Energy efficiency of gaming computers

[Bodhitree]

I came across this site featuring a couple of studies done by a father and son team about the energy efficiency of gaming computers. Apparently the son wanted to get into pc gaming and his engineer father agreed to bankroll him on the condition that he could measure every detail of the power consumption of the setup. Fabulous story.

The surprising thing is that it is possible to carefully choose components to create a much greener computer than the standard gaming pc. They reduced the PCs load by nearly half, although the claim of leaving performance largely unaltered is something you might want to take with a grain of salt.

I thought it might be interesting to compare with M1 machines. Anyway here is the article

Greening the Beast - Taming-the-Energy-Use-of-Gaming-Computers

Taming the Energy Use of Gaming Computers Nathaniel Mills GreeningTheBeast.org nathanielmills50@gmail.com Evan Mills Lawrence Berkeley National Laboratory evanmills.lbl.gov emills@lbl.gov Published in the journal Energy Efficiency, this new study presents a novel analysis of the energy use of

sites.google.com

 

[diamond.g]

I came across this site featuring a couple of studies done by a father and son team about the energy efficiency of gaming computers. Apparently the son wanted to get into pc gaming and his engineer father agreed to bankroll him on the condition that he could measure every detail of the power consumption of the setup. Fabulous story.

The surprising thing is that it is possible to carefully choose components to create a much greener computer than the standard gaming pc. They reduced the PCs load by nearly half, although the claim of leaving performance largely unaltered is something you might want to take with a grain of salt.

I thought it might be interesting to compare with M1 machines. Anyway here is the article

Greening the Beast - Taming-the-Energy-Use-of-Gaming-Computers

Taming the Energy Use of Gaming Computers Nathaniel Mills GreeningTheBeast.org nathanielmills50@gmail.com Evan Mills Lawrence Berkeley National Laboratory evanmills.lbl.gov emills@lbl.gov Published in the journal Energy Efficiency, this new study presents a novel analysis of the energy use of

sites.google.com

Carefully underclocking and undervolting makes a huge difference to power draw (and thus heat). If you are not trying to post big numbers in 3dmark most folk can get by just fine on a good 550 watt power supply with a 5950 and a RTX3090.

 

[velocityg4]

When I did my last one. I just looked up tests for idle power usage of the components, along with gaming loads to make my choices. I then fiddled with the vcore to lower below the default as much as possible. Without affecting turbo boost or stability.

For monitors I also factored in power consumption ratings. Going LED backlit was a big improvement over the old cfl backlit monitor.

Although power consumption was a concern. I was more interested in reducing heat output. Thus reducing fan noise.

Given current component availability. I didn’t have the luxury for my latest build. Just had to get what I could within my budget. Then tweak it as well as I could. Although I expect it’s a better performance per watt. Than my old build. Just because that’s improved in general over the last eight years. Although total watts is definitely higher under load. Idle should be better.

 

[diamond.g]

My current Windows rig has a 850W power supply which is overkill for my 1st gen Threadripper and a 6900XT. This weekend I plan to swap over to a 5900X and watercool the whole thing. At most I have pulled ~500 watts of power with a slight overclock on the GPU/CPU but cooling seems to be my limitation (fans are loud, lol).

 

[velocityg4]

My current Windows rig has a 850W power supply which is overkill for my 1st gen Threadripper and a 6900XT. This weekend I plan to swap over to a 5900X and watercool the whole thing. At most I have pulled ~500 watts of power with a slight overclock on the GPU/CPU but cooling seems to be my limitation (fans are loud, lol).

Can't do much with the GPU unless you try tweaking the VCORE or clock rates. Outside of a custom loop. Unless you aren't getting enough cool air in the case.

For general noise. Better case fans, more case fans and CPU coolers are the way to go. To get noise down. Since noise is most directly related to RPM. One fan moving really fast is louder than four fans moving slow.

 

[diamond.g]

Can't do much with the GPU unless you try tweaking the VCORE or clock rates. Outside of a custom loop. Unless you aren't getting enough cool air in the case.

