Power Consumtion (just got a new toy, a Kill-a-Watt)

[plonk420]

...and here's my findings:

new table (WIP): (w = World Community Grid)



CPU
GPU
Prime95
idle+Fur
P95+Fur
R@H / WCG
dimms
hdds
fans
setup notes
other notes


Ryzen 2400G
Vega 11 IGP



89w
2
ssd
2
b450-a pro
0.05v undervolt, full atx, 2933 RAM, 30 watt idle, 106 watt WCG+Einstein


Threadripper 1950X
HD4350



233w
4
ssd
2
designare ex
230 watts with WCG at 50%. idles at 67 watts.


Ryzen 1700X
GT 730 DDR5



136w
2
ssd
2
b450m mortar µ
145 watts with WCG + Einstein@Home


Ryzen 1700X
GT 730 DDR5



138w
2
ssd
1
ab350m-dh3s µ
idles at 45 watts. 147 watts with WCG + Einstein@Home


Ryzen 2400G
Vega 11 IGP



103w
1
ssd
1
ab350m-dh3s µ
idles at 32 watts. 125 watts with WCG + MilkyWay@Home


i7-4790
GTX1060 3GB
160
192
270
153w
2
1
2
lenovo c226 µ
1 HDD 1 SSD. idles at 65 watts (-35 watts on prime95 and WCG numbers for wattage without a dGPU)


i7-860
HD7790
143
153- 230
250- 330
128w
2
ssd
1
h55m-e21 µ
1 SSD. idles at 52 watts. wattage on tests with furmark started around the lower value but kept rising as heat kept rising. i stopped at 95C when there were visual artifacts


G3258
HD7790



79w
2
1
1
b85m-ds3h µ
52 watt idle


i7-920
HD5870
215
176
280*
195**
3
2
3
ex58-ud3r
*possibly bad testing method


i7-920
GTX560
207

370
185
2
2
2
ex58-ud3r
**extrapolated from the GTX560 test


i7-860
HD5870
180
330
410
160
2
1
2
h55m-e21 µ



x3 705e
HD3850
122


111
2
1
2
msi gf8200 µ



x2 4200
HD3850
119


115
2
1
2
msi gf8200 µ



x2 4200
9400GT
120


113
2
1
2
msi gf8200 µ



x4 9550
igp
196


174
2
2
2
zotac gf8200 µ



x4 9550
9400GT
220


198
2
2
2
zotac gf8200 µ



x4 9550
igp



124*
1
cf
1
m2a-vm µ, cf-to-ide card
*3 r@h, 1 wcg due to low ram


x4 9550
igp



135*
1
cf
1
m2a-vm, cf-to-ide, diff psu
*3 r@h, 1 wcg due to low ram


x4 630




116
1
cf
1
m4a-785 µ, cf-to-ide



E-350
igp
44
45
51
41
2
1
2
e35m1-m pro µ, pci sound
idle 35 watts


E-350
igp




1
1
2
e35m1-m pro µ
idle 29 watts!!!




















































R@H = rosetta@home
µ = microATX motherboard

old, original post:



i7-920 (1.25-1.3v), ex58-ud3r, 3x1 DDR3, hd5870, 2 hdds, 3 fans, HT on, pc power & cooling 750 watt
"default voltage": 122 watts idle, 215 p95, 176 idle+gpu load, 280 load+gpu load
bonus: 1.125v*: 121 watts idle, 187 p95, 180 idle+gpu load, 272 load+gpu load

1.125v, 9400GT, 113 watts idle, 181 p95 (bonus: 168 watts on rosetta@home)
1.125v, 8600GT, 115 watts idle, 183 p95 (bonus: 164 watts on rosetta@home)
default, 8600GT, 116 watts idle, 211 p95 (bonus: 184 watts on rosetta@home)
default, GTX560, 112 watts idle, 207 p95, 370 load+gpu (bonus: 185 watts on rosetta@home)

