Proposal for CPUs/graphic cards for PCs power draw standardisation

[Bill_Bright]

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

Does one reason having a higher priority make the other reasons irrelevant?

Having said that, when reading the article @Super Firm Tofu linked to in post #7 above and scrolling down to the "Background" section, it suggests this initiative was brought on by,

38% of consumers report having experienced problems charging their devices because no compatible charger was available. To address these (sic) issues, on 23 September 2021 the European Commission tabled a proposal for a common charger.

Regardless the reason for the new regulation, of course, minimizing e-waste is important. And that may have been the top, though perhaps hidden, agenda in their decision. I don't know. But e-waste, along with proprietary, non-standard devices is a consumer problem too - if you consider the consumers are the taxpayers footing the bill to deal with e-waste too - if not us, then our children and grandchildren.

 

[BoggledBeagle]

.... An average customer is a surprisingly non-demanding animal. ...

If I take the lineup from the first post:

universal 40W CPU
universal 65W CPU
universal 100W CPU
gaming 100W CPU
universal 160W CPU
"home workstation" 160W CPU

I would think that in terms of being able to satisfy real needs of consumers, the percentage of products may look like something like this:

50
30
10
5
4
1

The really sold percentages because people always buy better things than they really need could be something like this:

30
25
20
15
8
2

 

[TumbleGeorge]

No one should ban anything, only impose taxes on consumption above certain norms. The more you consume, the more you pay. Such general conditions already exist in some countries at the household level for consumed electricity. There probably is/or will be for the computer components as well. In household appliances, there has long been a division by efficiency. In some countries, specific measures may have already been taken, and devices that are too inefficient may not be profitable to use. Thus, although no one forbids you to buy a washing machine from the lowest level of efficiency, they already fall from the baskets of customers in a "natural" way.

 

[BoggledBeagle]

No one should ban anything, only impose taxes on consumption above certain norms.

No. If two manufacturers publish numbers which do not correspond directly to how the products behave and these numbers are incomparable between each manufacturer, you ban the manufacturers from using such numbers and force them to inform consumers in a way that enables consumers to make purchasing decisions.

 

[TumbleGeorge]

No. If two manufacturers publish numbers which do not correspond directly to how the products behave and these numbers are incomparable between each manufacturer, you ban the manufacturers from using such numbers and force them to inform consumers in a way that enables consumers to make purchasing decisions.

For this purpose, there are countless reviews that show the consumption of the system that has been tested and of the component itself, in different cases of use. In a work process, including games, without (additional) load (in idle).
The assembly options and use cases are so numerous that manufacturers are hardly able to test and describe each of them. Limited number of descriptions, may not match many cases of user choice. So, it would be a pointless endeavor. Additional personnel may even need to be hired to verify and describe variants within the manufacturers' staff, and all the different brands of components tested will need to be purchased. Which then what, will they be sold on the second-hand market so that they do not clog up their premises? It's just an unrealistic expectation to go to someone's site and get completely accurate numbers like for yourself.

 

[BoggledBeagle]

Official specifications and reviews are two different things. I just checked Intel specs of 14900K, and the 125W number involves running a MYSTERY WORKLOAD. Yeah. Not good at all.

Having clearly defined classes of products also simplifies choice between the combination of the products.

For example: If you wanted to get a lower end 100W graphic card, you would know that you do not need any more expensive CPU for it than the 65W model. You would also know that you should spend the extra money for the gaming version of the 100W CPU only if you are running the best 250W GPU. Etc.

 

[agent_x007]

That table is 100% fantasy...
On CPU side, it would be easier to simply lock models to max. voltages they can use (like X3D does now) :
"40W" = 0.7 - 0.75V max.
"65W" = 0.8 - 0.85V
"100W" = 0.9V - 1V
"125W" = 1V - 1.1V
125W+ = uncapped.
Frequency can be figured out by manufacturer, with ability to OC beyond it being option on some models.
For maximum efficiency, FIVR would need to come back (or to simply rekt any option of voltmodding things )

 

[ShrimpBrime]

Official specifications and reviews are two different things. I just checked Intel specs of 14900K, and the 125W number involves running a MYSTERY WORKLOAD. Yeah. Not good at all.

Having clearly defined classes of products also simplifies choice between the combination of the products.

For example: If you wanted to get a lower end 100W graphic card, you would know that you do not need any more expensive CPU for it than the 65W model. You would also know that you should spend the extra money for the gaming version of the 100W CPU only if you are running the best 250W GPU. Etc.View attachment 326889

Haha good thread. A chef might call it stiring the pot!!

