1250mV 93W 112W 131W
2200MHz 1300mV 98W 120W 144W
2400MHz 1300mV 98W 122W 149W
BTW, load is a pure CPU load. We found significantly higher power
consumption for a memory-bound load (~147W on the second system with
load 1, and up to 160W with load 2; additional core-intensive work and
maybe some I/O to a PCIe graphics card should increase the power some
more).
load 0 1 2 3 4 5 6 7 8 2800MHz 53W 80W 96W 118W 138W 145W 149W 151W 155WPerformance tests at different load levels showed that Turbo mode did not work.