1,200W

[Rob Williams]

The other night, Kayden mentioned that someone asked him if a GTX 590 could be paired up with a GTX 580 (for tri-GPU), and after a bit of laughter, an idea came to mind. While such a configuration isn't possible with NVIDIA, it is possible with AMD.

So, I decided to see how many fuses I could blow and installed a dual-GPU HD 6990 and dual HD 6970s. That's right... quad-GPU.

Ahem, excuse the dust:



Idle:



Load:



Given that the PSU in use is Corsair's HX1000 (1000W), I think my heart skipped a beat when I first saw this. But, thanks to Jamie and his mathematical excellence, he said that the PSU in fact can peak to 1,200W. With that knowledge in hand, I decided to do a couple of quick tests before disassembling that configuration.

3DMark Vantage Extreme:



Metro 2033:

2560x1600: DX11, Very High, 0xAA - 45 Minimum / 65.436 Maximum
1920x1080: DX11, Very High, 0xAA - 66 Minimum / 124.659 Maximum

By comparison, a single GTX 580 scores 12750 GPU in 3DMark Vantage and 81 FPS on average in Metro 2033 with Medium detail and DX10 mode. A single GTX 580 retails for about $500, while an HD 6990+6970+6970 setup would be about $1,400.

Now if you'll excuse me, I need some ice.

 

[OriginalJoeCool]

Nah, everybody knows real gamers have five GPUs.

 

[marfig]

LOL! Couldn't think of a better name to that voltage monitor in these circumstances.

Do you realize you just beat the top Extreme score on 3DMark, by a little more than 2.5x?

 

[Tharic-Nar]

For those interested and don't believe the PSU can handle it, check the specifications on Corsair's site, it says 500 watts under each channel, but do some maths (12v x 40amp = 480 watts + 5v x 30amp = 150 watts)... it's 630 watts on one channel, 580 on the other, plus miscellaneous, it's about 1250 watts of power. When you take the 80% efficiency rating into consideration... 1250 x 0.80 = 1000 watts of pure DC power.

I'm just amazed at the idle power... anyway, at least rob now has a way to stress test PSUs.... ;D

@marfig
Not quite, that was the 3DMark 11 score, Rob posted Vantage, the top 20 is a little more scary at 58K... http://www.3dmark.com/hall-of-fame/3dmark-vantage-users-top-20-extreme-preset/

 

[marfig]

>> Not quite, that was the 3DMark 11 score, Rob posted Vantage, the top 20 is a little more scary at 58K... http://www.3dmark.com/hall-of-fame/3dmark-vantage-users-top-20-extreme-preset/

Oh right!
Thanks for the correction.

 

[RainMotorsports]

For those interested and don't believe the PSU can handle it, check the specifications on Corsair's site, it says 500 watts under each channel, but do some maths (12v x 40amp = 480 watts + 5v x 30amp = 150 watts)... it's 630 watts on one channel, 580 on the other, plus miscellaneous, it's about 1250 watts of power. When you take the 80% efficiency rating into consideration... 1250 x 0.80 = 1000 watts of pure DC power.

All correct but im not sure why this hasn't been said. The PSU was not delivering more than its rated power, that we know of anyways.

With the approx 80% efficiency the power supply is drawing more power than it is putting out. Which is entirely normal. While getting close to the edge and never comfortable it was probably fine.

Best I remember PSU's are advertised at how much power they are capable on output not expected consumption at peak output. Which is little to their benefit in the way that advertising normally works theirs no foolery here.

 

[TheCrimsonStar]

 

[RainMotorsports]

Not sure what was up with that but yes, computer plugged into kill a watt, plugged into wall. Measuring consumption not output. The PSU's output will be about 80% of its consumption.

PSU was outputting about 930 watts on all rails total. Within its rated 1000 watts.

The Kill a Watt is a tool for measuring total power consumption from an AC device. At 1000 watts DC (all 12v 5v 3.3v rails being maxed) as stated by Tharic-Nar would have the Kill a Watt reading about 1250 watts.

 

[Kayden]

Surprisingly I managed to get pretty close to that. I had to turn off my OC on my CPU because of heat in my room and the Mojave isn't known for cool summers (c;

Anyways here is what I got.

 

[Kougar]

Now I know why you complain your house is always so hot...

 

[Kayden]

Now I know why you complain your house is always so hot...

Being in a room in the garage with no direct vent from AC doesn't help either. Tried to fix that but the small AC unit I got from storage was broke now I have to get a new one and until I can keep my room temp below 87f I don't want to run the OC cause at 93f in here I get BSOD while playing games, best to play it safe and keep my stuff running that's for sure.

 

[RainMotorsports]

Being in a room in the garage with no direct vent from AC doesn't help either. Tried to fix that but the small AC unit I got from storage was broke now I have to get a new one and until I can keep my room temp below 87f I don't want to run the OC cause at 93f in here I get BSOD while playing games, best to play it safe and keep my stuff running that's for sure.

93C BSOD's... crap you desktop guys have it rough lol. My 9800M GS running GTS clocks are stock voltage stable until about 90C and that why I have to run GTS Voltage on my particular sample. Because in a laptop thats a real temp. I dont keep tabs on it anymore but prior to the last repaste 93C was normal for GRiD max'd out. Desktop in the works still polishing out the case.

 

[Kayden]

93C BSOD's... crap you desktop guys have it rough lol. My 9800M GS running GTS clocks are stock voltage stable until about 90C and that why I have to run GTS Voltage on my particular sample. Because in a laptop thats a real temp. I dont keep tabs on it anymore but prior to the last repaste 93C was normal for GRiD max'd out. Desktop in the works still polishing out the case.

