Proposal for CPUs/graphic cards for PCs power draw standardisation

[BoggledBeagle]

The current situation is a mess:

• CPUs are marked with power draw numbers which do not mean real power draw.
• Real power draw of the CPU depends on the settings each motherboard manufacturer chose, the same processor can draw different amounts of power on boards from different manufacturers. You may be OK with a larger air cooler on one board and require water cooling on another.
• CPUs are sold with absurd power draw levels which have po place in personal computers.

Why not to standardise a few power levels, which would:

• represent real power draw
• enable customers so choose appropriate cooler
• indicate clearly the performance level of the CPU

Motherboard manufacturers would of course need to ensure that the CPUs draw the proper level of power.

I would propose the levels, which are factor of 1,6 apart from each other: 25?, 40, 65, 100 and 160W

25W should be enough for an office PC and multimedia playback

40W may enable the 25W parts to get onboard a more powerful integrated graphic card or other stuff as an AI accelerator, etc. I am not sure if having 25W and 40W separately is necessary.

65W is what a little "pancake" cooler with no heatpipes can handle and is plenty for any office or universal family computer, should be able to serve lower power graphic cards.

100W can offer a serious amount of computing power already and should be enough for gaming even with the most demanding graphic cards, it is coolable with a very small 2 heatpipe cooler

160W can provide huge amount of computing power, if somebody really needs more than that he probably would be better off buying a proper workstation. It can be cooled with a small 4 heatpipe cooler.

There are three more benefits to this:

This may greatly reduce CPU lineups, force manufacturers to optimise and come up with the best what they can develop for the given level of CPU power.

A producer may sell:
one 40W CPU, the best universal CPU they can make
one 65W CPU, the best universal CPU they can make
two 100W CPUs, the best universal CPU they can make and the best gaming CPU they can make
two 160W CPUs, the best universal CPU they can make and the best "home workstation" CPU they can make, with some specialised accelerators helping with productive workloads etc.

these are 6 CPUs that would serve the PC market. Does anybody really need more???

A third additional benefit would be that it would greatly simplify comparing performance of CPUs from different manufacturers.

Is there even a place of overclocking in this scheme? I doubt it. What is the point of allowing the 65W CPU to be overclocked e.g. to 100W, when you can get a native 100W CPU performing much better than the overclocked CPU? Only the 160W parts may be enabled to run at a maximal power draw of 200W at the cost of loosing energy efficiency. 200W could be a hard cap for PC CPUs. To anybody being offended to be limited by this cap I would say that having a CPU on a personal computer drawing e.g. 350W is OBJECTIVELY INSANE. Then you realise that this whole scheme is built on energy efficiency, increasing the power limit and overclocking really do not make any sense anymore.

The different power (and performance) levels could be clearly colour coded or marked with some performance number. Coolers needed for this level of perfomance could be marked the same way.

The energy efficiency is built in this scheme. The manufacturer needs to bring max performance for a given power level. It is a definition of efficiency. This is the best thing ever for enviroment and for consumers too, because manufacturers would be FORCED HARD to optimise and bring to the customer the best they can and also make it on the best technology they can use. No lazy-ass using old stuff over and over.

A similar scheme should be emloyed for graphic cards, forcing manufacturers to make the best say 100, 160 and 250W cards they can.

Discuss and share this to make this happen. Thanks!

I think I will write a letter to Europen commision. This stuff is green as a frog, benefits consumers greatly and pushes research and technology development hard.

 

[Bill_Bright]

Sorry, but in reality, nothing you said is either true, or makes any sense, from a technical standpoint, or otherwise.

• CPUs are marked with power draw numbers which do not mean real power draw.
• Real power draw of the CPU depends on the settings each motherboard manufacturer chose, the same processor can draw different amounts of power on boards from different manufacturers. You may be OK with a larger air cooler on one board and require water cooling on another.
• CPUs are sold with absurd power draw levels which have po place in personal computers.

1. Yes it does mean real power draw. But you must remember, power demands can go from minimal at idle to maximum in just a few clock cycles. And there typically are at least 3 billion clock cycles per second. Manufactures must go by the maximum one might see.
2. No its not. The motherboard makers do not "choose" this. Power demands depend on many factors beyond the control of the board makers. This includes the number of RAM sticks and amount of RAM, drives, graphics tasks the CPU must perform, I/O ops, the OS, and most significantly, the tasks the user is performing.

