R9 290 crossfire total system power consumption - AX860i results at different overclock speeds (Water Cooled)

[Tam3n]

Hello!

 

I've been doing some overclocking tests with my R9 290s after putting them under water.

 

You can check the water cooling build here: http://linustechtips.com/main/topic/167195-water-cooling-r9-290-crossfire/

 

I got curious about how much wattage am I actually pulling, since I only have 860 Watts PSU for these Hawaii cards.

 

The Corsair AX860i being a digital PSU has a feature to log various system data into Excel datasheets. This time around I concentrated only on the power usage, which will give power in ("from the wall"), power out (what PSU feeds to the components) and efficiency (out / in).

 

 

For all tests I used the Valley benchmark as I've found out it really pushes the system power usage. I ran all tests at the Ultra setting at 2560x1440 with 14.6 beta drivers (not the best for benchmark scores). For overclocking I used MSI After Burner and Sapphire Trixx for its added overclocking ranges.

 

System:

CPU: i7 3820 @4,3 GHz + 16GB 1833 MHz ram

GPU: 2x Gigabyte R9 290 OC WindForce (GV-R929OC-4GD rev 1.0)

MB: P9X79

PSU: Corsair AX860i (230 V in)

 

Before going into power consumption vs clocks comparisons, which are all done on water, lets take a look at water vs air. Before installing the water blocks, I did a Valley run on air at the factory oc settings 1040/1250, and then again after water cooling.

 

As you can see, the power consumption is over 10% higher on air than on water. Heat lessens components efficiency since higher temperatures increase electrical resistance which leads to higher heat output and power consumption. We are talking about 50-40 C difference in temperatures between a water cooled and an air cooled GPU.

 

 

Okay lets start with the stock (reference) R9 290 clocks 947 MHz on the core and memory at 1250 MHz.

As you can see the power in is about 560 Watts and power out 530 Watts.

 

Next would be the Gigabyte factory OC 1040/1250, but you can already see that from the air vs water comparison. Power in is about 650 W and out 600 W.

 

 

Then lets go to 1100/1350, still on stock voltage.

Power consumption doesn't rise much from 1040/1250

 

 

Next is a bigger leap to higher clocks and over voltting. 1200/1600 with +100 mV Vcore and Aux voltage.

With voltage increase and higher clocks the power consumption rises to an another level. Power in ~860 W and out ~780 W.

 

 

And finally at the limits of my PSU and the cards. 1230 MHz on the core and memory at 1650 MHz (6600 MHz effective) and +150 mV voltage increase.

Ouch. Total power consumption is over 900 Watts and PSU is putting out that 860 W and a little more at some point.

 

 

I didn't do any fine tuning on the OCs, just threw settings in, so the voltages could probably have been lower.

 

I hope this was informative for someone. I learned how my cards behave under different clocks.

 

 

 

Fame for the person who redoes these on air

[Freaky_spider]

very interesting read, but would be cool to see a MHz on the x axis and  wattage draw (the average) on the Y axis

 

that would be a cool graph to see! 

[Tam3n]

very interesting read, but would be cool to see a MHz on the x axis and  wattage draw (the average) on the Y axis

 

that would be a cool graph to see! 

 

Hmm yes, good suggestion. I could try different clocks at different voltages. I might do that if I have time

[VSG]

This is why I went with a 1200W PSU with dual 290x. Of course now, that isn't enough for dual KPEs.

[Tam3n]

This is why I went with a 1200W PSU with dual 290x. Of course now, that isn't enough for dual KPEs.

 

Haha yes. If you are doing extreme overclocking, a PSU what normally seems like complete overkill might not even be enough.

 

One funny thing I forgot to mention. When I was doing the highest OC (1230/1650) test, I forgot to rise the power limit to 150% when setting the clocks for the first run. I was wondering why are the clocks fluctuating from 1200 to 1230, I thought I was hitting some limit. And indeed, after the benchmark I checked and lolled. I had forgotten the power limit to default 100%, so the cards were just respecting that and slightly throttling on the heaviest parts of the benchmark. After rising the limit they throttled no more. I had no idea that the default 100% power limit is actually that high, I needed to do 30% OC with +150 mV overvoltage to start hitting that. Of course I'd believe you'd hit that limit sooner on air.

