My 4870 1GB power consumption numbers

[thilanliyan]

Power consumption:

-CPU (0.912v--967MHz **idle** ), GPU (1.08v--550/400 **idle** ) --> 175w
-CPU (0.912v--967MHz **load** ), GPU (1.08v--550/400 **idle** ) --> 191w
-CPU (0.912v--967MHz **idle** ), GPU (1.08v--550/400 **load** ) --> 207w
-CPU (0.912v--967MHz **idle** ), GPU (0.925v--550/400 **idle** ) --> 175w
-CPU (1.35v--3010MHz **idle** ), GPU (1.08v--550/400 **idle** ) --> 191w
-CPU (1.35v--3010MHz **load** ), GPU (1.08v--550/400 **idle** ) --> 320w
-CPU (1.35v--3010MHz **idle** ), GPU (1.26v--800/1000 **load** ) --> 335w
-CPU (1.35v--3010MHz **load** ), GPU (1.26v--800/1000 **load** ) --> 460w


These are full system results measured with a Kill-a-Watt meter (efficiency not factored). In addition to the specs in my sig, I have a COrsair HX620, 3 hard drives, 1 DVD-RW, 4x120mm fans, 3x40mm fans, and the Swiftech MCP355 pump all drawing power.

Load results for just the GPU was with Furmark (renamed) and there was a little bit of CPU usage on one core. Full load on CPU and GPU was done with Prime95 large FTT and Furmark running at the same time. I went lower with the GPU and CPU voltages but it didn't lower power consumption by any noticeable amount.

Remember this is at the socket, so assuming 80% efficiency, the fully loaded system would draw about 370w.

EDIT: According to S44 the kill-a-watt meter doesn't measure properly from active pfc PSUs.

 

[kmmatney]

Damn, amazing to see the power jumps with GPU speed and voltage. Your power consumption is a bit higher than this review:

http://hothardware.com/Article...Mainstream-GPU/?page=8

Thanks for the info though - useful for those wondering about their power supplies

 

[thilanliyan]

Originally posted by: kmmatney
Your power consumption is a bit higher than this review:

http://hothardware.com/Article...Mainstream-GPU/?page=8

Probably due to the different total systems used.

 

[evolucion8]

I'm seeing that in the hothardware review, the HD 4870 1GB consumes less power at load than the GTX 260+. It's the same phenomenom that happened before when the HD 4870 1GB was first launched and used less power compared to the HD 4870 512MB because of improved yields etc, seems that with newer RV770 chips, that keeps happening. But why it doesn't happen with the GTX 260+? Unless if that GTX 260+ card is older, or because since it was designed specifically to the 65nm process (That's why they weren't able to made it to 40nm) it won't see much improvement, like it happened with the 55nm besides overclocking.

 

[Wreckage]

Originally posted by: evolucion8
I'm seeing that in the hothardware review, the HD 4870 1GB consumes less power at load than the GTX 260+. It's the same phenomenom that happened before when the HD 4870 1GB was first launched and used less power compared to the HD 4870 512MB because of improved yields etc, seems that with newer RV770 chips, that keeps happening. But why it doesn't happen with the GTX 260+? Unless if that GTX 260+ card is older, or because since it was designed specifically to the 65nm process (That's why they weren't able to made it to 40nm) it won't see much improvement, like it happened with the 55nm besides overclocking.

The 4870 uses significantly more power at idle than the GTX260.
http://www.techreport.com/articles.x/16504/10

 

[s44]

Originally posted by: thilan29
Remember this is at the socket, so assuming 80% efficiency, the fully loaded system would draw about 370w.

Kill-A-Watts don't seem to properly measure draws from active PFC units...

 

[thilanliyan]

Originally posted by: s44

Originally posted by: thilan29
Remember this is at the socket, so assuming 80% efficiency, the fully loaded system would draw about 370w.

Kill-A-Watts don't seem to properly measure draws from active PFC units...

Ah okay i didn't know that. Any correction that can be applied? Would the relative numbers for my own comp still be valid?