The board has absolutely nothing to do with a CPU needing air or water cooling. First and foremost is case cooling, then the tasks the user is performing. Yes, clocking and voltage "options" may be determined by the board maker but the board maker does NOT know which CPU the user is mounting on the board. In a properly cooled case and with default settings, ANY commercially available CPU can adequately be cooled with air cooling, in a properly cooled case, in typical "normal" ambient (room temperature) environments.

3. It is NOT your decision what is appropriate, or not for personal computers. In fact, this comment makes the least sense at all! Sorry, but this just illustrates that you don't really understand what a "personal computer" even is. THAT IS NOT A CRITICISM, but simple observation. People use personal computers for all sorts of tasks, including extremely demanding CAD/CAE, development, compiling, editing and more.

Why not to standardise a few power levels

There are standards - just none that you seem to understand. Voltages are standard. And there are current capability standards. Plug those values into Ohm's Law formulas, Yadda, Yadda and there you have power level standards.

manufacturers would be FORCED
A producer may sell:

What? You want to "dictate" and "force" what manufactures can and cannot produce? You want to limit how manufacturers compete? Total nonsense! This is not communist North Korea or China, or Russia. Or fascist Nazi Germany or Iran.

To anybody being offended to be limited by this cap I would say that having a CPU on a personal computer drawing e.g. 350W is OBJECTIVELY INSANE.

No. Sorry. It is not them being objectively insane here.

Have a good day and Merry Christmas.

 

[Macro Device]

The board has absolutely nothing to do with a CPU needing air or water cooling. First and foremost is case cooling, then the tasks the user is performing. Yes, clocking and voltage "options" may be determined by the board maker but the board maker does NOT know which CPU the user is mounting on the board. In a properly cooled case and with default settings, ANY commercially available CPU can adequately be cooled with air cooling, in a properly cooled case, in typical "normal" ambient (room temperature) environments.

That got outdated forever ago. Many cases of LGA1200 and LGA1700 motherboards providing excessive voltage and frying the snot outta CPUs. Some Ryzen X3D CPUs got roasted by incorrect voltage settings as well.

1. Yes it does mean real power draw.

This also became outdated the day Intel invented Turbo Boost technology. This way the "125 W" 11700K consumes more than double this amount running at "stock" (= user didn't manually OC the CPU). Marketed TDP is a meaningless number. What makes it more meaningless is that different CPUs have different architectures and different heat dissipation approaches (some use thermal paste, some IHSes are soldered, some CPUs are monolythic, and some are clustered) which ultimately leads to them have major temperature difference despite consuming identical power.

This is not communist North Korea or China, or Russia

Russia is not communist since 1991.

100W can offer a serious amount of computing power already and should be enough for gaming even with the most demanding graphic cards

Giggled at this point. CPU power draw in gaming depends on the game you're playing and the resolution of your choice. Cyber athletic 720p gaming doesn't demand anything spectacular from a GPU, yet you are better off buying the fastest CPU on the market to ensure minimum input lag and best precision. Triple-A games at 4K are so GPU limited even if you pair your CPU with a 4090 that CPU is forced to wait and thus, it ends up drawing little power, sometimes below 20 W.

Cooling, once again, is now divided into four major categories:

• CPUs of monolythic layout with a soldered IHS or liquid metal under the IHS. This way you don't have to overthink, it's the most straightforward layout. Given you don't have ugly ventilation situation inside your PC and don't live in sauna you will be fine if you pick a cooler rated a tad higher than your CPU consumes at peak load.
• CPUs of monolythic layout with a thermal paste between the die and the IHS. Thermal pastes differ in quality and are overall worse than liquid metal/solder and this leads to the "no cooler can cool it" problem at some point. Since it's not a constant how good or bad this thermal paste is you have no way to find out the appropriate cooling other than tiral and error. This category is a subject to obsoletion though since more and more Intel SKUs get soldered.
• CPUs of chiplet design. Thanks to being chiplet designed they are cheaper in manufacturing and easier to build but they're effectively a nightmare of air coolers. Especially of those with direct heatpipe contact. This is currently a Ryzen thing (if we talk common desktop PCs, no idea what's going on in more advanced fields) and a cooler totally capable of keeping a 250 W i9-11900K cool can very much be happy to fail to cool a Ryzen that consumes half that amount only due to this geometrical quirk of the latter.
• CPUs without an IHS. Common in laptops. Even more straightforward than the first category. All modern GPUs also belong to this category since Kepler (GTX 600 series, 2012) era.