[billwanchalo]

thanks for the great info! i wanna build a pc and was wondering if i could get a mild overclock on a fx-8320,(to 8350's speeds) and have two msi r9 290 overclocked to around your first overclock 1100/1350 on air cooling of the msi twin frozer IV's and not exceed 95 C in a corsair 230t case with all fans blowing and the GPU fan going at 100% and still be good on the power side of things. the build would have a corsair ax860 watt psu ,6 case fans,(3 120mm and 3 140mm), 1 ssd, 1 hdd, 1 dvd reader, a MSI 970 gaming mobo, G. skill sniper low voltage Ram ,(2x4), a XIGMATEK dark knight 2,(with two fans) , plus a wifi card. Sorry for such a long post just wanted to put all the details in it.

[Tacitus]

Thanks a lot for this, I was concerned about running two 290s in crossfire on a decent 750W PSU at near stock speeds but it seems I'll be fine!

[n0x1ous]

thanks for the great info! i wanna build a pc and was wondering if i could get a mild overclock on a fx-8320,(to 8350's speeds) and have two msi r9 290 overclocked to around your first overclock 1100/1350 on air cooling of the msi twin frozer IV's and not exceed 95 C in a corsair 230t case with all fans blowing and the GPU fan going at 100% and still be good on the power side of things. the build would have a corsair ax860 watt psu ,6 case fans,(3 120mm and 3 140mm), 1 ssd, 1 hdd, 1 dvd reader, a MSI 970 gaming mobo, G. skill sniper low voltage Ram ,(2x4), a XIGMATEK dark knight 2,(with two fans) , plus a wifi card. Sorry for such a long post just wanted to put all the details in it.

The FX will be a bottleneck for those GPUs. I speak from experience.

[Sintezza]

The FX will be a bottleneck for those GPUs. I speak from experience.

 

Yes definitely, But basicly every cpu will do, at some certain point.

[Mongoose135]

Thanks for making these charts, they were helpful. Makes me tempted to pick a good 750W psu to save a bit of cash over an 850w. Then again I would like to have the option of adding a 3rd 290 later on when games get more demanding (which according to your results should be doable with a good 850w psu and very light OC)

 

I'm amazed to see the difference between air/water cooling.

 

Unigine Valley doesn't use that much CPU (mainly just the GPUs). Do you think you'd need to leave an extra 50-100W of headroom in case the CPU was also pushed up to 90-100% useage?

[Mongoose135]

thanks for the great info! i wanna build a pc and was wondering if i could get a mild overclock on a fx-8320,(to 8350's speeds) and have two msi r9 290 overclocked to around your first overclock 1100/1350 on air cooling of the msi twin frozer IV's and not exceed 95 C in a corsair 230t case with all fans blowing and the GPU fan going at 100% and still be good on the power side of things. the build would have a corsair ax860 watt psu ,6 case fans,(3 120mm and 3 140mm), 1 ssd, 1 hdd, 1 dvd reader, a MSI 970 gaming mobo, G. skill sniper low voltage Ram ,(2x4), a XIGMATEK dark knight 2,(with two fans) , plus a wifi card. Sorry for such a long post just wanted to put all the details in it.

 

Here's what I'd estimate:

 

1. GPU power consumption. Going by the OP's results, the windforce card pulled up to about 80w extra at 1040Hz, maxing out at about 680w. Cranking the card up to 1100Hz would add an extra 10-20w on to that. For 2 windforce cards you could expect 700w power consumption under air cooling.

 

Going by kitguru's reviews of the windforce and the MSI, the two cards pull almost exactly the same power, HOWEVER the gigabyte comes with a stock clock of 1040MHz, whereas the MSI stock speed is 977MHz. If we assume their power figures were both at stock speed, you can expect the MSI to consume more power when they're running at the same speed. Add on 50W to be safe (up to 750W requirement)

 

2. CPU power consumption. Now the test system uses an i7 3820. The difference in power consumption between that CPU and an fx-8150 is *about* 30W at medium load (very rough estimate based on 89/184W and 104/229W load/idle figures for a full pc system using the i7 3820 and fx8150 respectively, from guru3d). With a mild overclock, let's round it up another 50W to be safe.

 

- So in total you should be just fine power supply wise (my conservative estimate adds up to 800W full system useage).

 

3. Temperature wise, you're going to have issues. Taking into account the motherboard you're using and the case/fans, I see a potential issue in that there is no side fan on the 230T to blow directly between the graphics cards. Check out https://forum-en.msi.com/index.php?topic=180405.0 this guy has a similar set up to what you're considering, and is reaching 94 degrees on the top card at 1085/1350. The MSI cards run hotter than most other r9 290s (non-ref). If you had a case with a side window and 2 fans blowing on the gpus, it'd be a slightly different matter. Either way it's going to be pretty damn loud with all those fans blowing at 100%. It would probably work, and the cards would probably survive, but the VRM temps would probably hit 100+ degrees on the top card, so I wouldn't want to use this setup myself for any extended period of time.

