8350 Power Consumption

[Daveburt714]

I've been tuning my rig for power efficiency so I thought I'd share some results...
This chip will run daily stable ~5Ghz and bench ~5.3, but it takes a lot of v's and becomes very inefficient!
For daily use I've settled for the following clocks (IMC voltage is 1.275):



Just for reference, this is my main rig, it's water cooled and loaded for bear!
Here are some specs so you folks know what else is sucking power:

Five/120mm fans for the rads
Two/ 140mm fans for the case/HD cage
OCZ Mem cooler fans over the PWM heatsink
Aftermarket Zalman dual fan cooler for the 6870 vid card
MPC35X Pump and a C5F 990FX mobo.

With CnQ enabled here is my idle draw:


Here's the draw while running the WinRAR benchmark (represents max real world use, including WCG & encoding video):


And finally, the power draw while running P95 Blend as shown in the 1st SS:


Someone over in Xtreme News said the 8350 drew 200w by itself when OC'd...
That may be true when pushing the ABSOLUTE limits, but with a little tuning the 8350 isn't all that bad!

[FlanK3r]

Nice work Dave, could you mess power cons. in x264fhd and r11.5? thx

[Daveburt714]

Here you go Flank3r...

Nothing too impressive as far as MAX clocks, but these were all done with low v clocks (as shown above).





Keep in mind that when I ignore power requirements I've hit 9.2+ on CB11.5...

Here's a WinRAR bench @ 4.6 too, I know you didn't ask for it, but it's still pretty decent @ the low v clocks...

[FlanK3r]

yes, this C11.5 score is one og the best what I got with Thuban at maximum ...And at x264fhd is no change get this with "old" good Thuban

[behrouz]

Nice ,the Fx-8350 (at 4.6 Ghz) beat Core i7-2600 ( at 4.9 Ghz) ,

[stangracin3]

What he was claiming at first was that at stock it was a 200 Watt processor.

Then he moved on to claim turbo was over 200.
Then he moved on to saying overclocked.

[Evantaur]

Well my rig is sucking. About 480W from the wall (spiked at 608W, but i dont know when)

That is with 8350 @ 4,5GHz and 2x7970s

[Daveburt714]

Well my rig is sucking. About 480W from the wall (spiked at 608W, but i dont know when)

That is with 8350 @ 4,5GHz and 2x7970s

I assume that was while running WCG? If so, I'm sure those 7970's are drawing a lot of power, especially if your using the mod to allow multiple HCC wu's.

Just for reference, here's what I get while running WCG (w/HCC on 6870 VC).

[Evantaur]

I assume that was while running WCG?

of course

[tbone8ty]

so its safe to say vishera/piledriver cpus use much less power once OC'd then zmabezi/bulldozer?

[stangracin3]

so its safe to say vishera/piledriver cpus use much less power once OC'd then zmabezi/bulldozer?

yes very safe to say

[The Stilt]

8-core FX within 45W TDP envelope...

Note for Prime95 LFFT: At stock specs the rated TDP was exceeded by 15.5% during Prime95.
The same margin is allowed for 45W TDP test too.

Frequency: 2.7GHz
Voltage: 0.925V

Prime95 LargeFFT
DCR Imax (A): 47.25A
DCR Vsen (V): 0.919V
DCR Pmax (W): 43.42W

43.42 / 1.155 ("Prime95 margin") = 37.6W
37.6W + 6.75W (BaseTDP = NB, Logics) = 44.345W

X264 R2230 64/10-bit BD-RAW 1080p -> MKV
Preset: Slow, Tune: Film, Bitrate: 9000, Output: MKV

DCR Imax (A): 41.5A
DCR Vsen (V): 0.919V
DCR Pmax (W): 38.13W

38.37W + 6.75W (BaseTDP = NB, Logics) = 44.89W

[AMDforME]

With both P95 and OCCT stress testing I have seen the power consumption under full load change as much as ~90w depending on what is actually being run at the time you view the Kill-A-Watt meter. You also need to take into account the PSU efficiency and work the number viewed as a percentage of the actual PSU draw, i.e 90% for a good PSU.

[BeepBeep2]

^Woah!

Nice results "The Stilt"

[FlanK3r]

he is magic man. He got glory of name in my review

[cab]

371 Wattoosies?! In winter that's a good thing, you need the heat anyways..

[tbone8ty]

Nice stilt

8 core fx laptop chip!!!

[Daveburt714]

Very Impressive Stilt!

I was running mine ~4.85 @ 1.475v and power consumption was pretty bad...
That's why I decided to try and find a nice balance between OC and power usage.

At stock VID (1.3375), I could get stability ~4.3 Ghz. I never considered underclocking because I use this machine for serious work.
After some tuning and checking bench scores this seemed like the best balance for my uses. Yes, power consumtion is still a little high, but these clocks aren't bad...

This seems like a decent sample (1244 PGT). Next time I get bored I'll see how low I can get it by underclocking.
If it'll get down to the numbers you got it may make a nice server/HTPC chip down the road...