For general noise. Better case fans, more case fans and CPU coolers are the way to go. To get noise down. Since noise is most directly related to RPM. One fan moving really fast is louder than four fans moving slow.

Yeah that is why I am going with water cooling this weekend. The last parts showed up today. My Focus G case has somewhat poor airflow, so water cooling should see an improvement over air.

My Threadripper is using a Thermalright SilverArrow TR4 and will make your ears bleed when the CPU gets hot. The 5900X is an upgrade, lol.

 

[Bodhitree]

The new 24” iMac M1 does everything with a 143 Watt power supply. Not bad when you consider that that includes the monitor, speakers, microphones as well as the computer itself.

 

[LordVic]

Carefully underclocking and undervolting makes a huge difference to power draw (and thus heat). If you are not trying to post big numbers in 3dmark most folk can get by just fine on a good 550 watt power supply with a 5950 and a RTX3090.

yes, but how the heck am I supposed to be top 1% of users with similar builds as me if I'm not pumping through as much voltages through my components as possible

 

[LordVic]

The new 24” iMac M1 does everything with a 143 Watt power supply. Not bad when you consider that that includes the monitor, speakers, microphones as well as the computer itself.

it's not even remotely compariable.

the gpu in the M1 is not even comparable to the actual performance and capabilities in full blown discreet GPU's. your M1 GPU is not going to come close to the visual fidelity, or even CUDA / AI performance out of a RTX 3080.

yes, there's a massive difference in power consumption and heat generation, and if you're ok with a lower end gaming experience (which is absolutely fine for many), that's fine. But in no way is an M1 iMac comparable to a custom built gaming desktop for raw gaming performance.

 

[LordVic]

Yeah that is why I am going with water cooling this weekend. The last parts showed up today. My Focus G case has somewhat poor airflow, so water cooling should see an improvement over air.

My Threadripper is using a Thermalright SilverArrow TR4 and will make your ears bleed when the CPU gets hot. The 5900X is an upgrade, lol.

Be careful with this assumption. Watercooling is still heavily dependant on airflow. Different requirements than air cooling, but you still need to be able to get air efficiently past the radiators.

Looking at the Fractal G, it should be fine if you put the radiator on the top position. 2 fans in the front for intake, and a 240mm rad up top with 2x120's in push should do it.

as long as you can get good cool airflow yo'll be fine.

I did a custom water kit with 2 radiators, my CPU and GPU and the largest EKWB pump, all in an NZXT H510i case. The solid front definitely impacts airflow, requiring a slightly higher fan curve than i'd like to get reasonable idle temps.

Though, with my Overclocks in place, "reasonable" temperatures were thrown out a while ago. However, I'm able to hit top 5% in 3dmark for my hardware combination. GPU average temps around 72c under load. but hotspot still up at 100ish. I'm only going further with my OC if I can somehow get lower than ambient temps at this pint.

However, yes, it'll still be better than a fan at 100% on aircooling for a TR4

 

[diamond.g]

Be careful with this assumption. Watercooling is still heavily dependant on airflow. Different requirements than air cooling, but you still need to be able to get air efficiently past the radiators.

Looking at the Fractal G, it should be fine if you put the radiator on the top position. 2 fans in the front for intake, and a 240mm rad up top with 2x120's in push should do it.

as long as you can get good cool airflow yo'll be fine.

I did a custom water kit with 2 radiators, my CPU and GPU and the largest EKWB pump, all in an NZXT H510i case. The solid front definitely impacts airflow, requiring a slightly higher fan curve than i'd like to get reasonable idle temps.

Though, with my Overclocks in place, "reasonable" temperatures were thrown out a while ago. However, I'm able to hit top 5% in 3dmark for my hardware combination. GPU average temps around 72c under load. but hotspot still up at 100ish. I'm only going further with my OC if I can somehow get lower than ambient temps at this pint.