White papers. I take nobodies opinion over released white papers.
And typically, Intel and AMD hold onto those for a while after a generation is released. This I don't quite understand.

Motherboard manufacturers are Indeed in control of cpu operation within certain guidelines.

Most Bioses are written in sight of power savings. P-States and E-states are an EPA standard. The 100mhz BCLK droop is also an energy savings feature.

Processors capable of high wattage is not something new however.

To think a 2700X for example, is 105w. Yes, at its base frequency only. This does not account for idle, peak or even average wattage.

So for Ryzen 7000 or Intel 13/14th, without official white papers, I haven't a comment for these chips.

 

[TumbleGeorge]

To the extent that the chips even within a single wafer are not perfectly spec-matched to each other. What white papers are you expecting?

 

[ShrimpBrime]

To the extent that the chips even within a single wafer are not perfectly spec-matched to each other. What white papers are you expecting?

What are you talking about?

Each and every 14700K (for example) is exactly specced the same. They all will post from the box new exactly the same.

Maybe you're thinking about the binning between i3, i5, i7 and i9?? Doesn't matter. The white papers include the registers for the chips of the same generation. New white papers are released for each tic toc too.

 

[TumbleGeorge]

is exactly specced the same

Not so exactly how you think.
In fact, there is a choice based on quality not only with downgrading to a model with fewer cores, but also within the same model. There are companies that choose the best units, for example i9 xxxxx and sell them at a premium.

 

[ShrimpBrime]

Not so exactly how you think.
In fact, there is a choice based on quality not only with downgrading to a model with fewer cores, but also within the same model. There are companies that choose the best units, for example i9 xxxxx and sell them at a premium.

Architectural white papers.
Binning in house for an i3, all i3 chips are released the same based on the Architectural design for that model. The white papers cover all that.

 

[Totally]

This is a fool's errand. To reframe this proposal you are trying to standardize the output/results when the inputs by nature cannot be standardized.

 

[dj-electric]

Putting arbitrary limits on silicon based hardware is anti competitive, anti science and anti technology.
It is up to consumers to decide which product they pick and include their own judgement of efficiency and power consumption, and it is up to manufacturers to decide how they tune their chips.
This is how a free market should work. If a high power product gets poor reputation, this is up to the company behind it to improve upon it.

Who are engineering-illiterate, random online users who decide what's good for silicon and consumer and what's not? What makes a 250W 14900K sit at the same category of a 350W 64-core Threadripper chip? Do you understand the fine intricacies and endless bureaucracy that those arbitrary limitations impose?

 

[BoggledBeagle]

Putting arbitrary limits on silicon based hardware is anti competitive, anti science and anti technology.
It is up to consumers to decide which product they pick and include their own judgement of efficiency and power consumption, and it is up to manufacturers to decide how they tune their chips.
This is how a free market should work. If a high power product gets poor reputation, this is up to the company behind it to improve upon it.

Defining clear product categories with one fixed paramater greatly facilitates choosing products for consumers, because they can easilly judge performance of these products. Everything about performance of the products is more transparent and clear.

If you as a producer are forced to compete in a category with fixed power draw, you need to use the most efficient and best technology available, because not using it means that you can push less performance out of the product and you loose with the competing better performing products.

As a producer you are forced to:

• develop high performance building blocks of the products
• combine them together and optimise them to maximise performance with power draw constraint and also price constraint in mind
• make them on the best available most efficient technology

The end result for consumers is:

• they have great, highly performing, optimised and efficient products to choose from
• comparing performance and choosing products is easy
• they know what cooler they really need to choose

This all pushes producers to innovate and develop the best they can both in the CPU/GPU architecture and in the manufacturing process.

This disables them from hiding deficiencies in above mentioned by running the products at highly inefficient frequencies and with high power draw.

BTW what I wrote puts manufacturers in a very difficult position, in which any advantage and disadvantage of the product is very clear to the customer. The ability to hide these using marketing smoke screens and brain washing nearly disappeared.

Everything I wrote is directly the opposite of what you wrote. Are you by the chance from one of the producers? I cannot imagine any other reason why would you write what you did.

 

[Gica]

Russia is not communist since 1991.

Seriously? Is something moving there without Putin's consent, the "democrat" elected for life after pseudo-elections? In Russia it is Kadar's communism, with small free businesses but everything that matters is controlled by Putler... sorry... by the state, through the oligarchs.
Ukraine is paying the blood tribute because it wanted to get rid of the mess of this "democracy".

On topic
At this moment, I consider the consumption of today's processors to be super efficient. I say this because one of my systems uses the IGP and consumes as much as one light bulb for office, www and multimedia.
In short: the performance/watt ratio is much better than that offered by the processors of the past decades.