*the lowest i could go with prime95. 1.10v was stable with worldcommunitygrid but not prime. also i couldn't test "energy saver" mode due to some setting i'd set in bios and didn't feel like hunting down. who idles an i7 in powersaver, anyways?



i7-860, p55m-ud2 (micro atx), 2x1gb ddr3, 2 hdds 2 fans, HT, high performance, pc power & cooling 750 watt1.025v, 5870, 107 watts idle, 167 watts p95, (bonus: 149 watts on enigma@home*; 101 watts in power saver)

*for consistency, i'll have to add/replace rosetta@home later when i get more ram

i7-860, h55m-e21 (micro atx), 2x8gb ddr3, 1 hdds 2 fans, HT, default server 2008 r2 power, pc power & cooling 760 wattdefault, 5870, 88 watts idle, 180 p95, 330 idle+gpu, 410 p95+gpu (bonus: 160 watts on rosetta, 164 wcg)



msi gf8200, 2x2gb ddr2, 1 hdd, 1 fan, pc power & cooling 420 watt
x3 705e, hd3850, 90 watts idle, 122 p95 (bonus: 111 watts on rosetta@home, 76 watts on energy saver)
x2 4200+, hd3850, 76 watts idle, 119 p95, (bonus: 115 watts on rosetta@home, no energy saver)
x2 4200+, 9400GT, 74 watts idle, 120 p95, (bonus: 113 watts on rosetta@home, no energy saver)



x4 9550, zotac gf8200, 2x1gb ddr2, onboard, 2 hdd, 2 fan, antec basiq 550plus
default, onboard, 97 watts idle, 196 p95 (bonus: 174 on rosetta@home, 90 watts on energy saver)
default, 9400GT, 118 watts idle, 220 p95 (bonus: 198 on rosetta@home, 113 watts on energy saver)

the CPU stresser was prime95; the idle cpu+gpu load was distributed.net -bench, the CPU+GPU stresser was prime95+furmark

i used furmark only for CPU+GPU stresser as it seemed to max out a core. are there any GPU-only stressers? maybe a demoscene demo?

anyone else have a kill-a-watt?

anywho, more to come...

edit: more here:


x4 9550, m2a-vm, 1x1gb ddr2-800, onboard, 1 cf-to-ide running 8gb cf drive, 1x80mm fan, antec basiq 550plus, nic, as onboard is being flaky...also running linux
default, onboard, 78 watts idle, ?? p95 (bonus: 124 on rosetta@home* (sorta))


x4 9550, m2a-vm, 1x1gb ddr2-800, onboard, 1 cf-to-ide running 8gb cf drive, 1x80mm fan, nic, thermaltake TR2-430
default, onboard, 86 watts idle, ?? p95 (bonus: 135 on rosetta@home* (sorta))

*i could only get 3 instances of r@h in memory. the other i used a WCG unit, but the power usage didn't change


x4 630, m4a785-m, 1x1gb ddr2-800, onboard, 1 cf-to-ide running 8gb cf drive, 1x120mm fan, pc power and cooling 420 watt
default, onboard, 52 watts idle, ?? p95 (bonus: 116 on wcg)


E-350, ASUS E35M1-M PRO, 2x1gb ddr3-1333 (mushkin 1.5v), onboard, seagate 80gb 2.5" 5400 rpm sata drive, 2x120mm fan, pc power and cooling 370 watt, razer barracuda ac-1
default, onboard, 35 watts idle, 44 watts, p95, 45 watts furmark, 51 watts p95+furmark (bonus: 41 watts on wcg)


E-350, ASUS E35M1-M PRO, 1x2gb ddr3-1333 (mushkin 1.5v), onboard, seagate 80gb 2.5" 5400 rpm sata drive, 2x120mm fan, pc power and cooling 370 watt
default, onboard, 29 watts idle ... 31 watts idle with a DIAMOND Xtreme Sound 7.1 USB sound card



note: my kill-a-watt sits at 3 watts, even with nothing hooked up to it

[BlackAdder]

I saw one at Costco... will buy one once I'm back to work...280 watts isn't much at the outlet, just over 2 amps, nowhere near say a a/c window unit.