Overall, there is literally nothing you can do other than straight up ignore the marketed values and just get yourself an appropriate motherboard and cooling based on respective reviews and your intended way of use.

 

[Bill_Bright]

Russia is not communist since 1991.

Not my point - but sure you knew that.

 

[R-T-B]

This is not communist North Korea or China, or Russia. Or fascist Nazi Germany or Iran.

To be fair standards can and have been forced in markets in otherwise free countries. Think USB-C. But yes, I agree this is overbearing.

 

[Bill_Bright]

To be fair standards can and have been forced in markets in otherwise free countries. Think USB-C.

Not sure I would say USB-C has been "forced", at least not in the sense it limits options for the consumer, which the OP's proposal surely does.

For sure, there are "safety" standards that are forced on companies and consumers through government (including EU) laws and regulations. But those are there to protect consumers, not limit their options.

 

[Super Firm Tofu]

Not sure I would say USB-C has been "forced", at least not in the sense it limits options for the consumer, which the OP's proposal surely does.

For sure, there are "safety" standards that are forced on companies and consumers through government (including EU) laws and regulations. But those are there to protect consumers, not limit their options.

I believe that R-T-B was talking about USB-C being mandatory for all devices.

https://www.consilium.europa.eu/en/press/press-releases/2022/10/24/common-charger-eu-ministers-give-final-approval-to-one-size-fits-all-charging-port/

 

[Bill_Bright]

I believe that R-T-B was talking about USB-C being mandatory for all devices.

Right. But note that is to limit or prevent manufacturers intentionally ripping off consumers with "proprietary" parts. Anything that stops companies from forcing proprietary parts on us is a good thing, IMO.

 

[Macro Device]

Right. But note that is to limit or prevent manufacturers intentionally ripping off consumers with "proprietary" parts. Anything that stops companies from forcing proprietary parts on us is a good thing, IMO.

And the OP despite being a little confused and misinformed has the right thing on their mind.

I'd adjust their take though.

What I consider to be better than the actual state of things is companies listing:
• Mean wattage in CPU intensive gaming.*
• Mean wattage in CPU intensive workloads.*
• Peak wattage in CPU intensive workloads.*
• Suggested cooling solutions, as in types and calibres, as well as suggested motherboard tiers so the end user will need less google to make their decision.**

*with power consumption limits turned off.
**with a fair share of overkill like assuming +40C ambient and bad airflow in the PC case so the company is safe from "you recommended me this cooler and my PC still overheats"

This way we will have a 70/150/200 W rated "Amtel Terminator 7 20K" kind of CPUs.

 

[Bill_Bright]

And the OP ... has the right thing on their mind.

Totally disagree. Your "adjustments" are essentially totally different than his proposal.

You are talking about manufacturers publishing realistic figures. I agree with that. In other words, NO marketing hype.

The OP is suggesting manufactures being told what they can, and cannot make - regardless market demand or the state-of-the-art.

That would be like telling car makers they can ONLY make cars with 200 horsepower engines that can only go 60MPH - even if the car enthusiast wants a car with a 600HP engine that can go 200MPH.

Not only would dictating what the makers make greatly limit user options, it would essentially snuff out innovation and advances in technologies.

 

[Macro Device]

That would be like telling car makers they can ONLY make cars with 200 horsepower engines that can only go 60MPH - even if the car enthusiast wants a car with a 600HP engine that can go 200MPH.

They're only talking mass segment at stock clocks. There has been not a single word about overclocking limitations and bans on extreme CPUs. Let's not stretch things.

This is why I called it "a little confused and misinformed." The dude wants mass consumer hardware go the way of power efficiency rather than "our CPU is X% faster than competition" being silent about the fact these X percent are achieved by doubling the competition CPU's wattage. I totally dig this approach and I really want manufacturers to make it less likely to need powerful power supply, extremely high quality wires at home and ridiculously expensive UPS sets just for the sake of a gaming PC running safe and sound.