 

[Macro Device]

Seriously? Is something moving there without Putin's consent, the "democrat" elected for life after pseudo-elections? In Russia it is Kadar's communism, with small free businesses but everything that matters is controlled by Putler... sorry... by the state, through the oligarchs.

Totalitarism and communism are different things.

 

[AusWolf]

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.

True enough. The point is, though, that even with an $800 GPU (which isn't really mid-range, I agree), you won't notice any difference.

 

[Gica]

Totalitarism and communism are different things.

[off topic]
Niet, tavarisci!
You cannot develop a free functional market economy in Russia. Everything is controlled by the dictator and it shows. When the main directions come from the center, it is called communism.
There are many "suicidal" Russians (alone or with their families) who thought they had decision-making power in their companies and that they could go over the criminal's head.

Nothing moves in Russia without the Kremlin's consent.
Nothing moves in Kremlin without Putler's consent.
The difference between the Soviets and the current regime is only the illusion that, now, you have control over your life.

P.S. For what I wrote now, in Russia I was preparing for at least 10 years of Siberia.
[/off topic]

 

[BoggledBeagle]

... At this moment, I consider the consumption of today's processors to be super efficient. ....

As a general statement this is not correct.

Correct statements are:

• The technology used for making CPUs novaday enables CPUs to run efficiently.
• There are PC CPUs on the market, which are very or acceptably efficient out of the box.
• There are other CPUs on the market, that are allowed to consume as much power as possible and run as inefficiently as they can out of the box.
• Power draw numbers in the specifications do not express real power draw, are incomparable between manufacturers and mean nothing at all to consumers. For example 13900T has a base power of 35W with cores running at 1100 and 800 MHz running the mystery workload. Who cares about this number, when it involves the CPU running dog slow running God knows what. This is a totally useless number. Wait a minute, I wanted these to be as neutral and correct statements as possible, I must stop now!

 

[dj-electric]

Who defines what is efficiency?
You do? an organization is?
What would you define as efficient? Rendering ratings? gaming ratings? specific compile codes? Do you know how many use cases exist today for CPUs? its not several, its not dozens, its not hundreds.

Would a 14900K that can easily decode 10bit 4:2:2 video be categorized as inefficient while a Ryzen 5 7600 that can't do that be defined as efficient?

There is no simple way to define what compute efficiency is without endlessly branching its use-cases. Efficiency is use-case dependent. OP, you want to standardize something that has a million and one uses, and this whole thread is just an angry tangent about how some CPUs reach particularly high ratings when you use them for stuff like rendering. Nobody should get to tell people how to use their CPUs.

The idea to come out with standards has to come not just from a need, but also from understanding the products and their uses. Arbitrarily deciding which power consumption rating belongs to what market is not just technologically ignorant, its something I would expect from someone who mostly feeds off rage-bait youtube videos and articles about how product X dares to consume this much power, and product Y dares consuming that much power. We live in a drama and anger driven social algorithms clickbait and the fallout is self proclaimed freedom fighters.

Intel screwed up? Criticize them, let them know what you think of their products, tell them you move to AMD. That's what media outlets should do as well.
They make a product, they make the socket for it, they market it. People can vote with their wallets and be informed online about what the product can't or can do - that's exactly why media gets to cover these products unbiased before the product is even available. Ill-informed users today get spoonfed as much information as they want about a product just a few clicks away.

 

[Bill_Bright]

Defining clear product categories with one fixed paramater greatly facilitates choosing products for consumers

I don't think you understand how a free market society works. Not a criticism, just an observation.

You want the government to make the choices for the consumer through laws and regulations that force and dictate what the manufacturers can and cannot make.

You want to give consumers a "choice" (???) by graciously allowing them the choice of a red car or a blue car.

"But I want a green car!"
"Ha ha ha! That funny."

This all pushes producers to innovate and develop the best they can both in the CPU/GPU architecture and in the manufacturing process.

No it doesn't. When artificial limits (and that is what you are proposing) are imposed on a product, it totally stifles competition and innovation.

You really need to take a couple "basic" economics and business classes because you really don't understand how it works - at least not here in the free world. Again, not a criticism, just an observation. Nobody gets to choose where we were born, and sadly, many, still, don't get to choose where they live.

Again, I TOTALLY AGREE we consumers need better consumer protections laws (or at least better enforcement of those laws already on the books) when it comes to "marketing" and "truth in advertising" - no argument there.