[Jason1478963]

I have one of those and sometimes use it to check the power factor of power supplies and appliances. It can be very handy and it can be found for around 17 dollars at times.

[plonk420]

added 2 new video cards, 9550

[plonk420]

added i7-860

[Jason1478963]

Looks like some good numbers. I recently purchased another one as the other one is monitoring my main pharm circuit so I know when I have to stop turning things on. I've been wanting to check my PS3 when its crunching. It sure seems like it uses more then a PC when its crunching. I may also try and get some numbers running ubuntu on a flash drive vs a hard drive see what the difference would be.

[plonk420]

edit:

you use yours 24/7? :O i was a bit uncertain as to the heartiness of them, so i disconnected them asap after getting stable readings...

as for ps3, i think there's numbers out there, SOMEWHERE... i think there's like a 90nm, 65nm and 45nm (slim)...

ah, here's some numbers:
http://www.hardcoreware.net/ps3-power-usage-during-foldinghome/
http://en.wikipedia.org/wiki/PlayStation_3_hardware#Configurations
http://en.wikipedia.org/wiki/PlayStation_3_hardware#Form_and_power_consumption

[Jason1478963]

The first meter i purchased at an energy fair about 4 years ago is still working fine for me. I used it to monitor a 11,600 btu air conditioner a year ago. It is somewhat lighter display then the new one but that could have been like that since I purchased it. I also have a watts up pro that i use to monitor the pharm too. It was getting warm so i split the load in half and put the killawatt on half and the wattsup pro on the other half of the receptacle. It is nice to know what the crunchers are using.

[plonk420]

added another configuration, one relevant to S. Starbuck's configuration ;)

[AMDave]

Thanks for keeping these up to date plonk.
Since I don't have one of those monitoring units, this is a very helpful guide.
It seems my boxen are probably using only 2 thirds of what my estimate was...so I added in a couple of case fans :icon_twisted:
Temps are down about 2 degrees and with the Southern summer starting to kick in thats a relief.

[plonk420]

i'd happily exchange some solar heat from .. wherever you are! it was ~2F/-17F, here, and of all times it could happen, the heat broke :0 not even my small farm of computers could help noticeably... :S

also, added another PSU. odd findings...

[vaughan]

Computer room reached 41C last night and one of the computers decided to burnout - acrid electrical smell. Forced me to turn off every PC at the power switch and then switch back on one at a time until I found the culprit. A Shuttle SD37P2v2 with an Intel Q6600 cpu. :(

I was worried the smoke alarm would get triggered and "she who must be obeyed" would launch a tirade.

Bah-humbug :icon_santa: - greenhouse gases, more like brownhouse :icon_twisted:

What global warming :icon_wink:

[AMDave]

Ooh. That sucks.
I have been running at 37 to 40 ambient lately with a 13 inch box fan sitting in the open doorway of the study for the last month.
The logs show it has peaked to 41 and 42 briefly while at work.
Your event is making me think another strategy is in order.
This house is hot by design. We know that.
Maybe I can arrange to move some CPUs to a better location for a few weeks to pass the summer peak.
I guess I need to start some schmoozing this weekend ;)

[plonk420]

added Athlon II 630... 52 watts idle at Ubuntu desktop! (not sure if windows desktop in Power Saver/Balanced might save even more!)

edit: dang, i thought i dropped it to 110-112 watts after using casefan on 5V rail instead of 12V. it's back up to 116, though.

