Discussion New computer parts as an enviromental disaster?

[Kocicak]

The current broad trend in all technology is moving toward increased efficiency in order to decrease energy consumption, which has various negative effects.

I just saw a rumour that an upcoming Intel CPU will have special performance mode allowing 350W energy draw. It also seems that upcoming AMD CPUs will have higher TDP than the current generation.

What is the sense of this in the point of view of the broad trend I mentioned in the beginning? State regulation in my opinion is sometimes a good thing and at this point it seems that power draw of consumer computers should be limited.

I think that limits of 50W for CPU, 50W for integrated GPU and 150W for discrete GPU would satisfy needs of most people.

Parts with higher power draw could be heavilly taxed so that they would be financially viable only for proffesionals who would use them productively.

I understand that some people may find my idea not beneficial for whatever reasons but what is happening now in this part of PC market is ridiculous and needs to stop.

(and production of virtual currencies should be banned completely as a huge energy waste, but that is a different topic.)

 

[moinmoin]

I have put together, how performance and efficiency at different level of load (number of running computing threads), is affected when the CPU is limited by the total power it can draw.

The improvements in efficiency of the CPU while running heavy load are impressive. Unfortunatelly poor efficiency of the CPU remains poor at lower load.

Limiting power is useful to match available cooling solution or while running this CPU at maximal load to significantly improve its efficiency.

Unfortunatelly it cannot improve overall efficiency of the CPU, which most of the time will run at below maximal load.

A positive takeaway is that the performance of the CPU below maximal load is unaffected or just slightly affected by lowering the maximal energy draw.

View attachment 69247

I mentioned it in relation to 7950X in another thread: To achieve better per thread efficiency one likely needs to use fixed frequency. Depending on how temperature is measured a low-ish temperature limit may hinder single thread boosting past their efficiency infliction point as well.

 

[yottabit]

The average electric vehicle requires 30kWh to travel 100 miles. If my PC has a 1000W constant power draw (which most do not) I can run it for 30 hours and use the same amount of energy. Now where is the problem?

To other points you brought up. Most consumer PCs are likely laptops and use less power than what you list. Most business PCs are low end desktop models or laptops that use less power than what you list. The CPU you mention is an enthusiast or professional level processor that the vast majority of people aren't going to buy anyway.

BTW, are you also not the person "got lured from AMD to Intel with a high single thread performance" and are also overclocking your RAM in another thread?

That’s a ridiculous amount of energy though. You think it seems appropriate that blinking some lights on a screen for an hour should take the same amount of energy as driving 3.33 miles?

I’ll agree somewhat to the OP’s premise, but only in the context of the latest gen of CPU and GPU. It would be nice to see some sort of “gentleman’s agreement” like Japan had for their sports cars limiting stock power to 286 bhp.

So maybe everyone could agree that stock should ship at 120-140 W CPU, 275 W GPU or so. And similar to those tuner cars, Intel/AMD are free to day “definitely son’t pull that boost limiter jumper pin”

I think having the stock power limits so high is somewhat a negative for the consumer and environment. I’d be fine even if they want to advertise the PBO performance, but why not ship it in a more reasonable state?

Talking about cases where you can keep 95% of the performance at 70% of the power… that sounds like a factory overclock, not an efficient chip design

EDIT: meant to add that in my hypothetical scenario, HEDT and Server would be exempt of course

I think another great way to approach the power limit would be: two fully loaded desktop PC running off one 120 v 15 A wall socket. But I suppose it is all arbitrary, and as long as they don’t sacrifice idle performance and expose control for undervolting / limiting total power, I won’t complain (too much)

 

[Kocicak]

I played with Intel 13900K (power limited by 180W to run under an air cooler), this time I tried more underclocking than overclocking. I did not spend any time with optimising the offset, I randomly chose some numbers and they worked. Undervolting and underclocking the CPU brings some very interesting efficiency gains across the whole load intensity spectrum.