Like, we went from 100 W enthusiast CPU + 150 W enthusiast GPU to 150 W mid-tier CPU + 300 W mid-tier GPU combos in a matter of a decade and a half. That's not very economy and ecology friendly in my mind. Of course the modern stuff gets way more work done even if underscaled to 100 + 150 W but stock settings being pumped up higher and higher gen to gen are not a pleasant thing to behold, especially in our day of climate change awareness, Earth pollution awareness, global poverty explosion caused by C-bomb and other factors, and all that uncomfortable things to live with.

So yeah, leashing manufacturers so they don't spiral to only make 500+ W 2000+ USD extra enthusiast SKUs completely ignoring less adventurous audience would be nice. I know my numbers do look absurd and I understand they must be absurd but you get the idea:

• You see one wattage being marketed and get a different, much bigger number on your system you never bothered to manually tune and get "user error" kinda response from the manufacturer. Dat's just not cool.
• You are forced to buy a higher tier PSU to run your wares at stock compared to what you had to buy a dozen years ago.
• This also forces motherboard manufacturers to put more stress on their VRM, PCB quality and all that stuff, ultimately making your system even more expensive.
• This doesn't come free in terms of space and weight so boxes with motherboards and PSUs are becoming heavier and bigger so you also get logistics taxed harder.
• Same applies to GPUs since they became at least triple the size and triple the mass of what they used to be in 2010. More VRMs, more capacitors, more aluminium, more stuff, resounding more expensive by itself and logistics-wise.
• And yeah, UPS also cost money as well as household electricity renovations. Fans, too.

Of course you still can buy smaller, cheaper and less power hungry things but they are becoming worse and worse in terms of performance per dollar compared to same generation watt-happy monsters every year (Pentium G4400 was a slow 2-core CPU but it wasn't so far behind i7-7700K as Pentium G7400 is behind i7-12700K). Getting butted either way.

 

[Noci]

What I read between the lines in the OP, BoggledBeagle has more empathy towards environmental awareness than technical innovation.

What he proposes is to restrict chip manufacterers in a certain way with a non feasable standard. Next in line the consumer? Something like you're only allowed to have a max. 500W PC as home pc? You're not allowed to have more computing power, unless..... bla bla bla. A slippery slope as strongly put in words by Bill_Bright in his initial reaction.

Regarding to standards, don't the sockets on the mainboards (AM5, LGA1700 and so on) already dictate what the maximum powerdraw for that socket is, despite which suitable CPU you're using?

His intentions are good (providing you think that if all consumers will use less energy is going to save the world), but the underlaying thoughts are on the wrong website/forum.

In my opinion TPU is not the place for this discussion or at least not in the hardware section, but feel free to disagree

 

[R-T-B]

Right. But note that is to limit or prevent manufacturers intentionally ripping off consumers with "proprietary" parts. Anything that stops companies from forcing proprietary parts on us is a good thing, IMO.

And I agree on that front.

 

[BoggledBeagle]

Not only would dictating what the makers make greatly limit user options, it would essentially snuff out innovation and advances in technologies.

What I proposed is to corner makers and force them to inovate and force them to use the best most efficient technology to make their products on. For example, Intel has been making CPUs on the same process for how many years now? How is that benefiting consumers?

It would force them to put together the best performing and balanced CPUs at a given power level for the benefit of the consumers. For example, Intel has been making the same CPUs just with different core counts and frequencies for three years now. Where is the innovation in that and benefit for the consumers? There is none.

It would also greatly help consumers to choose better solution for their needs. The true sthrenghts and weaknesses of the products would become crystal clear. Competition between makers would intensify, for the benefit of consumers.

Producers are today placing products on the market with ridiculous out of the box settings, running horribly inefficiently, running at the insane power level of server CPUs with many dozens of cores. You buy a consumer PC CPU, put is in a normal consumer motherboard and it draws 370W?! That is insane.

Try to think once more about what I wrote and what benefits if would bring to consumers. Take your time.... You got it completely backwards the first time.

 

[Noci]

What I proposed is to corner makers and force them to inovate and force them to use the best most efficient technology to make their products on. For example, Intel has been making CPUs on the same process for how many years now? How is that benefiting consumers?

It would force them to put together the best performing and balanced CPUs at a given power level for the benefit of the consumers. For example, Intel has been making the same CPUs just with different core counts and frequencies for three years now. Where is the innovation in that and benefit for the consumers? There is none.