[plonk420]

added Zacate (E-350) ASUS E35M1-M PRO!

my god, it's quiet. O_O

notes:

can stream (edit:) SOME (realtime) decrypted blu-rays over a network

in MPC-HC:

~40% cpu usage with vc-1: Harry Potter 4
~20% cpu usage with h.264: MirrorMask (however, it would occasionally skip. seems like an app buffering issue?)

ONLY WITH iTMS* encodes can it do a 720p to 1080p bicubic scaling in ffdshow (superior to hardware scaling) with no hitches (tested with Castle)
"high profile level 4.1 h.264" (commonly "x264" mkv with ac3/dts) 720p to 1080p it struggles with bicubic scaling

sadly, Sinc scaling is hopelessly unusable

720p60 worked perfectly, sans scaling

edit:

xbmc could play sd xvid, sd h.264, 720p24/25/60, 1080p24 high profile level 4.1 h.264, wmv9 (from gametrailers)
xbmc couldn't play 1080p60 h.264

*iTunes Music Store

[Tamaster]

Thanks for the info. I've been looking at building an ITX based PVR using one of the new APUs from AMD.
(gotta find an extra input into the HT setup though).

I've a Sony BR player that it'll probably replace and I need Blu-Ray capability as well.

[plonk420]

it seems XBMC (in Win7) can play the unencrypted h.264 Mirrormask M2TS (1920x1080 on a 1440x900 monitor) on the E-350

edit: whereas it couldn't even play the M2TS straight from the (5400 rpm 2.5") HDD in MPC-HC. i need to try an official blu-ray app ... but i'm not the most motivated to move my only BD-ROM drive from one case to another... :S

[Jason1478963]

Interesting info guys.... I have been trying to find a decent PC based dvr and am currently using a windows vista laptop with mediacenter. I am still looking for a better setup as well.

[plonk420]

i wouldn't have minded a 25% cpu processing power increase at Radeon HD6250 speeds :S

[plonk420]

i got it to 29 watts, even with a Lenovo wireless mini keyboard plugged in (but without my new Diamond USB soundcard plugged in)! woot!

it probably can be gotten well under 29 watts even with a 3rd party soundcard with an SSD and probably 1.25v DDR3L ... a picoPSU might help a lot, too!

[Tamaster]

That's a pretty sweet power envelope! Not much cabinet heat in 29 watts.

I also have a BR laptop with Media Center that I use for Netflix, but as I've said I'd really like to build an HTPC/PVR for a more permanent installation.

What kind of HDD are you currently running?

[plonk420]

Seagate LD25.2 Series ST980210AS 80GB 5400 RPM SATA 2.5" it seems (with a warning on the label "not for laptop use!") heh

[plonk420]

added GTX560. not sure if my 5870 (once RMA'd already) is dying again. 2-3 lockups per League of Legends game. altho the day before my 560 arrived, i played fine all night.

also, lol, 100 watts more than my 5870 xD

[Jason1478963]

sorry to hear about your 5870 giving you trouble. I have a 5850 by Asus that does well for the game I played. Its hard to believe the Nvidia is using that much more power over a 5870

[plonk420]

updated with i7-860 at default voltages.

however, i think i've been incorrectly testing the GPU in the past... even tho Furmark was lower than 60fps, i wasn't using THAT much power... power consumption went WAY up when i used 1080p mode O_o

[plonk420]

oops, was testing tables and accidentally hit submit instead of preview. here goes, tho.



CPU
GPU
Prime95
idle+Fur
P95+Fur
R@H
setup notes
other notes


i7-920
HD5870
215
176
280
195*

*extrapolated from other tests


i7-920
GTX560
207

370
185




i7-860
HD5870
180
330
410
160




x3 705e
HD3850
122


111
msi gf8200 uATX



x2 4200
HD3850
119


115
msi gf8200 uATX



x2 4200
9400GT
120


113
msi gf8200 uATX



x4 9550
onbd
196


174
zotac gf8200 uATX



x4 9550
9400GT
220


198
zotac gf8200 uATX



x4 9550
onbd



124*
m2a-vm
*3 r@h, 1 wcg due to low ram


x4 9550
onbd



135*
m2a-vm, diff psu
*3 r@h, 1 wcg due to low ram


x4 630
onbd



116




E-350
onbd
44
45
51
41
uATX mobo, pci sound
idle 35 watts


E-350
onbd




uATX mobo
idle 29 watts

[Dirk Broer]

Can you test the power consumption under the various BOINC projects?

