News Intel's Core i7-14700K Benchmarked: More Cores, Higher Clocks

[TerryLaze]

In your text, you said "... AMD needs the same liquid cooling ...", but your link was to data only for 7950X. So, I was doing two things:

1. Pointing out the mismatch between your words and data.
2. Responding to the point you were apparently just making about the 7950X, by providing further data about the 7950X.

There are plenty of CPUs including but not limited to the 3xd ones that need liquid cooling, stated by AMD themselves.

https://www.amd.com/en/products/cpu/amd-ryzen-9-5900x

Here for the 5900x ,AMD knows what you need for their CPUs.
Recommended Cooler: Liquid cooler recommended for optimal performance

Although, what's funny is that nothing in that link directly supports your claim.

The 7950x is using up to 215w out of it's 230w budget because it can't be cooled enough when running ycruncher.
While the 13900k is easily running 30% above advertised numbers, and does it at 8 degrees lower temps.

They note that the Blender benchmark was measured to use up to 235 W, which is just over the 230 W PPT I assume you're referencing. That's where it loses a couple %. The average performance for the Noctua solution is 0.2% less than AIO. That's not significant.

That doesn't change the fact that you still need an AIO to reach the 230W that is being advertised to you.
The power drop between AIO and noctua at 100% is SIGNIFICANT.
And if you choose to not run any app that needs so much power then good for you but others might need it and not get it.

I already showed this statement is inaccurate (see table). TSMC claims their 5 nm node is 20% more efficient than their 7 nm node, at ISO-frequency, and offers 15% more performance at ISO-power. I know you saw the post, because your reply quoted it.

It's 20% more efficient IF YOU DON'T CHANGE ANYTHING so if you release the exact same CPU it would use 20% less power, which everybody would just laugh at and they would have zero sales.
The 15% are the theoretical higher clocks if you have a very simple design, for complex designs like CPUs it is much less.
But even if it would be 15% higher clocks it would be so at a smaller size increasing the need for cooling even more and there is no room for better cooling.

 

[bit_user]

There are plenty of CPUs including but not limited to the 3xd ones that need liquid cooling, stated by AMD themselves.

https://www.amd.com/en/products/cpu/amd-ryzen-9-5900x

Here for the 5900x ,AMD knows what you need for their CPUs.
Recommended Cooler: Liquid cooler recommended for optimal performance

That's a recommendation because not all air coolers are up to the task. Rather than try to specify which, it's easier for them just to specify liquid. That doesn't make it necessary, as the TechPowerUp data clearly shows.

In fact, according to the i9-13900K review, Intel does the same:

"Intel doesn’t include a cooler with the Raptor Lake processors but recommends a 280mm AIO watercooler"

Source: https://www.tomshardware.com/reviews/intel-core-i9-13900k-i5-13600k-cpu-review
​

That doesn't change the fact that you still need an AIO to reach the 230W that is being advertised to you.

The 230 W figure is a limit, not a feature! It's never guaranteed to hit that. Rather, that limit exists for safety reasons, to prevent motherboard and other components from being over-stressed.

The power drop between AIO and noctua at 100% is SIGNIFICANT.

You're confusing power consumption with performance. What matters, and what the user cares about, is maximizing performance. I already posted the performance data, which shows a negligible difference, even in that nThreads Cinebench case.

It's 20% more efficient IF YOU DON'T CHANGE ANYTHING so if you release the exact same CPU it would use 20% less power,

I think this statement explains a lot. Efficiency is performance divided by power. You get more efficiency by reducing power, not increasing it!

The 15% are the theoretical higher clocks if you have a very simple design,

No, it's not. The only place that constraint exists is in your imagination.

It's for porting a given design from one node to another, irrespective of complexity, so long as it's held constant. The article even says as much:

"It will also increase frequency by 15% (at the same complexity and power) or reduce power consumption by 20% power reduction (at the same frequency and complexity)."
​

But even if it would be 15% higher clocks it would be so at a smaller size increasing the need for cooling even more and there is no room for better cooling.

This statement explains a lot about why you think certain things. The cooler needed to dissipate 100 W is the same, irrespective of die size. The only challenge die size poses is the potential for a thermal bottleneck transferring energy from the die to the IHS. As people have shown with direct-die cooling of Ryzen 7000-series, the problem is their IHS.

As for process node, what we usually see is that some of the dividends of the node-shrink are re-invested into a more complex design which achieves higher IPC, because that's more efficient than a straight boost of clock frequency.