And we need better laws and oversight when it comes to product "safety". But we are not talking about faulty airbags, lead in our kid's applesauce, contaminated medicines, shoddy construction or unsafe shortcuts in the building of bridges or apartment buildings (or any corruption that may have led to those issues).

Nor are we talking about the pollution generated during the manufacturing of computer components and other products (including energy production)- where much better improvements are desperately needed.

And I am a strong and consistent proponent for better electronics recycling to keep our waterways and landfills clean of such hazardous waste.

But now you want to take away my freedom of choice? No way! For you to suggest we need a government that will dictate the specs of processors AMD and Intel can make, and therefore limit MY choice - well, I already spent 24 years in the Air Force defending our rights and freedoms, but it seems I need to re-up again.

Please note the second line in my signature.

In the meantime, be happy you have the freedom to choose the most energy efficient CPU out there - well, if you are willing to sacrifice a few FPS.

 

[BoggledBeagle]

Who defines what is efficiency? Efficiency is use-case dependent.

Sure. In my example lineup in the first post I proposed three different sorts of CPUs for home use: universal, gaming and home workstation. You would optimise each of the different kinds of CPUs differently according to the expected typical workloads of these CPUs.

Nobody should get to tell people how to use their CPUs.

Sure, but people should be able to buy a CPU well optimised for their specific use scenario.

The idea to come out with standards has to come not just from a need, but also from understanding the products and their uses. Arbitrarily deciding which power consumption rating belongs to what market is not just technologically ignorant...

Sorry, but car races need to happen between cars of comparable technical parameters and these parameters are arbitrarily set. There is nothing wrong about defining the classes of CPUs based on power draw somebody just made up. The serie 25, 40, 65,100 and 160W seems sufficient to me.

I already wrote, and I think it is very important to dwell on this for a while and think about it thoroughly:

Defining clear product categories with one fixed paramater greatly facilitates choosing products for consumers, because they can easilly judge performance of these products. Everything about performance of the products is more transparent and clear.

I am not trying to assign each power level to a specific market. I proposed in the first post just an example of simple product lineup. While universal CPUs occupy the whole power draw range, I thought that optimising the CPU specifically for some productive workloads makes sense only for the most powerful CPU and optimising for gaming could be practical only for 100W CPU, because I presumed that the 65W universal CPU could be quite useful for gaming on its own and the 160W CPUs may be underutilised for gaming.

People can vote with their wallets and be informed online about what the product can't or can do - that's exactly why media gets to cover these products unbiased before the product is even available. Ill-informed users today get spoonfed as much information as they want about a product just a few clicks away.

I have never seen in any of the initial reviews of 14900K and 7950X any advice for efficiency tuning. None of these CPUs are very efficient at 250W power draw.

Techpowerup only recently published a review of the 14900K limited to different power draws.

I don't think you understand how a free market society works. Not a criticism, just an observation. You want the government to make the choices for the consumer through laws and regulations that force and dictate what the manufacturers can and cannot make.

Consumers need clear and true information about the products to be able to make the correct decision which product to buy. Defining product categories based on a real parameter helps consumers a lot.

Current situation of producers lying about the power draw of their products and making their products incomparable with the competition using the numbers they publish helps "free market society" and consumers how exactly?

 

[ShrimpBrime]

If you forgot the Energy Star sticker for your case badge when you built the high end gaming machine, then you're not running an effecient machine. (Smart ass comment, but partially true)

 

[Gica]

As a general statement this is not correct.

Correct statements are:

• The technology used for making CPUs novaday enables CPUs to run efficiently.
• There are PC CPUs on the market, which are very or acceptably efficient out of the box.
• There are other CPUs on the market, that are allowed to consume as much power as possible and run as inefficiently as they can out of the box.
• Power draw numbers in the specifications do not express real power draw, are incomparable between manufacturers and mean nothing at all to consumers. For example 13900T has a base power of 35W with cores running at 1100 and 800 MHz running the mystery workload. Who cares about this number, when it involves the CPU running dog slow running God knows what. This is a totally useless number. Wait a minute, I wanted these to be as neutral and correct statements as possible, I must stop now!

From my garden:
i5-10500 - i5-12500
6c/12t - 6c/12t
PL1: 65W - 65W
PL2: 134W - 117W

Cinebench R23
Singles: 1210 - 1756
Multi: 8648 - 12750
i5-12500 much more efficient than i5-10500. Both consume more than the old i5-7500 in heavy tasks, but both effectively destroy it in terms of efficiency because they perform tasks much faster.
To calculate the efficiency, you must also take into account the computing power of the system. A system that consumes 100W is more efficient than one that consumes 80W but performs the same tasks in double time.