[Zeus]

Dave, if i'm not mistaken The Stilt does measure amps with a clamp at the EPS wires (4/8 pin connector). This way he will measure CPU only current. You measure power drawn by the entire system, not just the CPU.

I'm not sure all the current drawn by the CPU is fed by just the EPS wires but form what i've read from The Stilt, that's the way i understand it. If you multiply this current times the Vcore you're using you get a power number in watt.
Divide this by the efficiency percentage of your PSU and you get number of CPU power drawn.

Maybe The Stilt can clarify and/ or acknowledge this this?

[The Stilt]

Well under 25W average now, the peak is barely over...

Note for Prime95 LFFT: At stock specs the rated TDP was exceeded by 15.5% during Prime95.
The same margin is allowed for 25W TDP test too.

I did not underclock the NCLK domain as I wanted to keep the L3 cache at stock performance level.

Frequency: 1.8GHz
Voltage: 0.7875V
NCLK Frequency: 2200MHz
Voltage: 1.075V

I cannot make separate DCR measurements for VRM Loop 2 (CNB) currently, so the new BaseTDP is an estimate.
Default BaseTDP = 6.75W (CNB 1.1625V -> 1.075V): 6.75 * ((1.075 / 1.1625)^2) = 5.77W

Prime95 LargeFFT
DCR Imax (A): 28.75A
DCR Vsen (V): 0.781V
DCR Pmax (W): 22.45W
EPS12V Clamp Imax (A): 2.95A
EPS12V Vin (V): 12.13V
EPS12V Pmax (W): 35.78W

22.45 / 1.155 ("Prime95 margin") = 19.4W
19.4W + 5.77W (BaseTDP = NB, Logics) = 25.17W

X264 R2230 64/10-bit BD-RAW 1080p -> MKV
Preset: Slow, Tune: Film, Bitrate: 9000, Output: MKV

DCR Imax (A): 27.75A
DCR Vsen (V): 0.781V
DCR Pmax (W): 21.7W
EPS12V Clamp Imax (A): 2.58A
EPS12V Vin (V): 12.13V
EPS12V Pmax (W): 31.3W

21.7W + 5.77W (BaseTDP = NB, Logics) = 27.47W

DCR = Current flow over VRM inductors
EPS12V = Current draw from EPS12V connector measured with a separate clamp meter (VRM Iin, both loops included).

Cooling: Passive Thermalright Ultra 120 Extreme (not very well suitable for the purpose...), peak temperature after Prime95 LFFT 63 celsius (TSI).

[FlanK3r]

very nice man! Is it possible add dynamically profiles to PS Check software? I would like in next week rebuilding my Bulldozer for Vishera and after try PS check software (I plaiyng with this only for OC records etc)

[The Stilt]

It will be more dynamic if you use Turbo Core PStates with reconfigured TDP limits.
Cannot be done with PSCheck, however Orochi Control Interface (release TBD) supports this...

[FlanK3r]

Hope, we will see this utility soon

[tbone8ty]

Well under 25W average now, the peak is barely over...

Note for Prime95 LFFT: At stock specs the rated TDP was exceeded by 15.5% during Prime95.
The same margin is allowed for 25W TDP test too.

I did not underclock the NCLK domain as I wanted to keep the L3 cache at stock performance level.

Frequency: 1.8GHz
Voltage: 0.7875V
NCLK Frequency: 2200MHz
Voltage: 1.075V

I cannot make separate DCR measurements for VRM Loop 2 (CNB) currently, so the new BaseTDP is an estimate.
Default BaseTDP = 6.75W (CNB 1.1625V -> 1.075V): 6.75 * ((1.075 / 1.1625)^2) = 5.77W

Prime95 LargeFFT
DCR Imax (A): 28.75A
DCR Vsen (V): 0.781V
DCR Pmax (W): 22.45W
EPS12V Clamp Imax (A): 2.95A
EPS12V Vin (V): 12.13V
EPS12V Pmax (W): 35.78W

22.45 / 1.155 ("Prime95 margin") = 19.4W
19.4W + 5.77W (BaseTDP = NB, Logics) = 25.17W

X264 R2230 64/10-bit BD-RAW 1080p -> MKV
Preset: Slow, Tune: Film, Bitrate: 9000, Output: MKV

DCR Imax (A): 27.75A
DCR Vsen (V): 0.781V
DCR Pmax (W): 21.7W
EPS12V Clamp Imax (A): 2.58A
EPS12V Vin (V): 12.13V
EPS12V Pmax (W): 31.3W

21.7W + 5.77W (BaseTDP = NB, Logics) = 27.47W

DCR = Current flow over VRM inductors
EPS12V = Current draw from EPS12V connector measured with a separate clamp meter (VRM Iin, both loops included).

Cooling: Passive Thermalright Ultra 120 Extreme (not very well suitable for the purpose...), peak temperature after Prime95 LFFT 63 celsius (TSI).

so this is the new PS4 amd chip their using

[tbone8ty]

where do i find the x246 benchmark?

is this it?

http://www.techarp.com/showarticle.aspx?artno=520