 

[thestryker]

The only challenge die size poses is the potential for a thermal bottleneck transferring energy from the die to the IHS. As people have shown with direct-die cooling of Ryzen 7000-series, the problem is their IHS.

This isn't really a universal truth so much as what we're seeing in action today. AMD made an awful choice in maintaining AM4 compatibility so their IHS makes an outsized impact. Size of CPU die as well as location definitely comes into play because that changes how you design respective coolers (ex: Ice Giant's first cooler they basically said don't use it for desktop chips). The problem is also going to get a lot worse if clockspeeds stay high like they currently are.

I think we're going to see a lot more changes in cooler design over the next few years due to shrinking die size and Intel also moving the cores (hopefully to the same side as AMD uses just to make the cooler game easier). I don't think the consumer market needs to worry about needing exotic cooling like datacenter/HPC will, but things will change. We may also see some big increases in e/c-cores in lieu of high all core clocks so cooling simply isn't the problem it's barreling towards today.

 

[bit_user]

This isn't really a universal truth so much as what we're seeing in action today. AMD made an awful choice in maintaining AM4 compatibility

Did you see their prototype vapor chamber IHS? I think that explains a lot. They probably went with the taller spec to give them room for a vapor chamber. Then, by the time they finally cut their losses on the vapor chamber project (they said it was too expensive for the gains it provided), it was probably too late for them to revise the height.

AMD Had Ryzen 7000 Prototype CPUs With Vapor Chamber Cooling Embedded Within IHS

AMD initially planned to introduce a vapor chamber into its Ryzen 7000 CPU lineup to replace the traditional IHS but decided against it.

wccftech.com

Size of CPU die as well as location definitely comes into play because that changes how you design respective coolers

See, a vapor chamber is very good at distributing heat and reducing hotspots. They could use Z-oriented graphene to efficiently transfer heat from the die to IHS, but it'd be rubbish at spreading out the heat, the way a vapor chamber does. I hope the vapor chamber somehow comes back. Maybe they could use AI to optimize it?

Loosely related:

Cooler Master Builds World's Thinnest Vapor Chamber

Just slightly thicker than human hair.

www.tomshardware.com

 

[thestryker]

See, a vapor chamber is very good at distributing heat and reducing hotspots. They could use Z-oriented graphene to efficiently transfer heat from the die to IHS, but it'd be rubbish at spreading out the heat, the way a vapor chamber does. I hope the vapor chamber somehow comes back.

I could see vapor chambers potentially being the solution for consumer level chips. The only question I'd have there is the cost viability since laptop chip sales are quite a bit higher still and wouldn't leverage it. I'm also curious about Cooler Master's new vapor chamber prototype as I could see that being the way high end air goes.

 

[Ryrynz]

When you overclock the memory you dont need official intel support , and mind your language.

Not everyone has or can overclock buddy.
Mind your language and mind having basic PC knowledge before posting please.

 

[newtechldtech]

Not everyone has or can overclock buddy.
Mind your language and mind having basic PC knowledge before posting please.

the ones who dont want to overclock will not pick the "K" CPU and waste money on expensive overclocking CPU and motherboard.

and again mind your language . saying to people "go to sleep" has nothing to do with arguing about knowledge.

 

[bit_user]

the ones who dont want to overclock will not pick the "K" CPU and waste money on expensive overclocking CPU and motherboard.

That's not really true, because the spec difference between K and non-K is fairly substantial. Even just to run at stock settings, the K provides enough of a boost to justify the relatively small price difference.

I happen to believe the official RAM speed matters, because overclocking usually requires buying more expensive memory & motherboards and there's an element of luck to it. Whereas the stock settings are basically guaranteed. This is especially important for commercial PC builders, where they want a known configuration they can advertise and reliably achieve (and for the least $).

Personally, I stick to stock settings for the purpose of stability. For my purposes, the additional speed of OC isn't worth the risk of crashes or data corruption.

 

[thestryker]

That's not really true, because the spec difference between K and non-K is fairly substantial. Even just to run at stock settings, the K provides enough of a boost to justify the relatively small price difference.

I happen to believe the official RAM speed matters, because overclocking usually requires buying more expensive memory & motherboards and there's an element of luck to it. Whereas the stock settings are basically guaranteed. This is especially important for commercial PC builders, where they want a known configuration they can advertise and reliably achieve (and for the least $).

Personally, I stick to stock settings for the purpose of stability. For my purposes, the additional speed of OC isn't worth the risk of crashes or data corruption.

Also of note is that non-K SKUs have system agent voltage locked which means even just running higher speed DRAM is out (though less of a concern when using DDR5 instead of DDR4).