• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

i7 4770 Haswell (4 Gen) power consumption measurements on idle.

Joined
Dec 24, 2008
Messages
2,062 (0.34/day)
Location
Volos, Greece
System Name ATLAS
Processor Intel Core i7-4770 (4C/8T) Haswell
Motherboard GA-Z87X-UD5H , Dual Intel LAN, 10x SATA, 16x Power phace.
Cooling ProlimaTech Armageddon - Dual GELID 140 Silent PWM
Memory Mushkin Blackline DDR3 2400 997123F 16GB
Video Card(s) MSI GTX1060 OC 6GB (single fan) Micron
Storage WD Raptors 73Gb - Raid1 10.000rpm
Display(s) DELL U2311H
Case HEC Compucase CI-6919 Full tower (2003) moded .. hec-group.com.tw
Audio Device(s) Creative X-Fi Music + mods, Audigy front Panel - YAMAHA quad speakers with Sub.
Power Supply HPU-4M780-PE refurbished 23-3-2022
Mouse MS Pro IntelliMouse 16.000 Dpi Pixart Paw 3389
Keyboard Microsoft Wired 600
Software Win 7 Pro x64 ( Retail Box ) for EU
My compo is:
i7 4770 + DDR3 with XMP 2400 active (Jedec 1600) + GTX1060 6GB + six HDD.

I have my tower alone, this be connected to a low-cost power meter.
Idle consumption this is never less than 95 Watt at 240V.
I did one experiment and I did activate Haswell lowest power state, = all CPU core goes down to 800MHz instead of 3.7GHz (on idle).
The strange thing is that power consumption this does not decrease at all.

I do not need any advice to under-volt anything, I am simply asking for a confirmation if such behaviour this is to be expected.
Statistically, if I was using INTEL GPU instead, the numbers it would be lower.

Therefore the math changes:
a) minimum consumption of the NVIDIA VGA
Plus
b) minimum consumption of DDR3 at 2400
Plus
c) minimum consumption of the CPU it self.
Total of 95 Watts idle.
Approximately 250W when Gaming (bound to monitor refresh).
Approximately 300+W when Gaming ( GPU at max performance).
 
4790K, all power saving enabled, 800MHz idle
MSI H97-G43
16GB DDR3L 1.35V
GTX 970
1TB Hynix P31 Gold
500GB Evo 850
Seasonic G650 Gold

Idle power 240V = 43.5W
Max gaming load = 220W


Rig in profile: Idle = 52W, max gaming load = 200W
 
Last edited:
Six HDDs?

Are they spinning?
 
I did activate Haswell lowest power state
How did you enable it? In the BIOS, did you enable all energy-saving features and set them to their lowest values for minimum power consumption?
 
Motherboard + RAM consumes some power - 15-30W depending on the specifics (motherboard features, number of DIMMs, RAM speeds and voltages, etc.). CPU consumes some power, as does the dGPU. dGPUs are particularly inefficient at idle as the whole board still needs to be active, despite not doing much. If your HDDs are spinning, even at reduced speeds that consumes some power. Looking at your setup, your audio card and front panel will also be consuming some power. Fans consume a tad of power - a couple of watts in total at low speeds, most likely. Peripherals and other things also consume some power. It all adds up in the end. Measuring at the wall also includes PSU efficiency losses, which at such low power levels are likely disproportionately high - in the 80% range most likely, going by the Tom's review. Assuming 80% efficiency and that your power meter is accurate at 95 watts, that means you have an internal power draw of ~76 watts.
 
How did you enable it? In the BIOS, did you enable all energy-saving features and set them to their lowest values for minimum power consumption?
There is a Gigabyte software utility EasyTune.
I did use it for testing, but at Low power mode, this disengage the XMP of 2400, and returned the RAM at 1600.
I think that this is one undocumented detail, of what are actually the lowest possible core clocks and consumption of Haswell, when this is with XMP locked at 2400.

hardware monitor idle - due default Bios settings.

HRWD-monitor.jpg
 
Motherboard + RAM consumes some power - 15-30W

Say 5-40W, depending on NIC and USB and other things. (X58 being a exclusion where the bridge ate 50W alone and idled at 100C as a norm)

My H97 + Xeon E3 1285L v3 with 2x2GB DDR3 and Samsung oem nvme and idling two 2TB HDD NAS PC. Eats 25W from the wall. It would be even less if I would switch off two 120mm case fans gaining 3W.
 