However, yes, it'll still be better than a fan at 100% on aircooling for a TR4

I have a radiator up top, that i need to buy 2 120x15mm fans for as the 120x25mm fans I have are too thick, so I am just using 1 fan. I also have a radiator up front with 2 140mm fans (and a 280 radiator). My temps have been pretty awesome so far. While boredom mining ETH GPU hotspot max 61C, MemJ temps max 64.5C. The CPU tctl max hits 91C under Furmark CPU burn (also while mining ETH on GPU).

My loop is pump/res combo -> GPU -> Upper Rad -> CPU -> Front Rad -> pump/res combo.

I am fairly happy with the results so far.

 

[Bodhitree]

it's not even remotely compariable.

the gpu in the M1 is not even comparable to the actual performance and capabilities in full blown discreet GPU's. your M1 GPU is not going to come close to the visual fidelity, or even CUDA / AI performance out of a RTX 3080.

yes, there's a massive difference in power consumption and heat generation, and if you're ok with a lower end gaming experience (which is absolutely fine for many), that's fine. But in no way is an M1 iMac comparable to a custom built gaming desktop for raw gaming performance.

Perhaps if you’re building a computer primarily for gaming, then you might be right. But I think that for most average consumers the M1 is pretty capable. Even so, I think that as a general power consumption figure the 24” iMac is very efficient at what it does.

 

[Irishman]

Carefully underclocking and undervolting makes a huge difference to power draw (and thus heat). If you are not trying to post big numbers in 3dmark most folk can get by just fine on a good 550 watt power supply with a 5950 and a RTX3090.

Okay, but if you’re not trying to post big numbers in 3dmark, then I would question whether or not you need a 550watt PSU, a 5950 and a RTX3090. By noone’s measure would the components in your example be considered low-power, much less low-cost.

 

[Irishman]

it's not even remotely compariable.

the gpu in the M1 is not even comparable to the actual performance and capabilities in full blown discreet GPU's. your M1 GPU is not going to come close to the visual fidelity, or even CUDA / AI performance out of a RTX 3080.

yes, there's a massive difference in power consumption and heat generation, and if you're ok with a lower end gaming experience (which is absolutely fine for many), that's fine. But in no way is an M1 iMac comparable to a custom built gaming desktop for raw gaming performance.

But, it is more comparable to UMAs used in the PS5 and the XBOX series X, which is what Apple seems to be using as a guidepost in developing their own SoC.
In essence, they can achieve surprisingly high performance and surprisingly low TDP on the 24” iMac that others have mentioned, a 4.5K monitor built-in whose performance doesn’t suffer when compared to the other M1 Macs (even those with smaller, 1080p screens, or with no screen at all).

Have you had any experience gaming on an M1 Mac?

 

[diamond.g]

Okay, but if you’re not trying to post big numbers in 3dmark, then I would question whether or not you need a 550watt PSU, a 5950 and a RTX3090. By noone’s measure would the components in your example be considered low-power, much less low-cost.

True a 5600 is probably a better CPU choice. For turning RT on in games (that support it) you need all the performance you can get (yes DLSS helps).

 

[Irishman]

True a 5600 is probably a better CPU choice. For turning RT on in games (that support it) you need all the performance you can get (yes DLSS helps).

Yes, but, again, I don’t remember reading in the OP that he was expressing a need for RTX On in games. Remember, too, that RTX is not the only Ray Tracing game in town. If you ignore the PC market, AMD owns the vast majority of Ray Tracing in consoles, and consoles are a great model for Apple to follow if they are going the UMA route.

 

[diamond.g]

Yes, but, again, I don’t remember reading in the OP that he was expressing a need for RTX On in games. Remember, too, that RTX is not the only Ray Tracing game in town. If you ignore the PC market, AMD owns the vast majority of Ray Tracing in consoles, and consoles are a great model for Apple to follow if they are going the UMA route.

I used the 3090 in my example, personally I ahve a 6900XT, but I still see a hit in games that use DXR, more so than folks running 30-Series cards would. Even if RT performance is built around the consoles, RT on 30-Series cards is just plain faster. It think a lot of it has to do with the tradeoff of having RT Cores inline with the ALU's instead of a separate (independent) block.