Well that's my room temp when it BSOD, my CPU temp even with water cooling it was at 54c about 130f but I don't think that was the problem, I believe it was ram because they are air cooled and a little OC past their max of 1600 which is where I really hit my wall with my OC, I just need to get my memory, NB and Vreg liquid cooled then my rough spots will be over. (c;

That temp is about what I expect to see from a laptop tbh. What I hate is that Gaming laptops have very low life spans and most gamers don't think or can't afford the extended warranty's.

Good luck on the desktop, don't forget there's a thread to post your rig photos here

http://forums.techgage.com/showthread.php?t=1750

 

[RainMotorsports]

Thanks Ill be sure to post.

Yeah im going into the 3rd year of my G50VTX5 which has been cpu swapped for a T9600 and overclocked to 3.3 Ghz on a daily basis and the gpu was bios flashed. I voided the warranty week one and it hasn't failed me yet. Battery still holds an hour plus charge but its showing signs of complete failure soon. Mainly because in building the desktop I was doing 1.5+ hour runs searching for parts on the couch lol. Took me about 48 hours of battery conditioning to get it to last more than 30 minutes. Somewhere between wear and just the controller needing a good reset.

 

[GFreeman]

My.. goodness.. Close to 1200 watts that is quite the lot! That is a beast! I am surprised to see that 1000watt PSU cope so well with that kind of punishment really..

 

[Kayden]

I just did another benchmark with my OC turned back on and that got me much closer to what Rob got, I also disabled Aero this time.

My.. goodness.. Close to 1200 watts that is quite the lot! That is a beast! I am surprised to see that 1000watt PSU cope so well with that kind of punishment really..

That isn't something I would want to push on that PSU for very long that's for sure. /c:

 

[GFreeman]

I hear ya! I might say "POOF" to some point haha! That's a astonishing result there.. This got me thinking.. I wondered if it was the lack of Catalyst support running all these cards together or would that I7 make up for it compared to that AMD setup.. That second guess might be it I reckon.

 

[CorsairGeorge]

View attachment 1235

Given that the PSU in use is Corsair's HX1000 (1000W), I think my heart skipped a beat when I first saw this. But, thanks to Jamie and his mathematical excellence, he said that the PSU in fact can peak to 1,200W.

Hah, nice! But like some others pointed out, that's measuring input from the wall, not PSU output.

Here's the efficiency curve of the HX1000W. Notice at 100% load, it's about 83% efficient?

Just for fun, let's take the 1163W input and multiply times .83. That'd give us about 965W output, so let's round that up to 97% load on the 1000W PSU. This is only a ballpark number, since you really have to do it the other way. (figure out the output, then the percentage of capacity, then look at the input - and do the math that way. But for our purposes, it's pretty close)

That's pretty impressive, to say the least. Some statistics.

1163W - 965W = ~198W of heat being generated by the PSU alone. That's enough to run most low-end to mid-range gaming PCs at load by themselves.

So assuming that PC is drawing 965W at full load, here's some interesting statistics.

The HX1000W would be around 83% efficient, drawing 1163W from the wall, and producing 198W of heat.
The HX1050W would be around 88% efficient, drawing about 1097W from the wall, and producing only 132W of heat.
The AX1200W would be around 90% efficient, drawing about 1073W from the wall and producing only 108W of heat.

The HX1000W and HX1050W would be running their fans at full load, since they are near 100% utilization. That's 44dBA for the HX1000 and 35dBA for the HX1050. But the AX1200 would only be around 3/4 of its full utilization and so would only be running its fan at about 2/3 speed, or 27dBA.

Of course, at that point, the GPUs are going to be louder than any PSU anyway, so it doesn't really matter.

 

[Brett Thomas]

Thanks for the technical reply, George!

I knew something was a little wonky with the earlier maths because PSUs get more efficient under high load than they do at low loads (part of why people shouldn't put a 1000w PSU in a 200w system)...but it's great to see the actual efficiency graph. It's also really interesting to note just how much energy is getting completely lost to heat!

When Rob told me how much the Kill-a-Watt read, I begged him to go video his power meter...I bet that thing was practically spinning...

 

[CorsairGeorge]

Thanks for the technical reply, George!

I knew something was a little wonky with the earlier maths because PSUs get more efficient under high load than they do at low loads (part of why people shouldn't put a 1000w PSU in a 200w system)...but it's great to see the actual efficiency graph. It's also really interesting to note just how much energy is getting completely lost to heat!

When Rob told me how much the Kill-a-Watt read, I begged him to go video his power meter...I bet that thing was practically spinning...

Haha, no joke!

Somebody asked me on another forum once - "Well, if I get a more efficient PSU, will my room get colder? Because I use my system to heat my room in the winter."

I laughed, I'd never thought of that. It works pretty well as a heater.

The answer, of course, is no, not really. I mean, the output of 965W in Rob's test rig above is STILL being generated as heat by the other components in the case. GPUs, CPU, hard drives, etc, are still pumping the heat out. And their coolers are pumping it into the air anyway. So while the 90W difference between the HX1000 and the AX1200 above wouldn't be converted to heat, you'd still have over 1000W of heat being generated in total in the room.

Better hope your circuit breaker doesn't have too much other stuff on it - triple 30" LCDs and a couple NAS boxes, you'd trip the circuit breaker pretty quick.