 

I'd recommend attaching an NZXT G10 + liquid cooler (e.g. corsair h55) on to the top card, and leave the bottom card as it is. That should safely keep both cool and allow for potentially higher clocks. (would set you back another $100 or so)

[Cheddle]

Images are dead - any chance of a repost?

[SkilledRebuilds]

-Snipped-

 

Hello!

I've been doing some overclocking tests with my R9 290s

For all tests I used the Valley benchmark as I've found out it really pushes the system power usage. I ran all tests at the Ultra setting at 2560x1440 with 14.6 beta drivers (not the best for benchmark scores). For overclocking I used MSI After Burner and Sapphire Trixx for its added overclocking ranges.

 

System:

CPU: i7 3820 @4,3 GHz + 16GB 1833 MHz ram

GPU: 2x Gigabyte R9 290 OC WindForce (GV-R929OC-4GD rev 1.0)

MB: P9X79

PSU: Corsair AX860i (230 V in)

 

 

As you can see, the power consumption is over 10% higher on air than on water. Heat lessens components efficiency since higher temperatures increase electrical resistance which leads to higher heat output and power consumption. We are talking about 50-40 C difference in temperatures between a water cooled and an air cooled GPU.

 

 

Stock (reference) R9 290 clocks 947 MHz on the core and memory at 1250 MHz.

As you can see the power in is about 560 Watts and power out 530 Watts.

 

Gigabyte factory OC 1040/1250,  Power in is about 650 W and out 600 W.

 

 

Loved that reading, informative and clean. I have a query.

 

I have an EVGA SuperNOVA 750Watt 80Plus Bronze Semi-Modular PSU /w 4x 8pin (6+2) from the get go.

More Info - http://www.msy.com.au/sa/elizabeth/pc-components/14162-evga-supernova-750-b2-110-b2-0750-vr-750watt-80plus-bronze-semi-modular-power-supply-unit.html

 

Already having my 1040Mhz stock Windforce card now, I ordered another, however plan on running them @ 947Mhz each until my PSU is 'up to snuff' & upgraded to something with more headroom soon.

Do you think it's wise using them until then...... as going from your info above, it seems pretty reasonable to have no issues.... right?

 

My i5 4690 is non-OC'able however is pushed to 4.05Ghz with higher BCLK 'at the moment' & Ram is 1866Mhz 1.5v. /w 2 SSD's, 2HDD's and 5 Fans (1 x CPU, 2 intakes, 1 under GPU, 1 @ top of Case)<- Needs reworking anyway

Still seems reasonable on power compared to yours base system and testings.

[Tam3n]

Loved that reading, informative and clean. I have a query.

 

I have an EVGA SuperNOVA 750Watt 80Plus Bronze Semi-Modular PSU /w 4x 8pin (6+2) from the get go.

More Info - http://www.msy.com.au/sa/elizabeth/pc-components/14162-evga-supernova-750-b2-110-b2-0750-vr-750watt-80plus-bronze-semi-modular-power-supply-unit.html

 

Already having my 1040Mhz stock Windforce card now, I ordered another, however plan on running them @ 947Mhz each until my PSU is 'up to snuff' & upgraded to something with more headroom soon.

Do you think it's wise using them until then...... as going from your info above, it seems pretty reasonable to have no issues.... right?

 

My i5 4690 is non-OC'able however is pushed to 4.05Ghz with higher BCLK 'at the moment' & Ram is 1866Mhz 1.5v. /w 2 SSD's, 2HDD's and 5 Fans (1 x CPU, 2 intakes, 1 under GPU, 1 @ top of Case)<- Needs reworking anyway

Still seems reasonable on power compared to yours base system and testings.

 

You should be fine as long as you don't do much overclocking. With 860 W I can go ~1220-1240 MHz +150 mV for benchmark runs, but that's already a bit too much for the psu, and can't oc cpu too much. In normal use - medium oc I have no problems. If you want to be sure power will not be an issue (in high oc) I'd advise go 1000 W +

[SkilledRebuilds]

You should be fine as long as you don't do much overclocking. With 860 W I can go ~1220-1240 MHz +150 mV for benchmark runs, but that's already a bit too much for the psu, and can't oc cpu too much. In normal use - medium oc I have no problems. If you want to be sure power will not be an issue (in high oc) I'd advise go 1000 W +

I'd be happy with 2x 950mhz Cores for quite some time and pretty sure @ 950mhz on both it seems I'd be fine.

[sci.fi.man]

wow, i thought 750+ Watt PSU will be enough for 2x R9 290 i'm totally wrong

 

very interesting review by the way. very interesting to watch