 

[AzN]

very informative

thanks

 

[Schmide]

Originally posted by: Wreckage

Originally posted by: evolucion8
I'm seeing that in the hothardware review, the HD 4870 1GB consumes less power at load than the GTX 260+. It's the same phenomenom that happened before when the HD 4870 1GB was first launched and used less power compared to the HD 4870 512MB because of improved yields etc, seems that with newer RV770 chips, that keeps happening. But why it doesn't happen with the GTX 260+? Unless if that GTX 260+ card is older, or because since it was designed specifically to the 65nm process (That's why they weren't able to made it to 40nm) it won't see much improvement, like it happened with the 55nm besides overclocking.

The 4870 uses significantly more power at idle than the GTX260.
http://www.techreport.com/articles.x/16504/10

It's basically a tie.

xbitlabs 4870 130w vs. 260 136w

You win some you loose some.

 

[dug777]

Originally posted by: s44

Originally posted by: thilan29
Remember this is at the socket, so assuming 80% efficiency, the fully loaded system would draw about 370w.

Kill-A-Watts don't seem to properly measure draws from active PFC units...

Explain?

 

[evolucion8]

Originally posted by: Schmide

It's basically a tie.

xbitlabs 4870 130w vs. 260 136w

You win some you loose some.

Yeah, your review is more valid because is newer, proves that during time, manufacturing process matures and less leakage will happen, that also is happening with the GTS 250 which consumes less power than the 9800GTX+ and both uses the same old, tired chip.

 

[Wreckage]

Originally posted by: evolucion8

Originally posted by: Schmide

It's basically a tie.

xbitlabs 4870 130w vs. 260 136w

You win some you loose some.

Yeah, your review is more valid because is newer, proves that during time, manufacturing process matures and less leakage will happen, that also is happening with the GTS 250 which consumes less power than the 9800GTX+ and both uses the same old, tired chip.

Really? The review I linked is from March 3rd 2009. His link is from August 19th 2008. I'm guessing you did not even read the links.


Idle power consumption:

GTS250 - 125W
GTX260 - 148W

4850 - 154W
4870 - 175W :shocked:

 

[josh6079]

Originally posted by: Wreckage

Originally posted by: evolucion8
I'm seeing that in the hothardware review, the HD 4870 1GB consumes less power at load than the GTX 260+. It's the same phenomenom that happened before when the HD 4870 1GB was first launched and used less power compared to the HD 4870 512MB because of improved yields etc, seems that with newer RV770 chips, that keeps happening. But why it doesn't happen with the GTX 260+? Unless if that GTX 260+ card is older, or because since it was designed specifically to the 65nm process (That's why they weren't able to made it to 40nm) it won't see much improvement, like it happened with the 55nm besides overclocking.

The 4870 uses significantly more power at idle than the GTX260.
http://www.techreport.com/articles.x/16504/10

Evolucion8 was discussing load power draws.

That said, I haven't seen the 4870 1GB draw less power at either idle or load when comparing it to a GTX 260+.

 

[evolucion8]

Well said , but the GTX 260+ 55nm is a different story

 

[jamesbond007]

I don't have a scientific method of measuring power consumption, but I did install a second HD4870 in my computer last night as well as upgrading my PSU. I had an Antec 550W and replaced it with a PC Power and Cooling 750W (INSANELY AWESOME UNIT!) and boy does that thing rock.

Idling while I type this message, my Killawatt meter reads 342W. When I loaded up Crysis earlier, I was jumping between 580W-640W, depending on what was on the screen and where I was in the game. My system specs are:

ASUS Rampage Formula (X48 Chipset)
Intel Q9450 2.66GHz @ 3.2GHz (QX9770 speed, 12MB L2, 1600MHz FSB, stock 9450 voltage)
8GB G.Skill DDR2 at PC1066 speeds
Two WD 150GB Raptors in RAID0
Two Radeon HD4870 512MB in Crossfire (one ASUS brand and one HIS, both stock cooler designs, clocked to 775/1050, 35% continuous fan speed) 750/900 is default
Windows Vista Ultimate /SP1 64bit Edition

What is really funny is when I was playing Crysis last night, my APC unit started freaking out and gave me a warning message saying that if it switched to battery power, I'd have less than 5 minutes of battery life!!! :laugh: :laugh: :laugh: I'm using an APC 1500VA Back-UPS XS series that puts out up to 865W! My Killawatt meter is hooked up only to my computer tower, so my monitor and other peripherals are not factored into my wattage rating.