[plonk420]

i'm not THAT interested in doing so, but i might if it's quick and easy and i already have an account.

i feel pretty bad about aborting a unit, though. do any projects really care if you abort a unit?

[Dirk Broer]

I don't think aborting is part of testing what the power consumption of a WU of any given project is.
Please feel free to test for the projects you are already participating in.

[plonk420]

not really wanting to tie up the cpu too much as i've been gaming a lot lately...

like, for example, i think R@H (current 4/2012 version) was 160 watts on the i7-860 and wcg 164 watts, oddly. i'm sure BURP will be near-x264, but i don't have enough space on the HDD :(

edit: poem@home is 160 watts

edit2: burp fluctuates from as low as 106 watts up to (high average) 160 (154-164)... max sustained (for more than a few seconds) i saw was 175.

[Dirk Broer]

Well, that's less than I feared. I'll try to get hold of a kill-a-watt or equivalent here and run it for the 20-odd projects I'm involved in.

[plonk420]

some 3DMark (2013) tests with a Haswell Pentium G3258 + 7790 were mildly interesting:

52 watts at idle

x264: 80 watts

world community grid: 79 watts

with a Radeon 7790:

as i ran 3DMark, the wattage usually increased: Ice Storm and Cloud Gate (complex scenes) started at 145 watts, then hit up to 160 (ignoring peaks of 165; one or two moments higher) ...also 95 watts on CPU physics

Sky Dive hit 187-198 watts (103 watts on CPU physics)

Fire Strike actually started at 150 or so, and edged its way up to 180 (plus a few spikes into 190), but was overall lower than skydive o_O

[Dirk Broer]

Sky Diver is really pushing the limits (http://www.3dmark.com/hall-of-fame-2/sky+diver+3dmark+score+performance+preset/version+1.0) -if only Watt-wise...

[plonk420]

added new HDDless i7-860 setup, i7-4790, Ryzen 2400G.

yay necroposting! \o/

oh, Enigma@Home on H7790 only added about 65 or 75 watts to the 860. can't remember exactly

anyone else have a Kill-A-Watt and want to pitch in their numbers?

[plonk420]

added Ryzen 7 1700X with GT730

[Dirk Broer]

added Ryzen 7 1700X with GT730

Where can we see the results?

[plonk420]

see the first post. also my new board arrived sooner than expected (was monday last i checked). power consumption lowered a few watts just with a different board, VRMs!

AB350M-DS3H = crappy 4C10N/4C06N, teeny wickle VRM cooler = 86C on the VRMs

B450M Mortar = slightly better VRMs with 25% less heat, better VRM cooler = 65C on the VRMs

the DS3H is ok with the Ryzen 2200/2400G, tho.

[Dirk Broer]

see the first post. also my new board arrived sooner than expected (was monday last i checked). power consumption lowered a few watts just with a different board, VRMs!

AB350M-DS3H = crappy 4C10N/4C06N, teeny wickle VRM cooler = 86C on the VRMs

B450M Mortar = slightly better VRMs with 25% less heat, better VRM cooler = 65C on the VRMs

the DS3H is ok with the Ryzen 2200/2400G, tho.

I use reverse sorting (most recent message on top), for me it was the last message -and I encountered another table sooner.

[plonk420]

Threadripper 1950X added. interestingly, same power consumption for Blender (Sheep-it) opposed to 32 threads of WCG. and 16 threads of WCG is just 3 watts lower