AMD Zen 3 (5000 Series), rumored 17% IPC gain.

[LePhuronn]

Why do people keep banging on about performance per watt in a desktop computer?

In reality performance and price and price is all that matters in a desktop form factor. A desktops electricity usage is frankly incidental on your household bill. As long as it can be cooled sufficiently with a reasonable heat sink/fan it really doesn’t matter.

Wrong. You should've saw the spike in my electricity bill when I switched to some rubbish CPUs. And given I run multiple computers at once, I don't want a bazillion watt monsters burning through my power. Plus, an efficient processor means better thermal management, important for small systems, confined spaces and generally not spending a fortune on unnecessary cooling and power supplies.

If you're content to run 1 computer in a massive case and burn 400W through an overclocked Intel POS then that's entirely your choice. Personally, I have more stringent requirements.

 

[clv101]

Why do people keep banging on about performance per watt in a desktop computer?

In computing, performance per watt is all important, from phones and laptops, to multimillion dollar supercomputers and data centers.

Desktops are not immune. In an office environment, it impacts power supply on the floorplate and the buildings air conditioning. On an individual level it impacts noise, heat, cost of cooler and space constraints.

Everyone is impacted by power cost and we should all be concerned about the additional CO2 emissions.

As we simply can't increase power consumption, all performance improvements need to come with power efficiency improvements.

It's important.

 

[Robert896r1]

Use cases matter. If I want a track car, while it'd be great if it was more efficient, I'm willing to give up efficiency for a lower laptime as the laptime is a more important criteria to me in this use case. It's best to be realistic with your use cases before making any purchase that will span years.

 

[Journey]

If 3 million desktop PC's were shipped for one year, with a CPU that consumes on average 20w more per CPU, and they were used for 2 hours per day every day of the year, that is 120 million Wh per day, or over 43 GWh of electricity per year.

Drop in the pond, no effect at all on anyone, nope lets just keep making stuff inefficient and worry about the consequences later.

 

[Random Guy Esq.]

If 3 million desktop PC's were shipped for one year, with a CPU that consumes on average 20w more per CPU, and they were used for 2 hours per day every day of the year, that is 120 million Wh per day, or over 43 GWh of electricity per year.

Drop in the pond, no effect at all on anyone, nope lets just keep making stuff inefficient and worry about the consequences later.

This.

 

[b0rn2sk8]

That logic is flawed, it assumes the CPU is running balls to the wall and the reality is quite different. The CPU spends most of its life idle, even when the computer is being used. It’s only the last few % of the performance is what takes all the power.

We’re not talking about a data centre environment either, we’re all consumers here looking at a consumer desktop CPU. If it consumes a few more watts over an equivalent processor is incidental as long as it can be cooled adequately. There is much more that goes into the die design than overall power consumption which impacts cooling.

For example, before the 10th gen Intel parts launched everyone was predicting if they would be a thermal nuclear meltdown but the reality was very different. Intel changes the IHS and die design which improved thermal transfer and it was a complete non issue.

The point I was making that in the real world performance per watt isn’t a big factor in purchasing decisions for consumers buying desktop CPUs. The only time it would be is if it’s massively out of step with the alternative. If AMD is faster people are just going to buy it regardless of if it’s power consumption is slightly higher than intel or the other way round.

For example if AMD release a new GPU and it’s slower than a 3070/80 but consumes 50w less. If people want/need the performance they aren’t going to buy it over the 350w Nvidia because it uses less power.

 

[humbug]

If this is true that's quite a chonky jump over Zen 2.

166.6 / 125.6 = 1.326 (+32.6%)




https://twitter.com/CapFrameX/status/1310823769787846658

 

[RavenXXX2]

My B550 is ready!

 

[Robert896r1]

That logic is flawed, it assumes the CPU is running balls to the wall and the reality is quite different. The CPU spends most of its life idle, even when the computer is being used. It’s only the last few % of the performance is what takes all the power.

We’re not talking about a data centre environment either, we’re all consumers here looking at a consumer desktop CPU. If it consumes a few more watts over an equivalent processor is incidental as long as it can be cooled adequately. There is much more that goes into the die design than overall power consumption which impacts cooling.

For example, before the 10th gen Intel parts launched everyone was predicting if they would be a thermal nuclear meltdown but the reality was very different. Intel changes the IHS and die design which improved thermal transfer and it was a complete non issue.

The point I was making that in the real world performance per watt isn’t a big factor in purchasing decisions for consumers buying desktop CPUs. The only time it would be is if it’s massively out of step with the alternative. If AMD is faster people are just going to buy it regardless of if it’s power consumption is slightly higher than intel or the other way round.

For example if AMD release a new GPU and it’s slower than a 3070/80 but consumes 50w less. If people want/need the performance they aren’t going to buy it over the 350w Nvidia because it uses less power.

A vocal group on here find the highest power draw figure they can find and build an entire narrative on the back of it.

Your logic is what most people use for personal usage, esp given the common use case of gaming which dominates the DIY space. "What's the best performance I can get for my budget?"

 

[LePhuronn]

We’re not talking about a data centre environment either, we’re all consumers here looking at a consumer desktop CPU.

The logic is perfectly fine. OK, let's ignore data centre and office, let's focus on consumer desktop. Journey's numbers still hold true for 1 million consumer desktops. It's your scope of the situation that seems flawed.

 

[humbug]

My B550 is ready!

lol...