 

[thilanliyan]

Originally posted by: jamesbond007
Idling while I type this message, my Killawatt meter reads 342W. When I loaded up Crysis earlier, I was jumping between 580W-640W, depending on what was on the screen and where I was in the game. My system specs are:

:Q

Wow...that power draw is really high...and that's not even with the CPU fully loaded?

 

[jamesbond007]

Originally posted by: thilan29

Originally posted by: jamesbond007
Idling while I type this message, my Killawatt meter reads 342W. When I loaded up Crysis earlier, I was jumping between 580W-640W, depending on what was on the screen and where I was in the game. My system specs are:

:Q

Wow...that power draw is really high...and that's not even with the CPU fully loaded?

That's right - it says 351W at the moment with 0-3% CPU usage as I'm typing. One thing I did notice is that my secondary 4870 is clocked at 775/1050 all the time while my primary seems to be the only one that 'steps' down. Additionally, my CPU seems to always be at 3.2GHz, so I'm going to go check my BIOS settings and see if SpeedStep is disabled or not. CPU-Z says that my core voltage is 1.192V - does that seem about right?

 

[toyota]

Originally posted by: jamesbond007

Originally posted by: thilan29

Originally posted by: jamesbond007
Idling while I type this message, my Killawatt meter reads 342W. When I loaded up Crysis earlier, I was jumping between 580W-640W, depending on what was on the screen and where I was in the game. My system specs are:

:Q

Wow...that power draw is really high...and that's not even with the CPU fully loaded?

That's right - it says 351W at the moment with 0-3% CPU usage as I'm typing. One thing I did notice is that my secondary 4870 is clocked at 775/1050 all the time while my primary seems to be the only one that 'steps' down. Additionally, my CPU seems to always be at 3.2GHz, so I'm going to go check my BIOS settings and see if SpeedStep is disabled or not. CPU-Z says that my core voltage is 1.192V - does that seem about right?

thats crazy. my pc only draws 81 watts from the wall while idle and typically less than 200 while gaming. the most I have ever seen is 236 watts during furmark.

 

[thilanliyan]

Originally posted by: toyota
thats crazy. my pc only draws 81 watts from the wall while idle and typically less than 200
while gaming. the most I have ever seen is 236 watts during furmark.

Less than 200w with a GTX260? That's really good. I think my Phenom is a bit of a power pig.

 

[toyota]

Originally posted by: thilan29

Originally posted by: toyota
thats crazy. my pc only draws 81 watts from the wall while idle and typically less than 200
while gaming. the most I have ever seen is 236 watts during furmark.

Less than 200w with a GTX260? That's really good. I think my Phenom is a bit of a power pig.

yeah just under 190 watts in Crysis benchmark loop. my cpu only has a tdp of 65 watts and I dont think it even gets near that. I have put both cpu and gpu at 100% load and still didnt hit 250 watts.

 

[jamesbond007]

What could I possibly doing wrong? Is there a setting on the Killawatt that needs to be changed? LOL

I rebooted and I'm thinking it was some tinkering I was playing with my 4870 OCs that kept my second card at the OCed values all the time. It is now stepped back down to 500MHz on the core when idling on the desktop. My current meter reading is 350-355W while I'm typing this.

Is there something specific that I can measure to find out what's sucking all the juice? Is it the PC Power and Cooling 750W PSU? My CPU is near idle with 0-2% CPU as always. I just don't get what could be sucking up all the power.