It would also greatly help consumers to choose better solution for their needs. The true sthrenghts and weaknesses of the products would become crystal clear. Competition between makers would intensify, for the benefit of consumers.

Producers are today placing products on the market with ridiculous out of the box settings, running horribly inefficiently, running at the insane power level of server CPUs with many dozens of cores. You buy a consumer PC CPU, put is in a normal consumer motherboard and it draws 370W?! That is insane.

Try to think once more about what I wrote and what benefits if would bring to consumers. Take your time.... You got it completely backwards the first time.

I dig the minimal performance per Watt parameter, that could have potential to force manufacterers to think in a slightly different pattern.

But on the other hand this trend was already set a couple of years ago when power prices went throug the roof and the aspect of power efficiency became a strong selling point. It also has everything to do with technical (im)possibillities with the latest techniques on chip manufactering. It is more complex than you think, but believe me, if some chip maunfacterer could produce a CPU with i9 14900K capabillities when only using 160W @ peak performance they would start straight away. It's about market forces at work, they will sell more, not to mention it would obliviate the competition,

Nevertheless, what you're proposing is close to economical dictatorship and mentioned (consumer) benefits are merely assumptions, so in my opinion not realistic.
Basic question you're asking: does an induvidual need so much computing power? (probably not)

But do we want to have the freedom to choose for ourselves? hell yes .

Right. But note that is to limit or prevent manufacturers intentionally ripping off consumers with "proprietary" parts. Anything that stops companies from forcing proprietary parts on us is a good thing, IMO.

If I do recall the EU-commitee hearings right, it was more about preventing from excessive e-waste, consumer protection was just a beneficial political selling point

 

[Wirko]

You buy a consumer PC CPU, put is in a normal consumer motherboard and it draws 370W?!

Well, the answer isn't "no", but there's much more to be said here: with a lot of dedication, money and time spent, you build a PC with a cooling setup that can dissipate 370 W for an extended period of time. Then you find out that your precious 600€ CPU on a 700€ motherboard can't be pushed over 280 W until you spend more time and patience overclocking it.

Less determined folks will buy a processor for around 300 €$£ and stay in the 50 - 70 W range when gaming, and below that range when browsing and using various applications - except when those applications are actively doing something useful, such as transcoding, rendering, code compilation and so on. Those who transcode, render, compile and so on many hours a day, every day, will likely buy or build a workstation with a Threadripper or Xeon, but those processors won't consume significantly less energy for the same amount of work done than consumer CPUs.

 

[Macro Device]

if some chip maunfacterer could produce a CPU with i9 14900K capabillities when only using 160W @ peak performance

They would still push it to 5 million watts because they need the max performance to show off.

 

[BoggledBeagle]

I could also compare the situation on the CPUs for PCs market as a messy chaotic deafening car show with cars with mislabeled engine displacements, with engines running revved up to the melting points, all competing in the same time for consumer attention. A hell on earth, really.

What I am proposing is to force car makers to correctly mark engine displacements, define car race classes based on engine displacement and let manufacturers optimise and develop the best car they can for the given class and then in an orderly manner display their creation in the race, allowing all consumers to assess the performance of the cars and then depending on the price make the correct purchase decision in the performance class they are interested in.

It that clearer now?

As in the car races, CPU manufacturers would be forced to optimise and use the best technology available, all marketing tricks and smoke screens are not usable here, because in a car race everything depends only on the reached time. It is very clear who won.

Let us presume a customer wants to buy a 100W universal CPU. He is interested mainly in tasks C and D, CPU Amud has 115% performance in C and 90% in D of the performance of the CPU Incel. Customer values performance in D twice as much as performance in C, and Amud it 20% less expensive than Incel. A homework for you here: Should consumer choose Amud or Incel? I came to conclusion that even that Amud has only 98% performance (as valued by the customer) of Incel, customer would choose Amud, because it is 20% cheaper than Incel.

Here you go. 100% rational and correct decision making between two great and efficient optimised products. Customer got exactly what he needs for a great price, thanks to intensive competition between manufacturers.

 

[AusWolf]

• enable customers so choose appropriate cooler

That was the original idea behind AMD's TDP number - to provide guidance as to what kind of cooler one needs for a certain processor. But it clearly doesn't work when a 7800X3D with a TDP of 120 W throttles with a be quiet! Shadow Rock LP (while consuming less than 80 W) which happens to be rated at exactly 120 W by the manufacturer.