Last edited:
Honestly, Intel Power Gadget is best if you want just CPU power readings.
 
It's probably the HDDs that are causing your woes, Haswell & 1060 are pretty good at idle.
 
I have a snow storm outside and I need to leave.
I got a screenshot of software measurements, so to compute them later on.
Electrical measurements due software they worth something only when they are related with real world measurements.
My current PSU its not fresh, but neither I would expect this wasting that much energy in comparison to a fresh at idle.

sens.jpg
 
Everything uses some power, even when idle. So with just your tower case connected to your power meter, the CPU, GPU, RAM, drives, fans, and the motherboard itself are all consuming some power.

While my modem and router are also connected, when I power off my monitors, my little i5 system with 4 sticks of DDR4 RAM and 2 SSDs are still pulling 65w, according to my UPS.

You have an i7, DDR3 (which eat more power than DDR4) and 6! hard drives.

And BTW, it is the number of RAM sticks that matters most, not the size. That is, a 16GB stick consumes essentially the same as a 8GB stick. This means 2 x 8GB will pull significantly more than 1 x 16GB. I don't see where you said how your RAM is configured. But the more sticks, the more power - even at idle.

I think your 95w seems about right.
 
I have a snow storm outside and I need to leave.
I got a screenshot of software measurements, so to compute them later on.
Electrical measurements due software they worth something only when they are related with real world measurements.
My current PSU its not fresh, but neither I would expect this wasting that much energy in comparison to a fresh at idle.

[snip]

FYI: according to an old ixbt review, wd740gd specifies < 8 watts idle, wd1600js < 1 watt and sa400 < 1 watt.
 
FYI: according to an old ixbt review, wd740gd specifies < 8 watts idle, wd1600js < 1 watt and sa400 < 1 watt.
I have the real thing at my safe box.
But the math of DCV Watt, this changes at ACV Watt (They are much less).
Never the less, one pair of RAPTOR's will be replaced soon with a pair of WD Gold 1TB = 7W its one too.



WDD.jpg



EDIT:
My older system was always slightly OCed and I did not care much of Windows power plan and of Intel speedstep.

The new system it is fastest out of the box, CPU and memory performance both extremely impressive, especially the memory bandwidth.
I had to redo my homework, by testing Bios and Windows power plans.

Windows Balanced power plan + HDD-off after some time ... this saved only 5W ( down to 85W)
Windows Balanced power plan + HDD-on = about 90W
Windows Balanced power plan: this aloud Intel SpeedStep to get active = all CPU core's at 800MHz ( no reduction of power consumption on idle)

BIOS: CPU Turbo Boost ( On / Off) = ( no reduction of power consumption on idle)
BIOS: RAM DDR3 at Jedec DDR-1600 vs XMP-2400 ( no reduction of power consumption on idle)

So, I will simply have to understand and or accept, that only one fresh PSU with high efficiency rate at Low-Watts, this will lead to a measurable lower consumption.
Because I am now fully aware, that PC individual components, they do the best that they can all ready, at my Pro workstation.
 
Last edited:
I have the real thing at my safe box.
But the math of DCV Watt, this changes at ACV Watt (They are much less).
Never the less, one pair of RAPTOR's will be replaced soon with a pair of WD Gold 1TB = 7W its one too.


EDIT:
My older system was always slightly OCed and I did not care much of Windows power plan and of Intel speedstep.

The new system it is fastest out of the box, CPU and memory performance both extremely impressive, especially the memory bandwidth.
I had to redo my homework, by testing Bios and Windows power plans.

Windows Balanced power plan + HDD-off after some time ... this saved only 5W ( down to 85W)
Windows Balanced power plan + HDD-on = about 90W
Windows Balanced power plan: this aloud Intel SpeedStep to get active = all CPU core's at 800MHz ( no reduction of power consumption on idle)

BIOS: CPU Turbo Boost ( On / Off) = ( no reduction of power consumption on idle)
BIOS: RAM DDR3 at Jedec DDR-1600 vs XMP-2400 ( no reduction of power consumption on idle)

So, I will simply have to understand and or accept, that only one fresh PSU with high efficiency rate at Low-Watts, this will lead to a measurable lower consumption.
Because I am now fully aware, that PC individual components, they do the best that they can all ready, at my Pro workstation.