My CPU is not speedstepping as I recall it's disabled in the BIOS, but I cannot get into my BIOS at the moment because my keyboard isn't being recognized or responsive during boot. (USB nor PS/2 keyboards) I've learned in many years of experience to not do something like reset the BIOS at near midnight when I'm sleepy cuz it has always yielded bad results!

 

[Schmide]

Originally posted by: jamesbond007
I don't have a scientific method of measuring power consumption, but I did install a second HD4870 in my computer last night as well as upgrading my PSU. I had an Antec 550W and replaced it with a PC Power and Cooling 750W (INSANELY AWESOME UNIT!) and boy does that thing rock.

Idling while I type this message, my Killawatt meter reads 342W. When I loaded up Crysis earlier, I was jumping between 580W-640W, depending on what was on the screen and where I was in the game. My system specs are:

ASUS Rampage Formula (X48 Chipset)
Intel Q9450 2.66GHz @ 3.2GHz (QX9770 speed, 12MB L2, 1600MHz FSB, stock 9450 voltage)
8GB G.Skill DDR2 at PC1066 speeds
Two WD 150GB Raptors in RAID0
Two Radeon HD4870 512MB in Crossfire (one ASUS brand and one HIS, both stock cooler designs, clocked to 775/1050, 35% continuous fan speed) 750/900 is default
Windows Vista Ultimate /SP1 64bit Edition

What is really funny is when I was playing Crysis last night, my APC unit started freaking out and gave me a warning message saying that if it switched to battery power, I'd have less than 5 minutes of battery life!!! :laugh: :laugh: :laugh: I'm using an APC 1500VA Back-UPS XS series that puts out up to 865W! My Killawatt meter is hooked up only to my computer tower, so my monitor and other peripherals are not factored into my wattage rating.

Have you tried not using your APC? You seem to be about 50-100w more than you should be.

Idle
4870 70w x2 = 140w
q9450 = 33w
x48 = 20w
Drives = 30w
Ram = 10w

total 233w + 25% inefficiency of 58w = 291w

Load
4870 130w x2 = 260w
q9450 = 115w
x48 = 30w
drives = 30w
Ram = 10w

total 445w + 25% inefficiency of 111w = 556w

Sources for power figures xbitlabs above and lostcircuits

Edit: fixed cpu name and added sources.

 

[thilanliyan]

Originally posted by: josh6079
That said, I haven't seen the 4870 1GB draw less power at either idle or load when comparing it to a GTX 260+.

http://www.xbitlabs.com/articl...d-hd-4870_5.html#sect0

Here the 4870 draws less...it's mostly in the 260s favour but not by a large margin.

EDIT: Were you talking about the 55nm 260? If so then yeah the 260 would win every time.

 

[Cookie Monster]

Thing about the HD4870 and rather high idle power draw can be blamed on the use of GDDR5. With the initial implementation of GDDR5, i dont think AMD took advantage of the "power saving" features of the GDDR5. If you can successfully lower its memory frequencies (not sure about voltage), huge gains can be had for the idle power draw. The only problem is that the current boards in my knowledge, they lack the stability when reduced to a certain memory frequency threshold. So AMDs second implementation of this new memory module will probably be much more refined, i.e include these power saving techniques.

 

[jamesbond007]

Originally posted by: Schmide
Have you tried not using your APC? You seem to be about 50-100w more than you should be.

Idle
4870 70w x2 = 140w
q9450 = 33w
x48 = 20w
Drives = 30w
Ram = 10w

total 233w + 25% inefficiency of 58w = 291w

Load
4870 130w x2 = 260w
q9450 = 115w
x48 = 30w
drives = 30w
Ram = 10w

total 445w + 25% inefficiency of 111w = 556w

Sources for power figures xbitlabs above and lostcircuits

Edit: fixed cpu name and added sources.

I use this power supply, which has an 83% efficiency rating, so 25% would be a bit much to estimate with. Additionally, thilan29 said that the Killawatt meters have troubles with active PFC units, which my PSU is. Perhaps that is what's throwing the numbers off a bit? Is there any real way to measure the power draw?

Also, I'm not sure what the difference my APC unit would make since my meter is between the UPS and my tower.