 

[Richdog]

Wrong. You should've saw the spike in my electricity bill when I switched to some rubbish CPUs. And given I run multiple computers at once, I don't want a bazillion watt monsters burning through my power. Plus, an efficient processor means better thermal management, important for small systems, confined spaces and generally not spending a fortune on unnecessary cooling and power supplies.

If you're content to run 1 computer in a massive case and burn 400W through an overclocked Intel POS then that's entirely your choice. Personally, I have more stringent requirements.

By running 'multiple computers at home' then you hardly represent the average use case of a PC gaming enthusiast, which is running one computer for gaming and general use. Your comment doesn't mean much without knowing what spec these 'multiple computers' are and what are they used for.

The logic is perfectly fine. OK, let's ignore data centre and office, let's focus on consumer desktop. Journey's numbers still hold true for 1 million consumer desktops. It's your scope of the situation that seems flawed.

It's a similar principle to someone buying a performance car... some people buy the fastest they want and are fine for it to have increased running costs. In the end, proportionally speaking, not many people are running 5ghz overclocked Intel CPU's.

 

[kevman]

Your comment doesn't mean much without knowing what spec these 'multiple computers' are and what are they used for.

Probably mining

 

[LePhuronn]

Not sure how running a couple of machines invalidates the point that power efficiency and thermals are actually (or should be) integral to component choice and purchasing. If anything it serves to highlight further their importance. It doesn't matter what I'm running, does it? Why would I, you or anybody knowingly run a 45W CPU when they can achieve the same goal at 25W? Of course there is a balancing act between performance, cost and everything else, but to just say "it's 1 consumer desktop, power and heat don't matter" is just daft. Power efficiency matters in every single computing application.

 

[LePhuronn]

Probably mining

Nah, I didn't even know it was a thing until it was too late to capitalise on it, plus I've never had the money to even make a start getting the kit in.

I'm talking about a combo web dev server/NAS, primary system for gaming, CAD and animation, couple of older machines slaved into a quick n dirty render farm for animation (not used often), secondary machine for generally pottering about if the primary one is rendering. Plus the other half running her system downstairs (that RX 570 seriously needs a tune up).

A lot of these can be running 24/7 and even if just ticking over quietly, why waste the energy? And certainly I don't want to be sat in a hot and sweaty room for 12-24 hours during a big batch of renders.

 

[FoxEye]

What nonsense you talk. Zen 1 is an upgrade and you know it is. You’re just but hurt by that fact, and want to argue for some strange reason, although I’m not sure why.

e: The follow is wrt gaming performance and nothing else, as that was the metric in question. I know you're one of the "nobody _just_ games" crowd.

By the same strange logic you could make some really daft statements.

"Skylake is an upgrade to the 7600K."
"Haswell is an upgrade to the 7600K."

That is just as true as "Zen 1 is an upgrade to the 7600K."

Both statements require taking a processor with more cores/threads from an inferior architecture and using it to show increased performance in some modern games.

Both statements are essentially a load of crap. Zen1 is an inferior gaming architecture to Kaby Lake. Compare 4c/4t with 4c/4t, and 8c/8t with 8c/8t.

Don't downclock the Kaby to Zen1 clocks because we're not comparing IPC here. Just real world *gaming* perf.

e2: My challenge to you @jigger. Show a 4c/4t Zen1 CPU beating a 7600K in gaming perf. Come on then

 

[b0rn2sk8]

The logic is perfectly fine. OK, let's ignore data centre and office, let's focus on consumer desktop. Journey's numbers still hold true for 1 million consumer desktops. It's your scope of the situation that seems flawed.

No, the numbers are flawed and assume a cpu is running flat out all the time when in reality they consume a handful of watts unless they clock up to keep performance. In reality this happens for a fraction of the time they are active. It’s the last few hundred MHz takes significantly more power than the first.

Again, I think people are getting away from the point I was making and going complete hyperbole on this. Let’s get back to reality for 5 minutes.

If the 10900k ends up being slower than a 5900x but has better power consumption relative to its performance no one is going to care. People will just buy what ever is faster. Neither product is going to operate outside of a reasonable thermal or power envelope so as a consumer it’s not a really a purchasing consideration. That’s the point I was making.

In reality it looks like AMD may well be better on both power and performance but no one is thinking ‘I’m going to go AMD because it uses less power’ there thinking ‘AMD is faster I’ll get that’ or ‘this one has a bigger number on X benchmark, I’ll get that’.

The same thing applies to GPUs.

 

[Richdog]

Nah, I didn't even know it was a thing until it was too late to capitalise on it, plus I've never had the money to even make a start getting the kit in.

I'm talking about a combo web dev server/NAS, primary system for gaming, CAD and animation, couple of older machines slaved into a quick n dirty render farm for animation (not used often), secondary machine for generally pottering about if the primary one is rendering. Plus the other half running her system downstairs (that RX 570 seriously needs a tune up).

A lot of these can be running 24/7 and even if just ticking over quietly, why waste the energy? And certainly I don't want to be sat in a hot and sweaty room for 12-24 hours during a big batch of renders.

So lets get this straight... you are running around 5-6 separate machines and then criticizing people for running 1 machine containing an overclocked Intel CPU? You have to laugh at that kind of oblivious hypocrisy.

 

[humbug]

Unless you have a high budget the only thing people care about when looking at CPU's is "How much CPU can i get for my money? That one is more CPU than this one and its cheaper, sold!"

 

[alec]

Unless you have a high budget the only thing people care about when looking at CPU's is "How much CPU can i get for my money? That one is more CPU than this one and its cheaper, sold!"

Don't forget a hidden added cost of more expensive cooling system, possibly more expensive case that comes with high power CPU.