I agree that some form of standardisation would be nice, I'm just not sure how to go about it. You cannot draw a correlation between power draw, temperature and heat dissipation, especially if you look at how different Intel and AMD chips are in this regard. Your cooler may be good for an i7 that consumes up to 200 W, but could throttle an AMD chiplet CPU at 150 W.

The only thing I can come up with from the top of my head is categorisation.
For example, if you have an A-class CPU, it'll need an A-class cooler not to throttle by default. But if you have a C-class CPU, you'll be fine with A, B or C-class coolers.
The problem is, there needs to be an independent board testing and giving out these category marks, and it's bad marketing for certain models, so it'll never happen.

 

[Onasi]

Let us presume a customer wants to buy a 100W universal CPU. He is interested mainly in tasks C and D, CPU Amud has 115% performance in C and 90% in D of the performance of the CPU Incel. Customer values performance in D twice as much as performance in C, and Amud it 20% less expensive than Incel. A homework for you here: Should consumer choose Amud or Incel? I came to conclusion that even that Amud has only 98% performance (as valued by the customer) of Incel, customer would choose Amud, because it is 20% cheaper than Incel.

That’s cool and all, except there is no such thing as a 100W universal CPU because of boost algorithms, sleep states, race-to-idle, race-to-sleep, core parking and the fact that different tasks will stress the CPU differently in terms of wattage. The whole premise is asinine. I would be the first one to agree that Intel flagship chips have absolutely bonkers defaults or that AMD shot themselves in the foot with the THICC IHS on AM5 chips to preserve cooler compatibility, but trying to find some arbitrary inflection point around which all CPUs would have to balance their power budget is impossible. You CANNOT adequately compare current Intel and AMD chips even at fairly close wattages because they just behave differently.

 

[AusWolf]

That’s cool and all, except there is no such thing as a 100W universal CPU because of boost algorithms, sleep states, race-to-idle, race-to-sleep, core parking and the fact that different tasks will stress the CPU differently in terms of wattage. The whole premise is asinine. I would be the first one to agree that Intel flagship chips have absolutely bonkers defaults or that AMD shot themselves in the foot with the THICC IHS on AM5 chips to preserve cooler compatibility, but trying to find some arbitrary inflection point around which all CPUs would have to balance their power budget is impossible. You CANNOT adequately compare current Intel and AMD chips even at fairly close wattages because they just behave differently.

Not to mention that if the difference in performance and price between the two CPUs is around the 10-15% mark, then said customer won't notice any difference regardless of which model they choose.

 

[Onasi]

Not to mention that if the difference in performance and price between the two CPUs is around the 10-15% mark, then said customer won't notice any difference regardless of which model they choose.

I would go as far as to say that an average customer who is not an enthusiast and not a hardcore gamer will fail to notice a difference in the A/B test even between, say, a 7600X and a 14900K. An average customer is a surprisingly non-demanding animal. He cares whether he can do his office tasks well enough and if cat videos on YouTube or a hyped series on Netflix streams without problems. That’s kind of it. In terms of “average computing” we had more than enough performance even a decade ago. You can throw an SSD into an Ivy Bridge rig and it will satisfy 99% of said average consumers.

 

[Vayra86]

How about no?

This doesnt make a sliver of sense

 

[AusWolf]

I would go as far as to say that an average customer who is not an enthusiast and not a hardcore gamer will fail to notice a difference in the A/B test even between, say, a 7600X and a 14900K.

I will go even further. Even an enthusiast won't notice any difference in games between a 7600X and 14900K, provided a mid-range GPU, such as a 7800 XT or 4070 Ti is being used.

 

[Onasi]

I will go even further. Even an enthusiast won't notice any difference in games between a 7600X and 14900K, provided a mid-range GPU, such as a 7800 XT or 4070 Ti is being used.

Not without an FPS counter on screen, no. And considering that most respectable displays nowadays come with Variable Refresh this is even more of a non-issue.
I laughed, by the way, how we just started casually referring to an 800 dollar GPU as “mid-range”. Like, sure, technically in terms of model number, but… let’s not enable NVidias nonsense any further than we have to.