About your tests, I'm not sure old HDDs like raptors can be physically turned off by the OS, so I wonder if what you're seeing is that only the newer drives are powering down, hence very modest power savings?

I'm not surprised about turbo boost, but I'm a little surprised about the RAM, I thought it might save a watt or two to operate at a lower voltage, but maybe 1600 is still 1.5v?

About the CPU frequency, that's unexpected, I'd have thought that at a lower power state and voltage it would save quite a lot. Did you tweak other settings? I think there's a program that can check what power states are enabled (throttlestop is it?)?
 
About your tests, I'm not sure old HDDs like raptors can be physically turned off by the OS, so I wonder if what you're seeing is that only the newer drives are powering down, hence very modest power savings?

It is actually undocumented of how Windows control multiple HDD power.

In my box:
Windows does not turn off the RAID Array. (OS drive) for Power savings on idle
One Raptor along one WD for storage, they are set as Pagefile locations. (interesting enough windows use them simultaneously, at to was Pagefile over RAID-0)
The SSD this is game storage.

I am starting to believe that Windows use different power-management setting at.
HDD OS container'
HDD Pagefile location
HDD / SSD plain storage location

In this example, Windows they probably choose to turn off, the WD for storage and the SSD.
I did wrote about six HDD, they are actually four , plus the SSD and one optical media DVD recorder ( sum of consumption almost equal to six HDD)


About the CPU frequency, that's unexpected, I'd have thought that at a lower power state and voltage it would save quite a lot. Did you tweak other settings? I think there's a program that can check what power states are enabled (throttlestop is it?)?
Active Speedstep this resulted 2C temperature reduction on idle (this is the only non-important but measurable benefit :) )

There is no other settings effective for the idle state.
Other than that, the system can sleep normally, and or hibernate normally, and recover back from both states normally.
Over all its perfect.
 
Last edited:
Active Speedstep this resulted 2C temperature reduction on idle (this is the only non-important but measurable benefit :) )

There is no other settings effective for the idle state.
Other than that, the system can sleep normally, and or hibernate normally, and recover back from both states normally.
Over all its perfect.

I guess it depends on the motherboard, but usually there are: EIST (enable/disable), C states, highest (or lowest) C state permissible and cTDP. C1E is often separate too. And some motherboards (at least, I have seen it on asus and msi) have an idle state control for keeping compatibility with older PSUs.
 
I guess it depends on the motherboard, but usually there are: EIST (enable/disable), C states, highest (or lowest) C state permissible and cTDP. C1E is often separate too. And some motherboards (at least, I have seen it on asus and msi) have an idle state control for keeping compatibility with older PSUs.
GA-Z87X-UD5H this has a special BIOS setting, this mentioning activation of special mode, for older PSU compatibility.
At sleep mode, it can increase the level of lowest electrical consumption ( by activation of an simulation circuit this causing higher DC load).
I did disable this workaround trick so to test of how my old Corsair CX750 (of 2013) this will react.
My find is that the specific PSU does not need this workaround.

The system sleep with 2.5 ~ 3.5 W total consumption from the wall plug.
I am now in research to get a much more serious digital power meter (high precision one) because the cheap wall plug type, they are inaccurate at low-watt and tremendously slow.
 
Last edited:
Starting to feel like you just need some Atom, Celeron, or Kabini chip since you're so worried about power draw.
 
I am just exploring its room of my new home and of what all these power switches are there for.
8-series-chipset-pch-datasheet.pdf this talks about everything at the finest detail.
 
I am just exploring its room of my new home and of what all these power switches are there for.
8-series-chipset-pch-datasheet.pdf this talks about everything at the finest detail.
Power savings is one of the major advantages of more recent platforms - there's a reason laptop battery life has doubled or more since Haswell. Still, there are always optimizations to be made on desktops, as they are rarely configured for efficiency out of the box.
 
I think that I had to write it up front, that all comparison points are for high performance professional workstations.
My only problem so far, this is that I am surrounded by three huge PSU (750W ) all of them not better than 83% efficiency at full load.

Today I got my third and I am confused if this worth to be refurbished and get back in action, due it electronic design age of 2007.
My HIPER 780W has almost the same age.
My Corsair CX750, a bit more modern but more weak due it DC/DC 12V, that is not that well made.

W0116RE-750W_2.jpg


W0116RE-750W_3.jpg
 
Last edited:
Back
Top