News AMD deep-dives Zen 5 — Ryzen 9000 and AI 300 benchmarks, Zen 5, RDNA 3.5 GPU, and XDNA 2 microarchitectures

[TheHerald]

In the article, same cooler (FROZN A620 PRO SE Max Speed) I see :
300W @ 100°C for Intel
229W @ 95°C for AMD
This confirm the fact that is more difficult to cool a CPU that have more Wattage.
What I continue to say but somebody seems to not understand.

The temperature is irrelevant since they are both thermal throttling. The cooler can cool 300w on intel and 229 on and amd chip. Obviously intel is easier to cool. Your own numbers prove that, lok

 

[TheHerald]

At same Wattage, yes, it is obvious.
But the problem is that Intel use a lot more power and so, it is more difficult to cool.
It is simple, more watt = more cooling problems.
I could confirm your point only with a marginal difference in power consumption, but in case of 7950X and 14900k the difference is huge (insane imho) and the more efficient IHS or the larger area will never compensate the wattage difference.

No, intel doesn't use more power. I have an Intel 12900t that is drawing 35 watts. Please, explain to me how exactly does it draw more power than amd chips?

There are Intel cpus that draw less power than amd and others that draw more. So that doesn't matter. What matter is Intel chips are easier to cool.

 

[jeremyj_83]

At same Wattage, yes, it is obvious.
But the problem is that Intel use a lot more power and so, it is more difficult to cool.
It is simple, more watt = more cooling problems.
I could confirm your point only with a marginal difference in power consumption, but in case of 7950X and 14900k the difference is huge (insane imho) and the more efficient IHS or the larger area will never compensate the wattage difference.

This is why server CPUs with 500W TDPs are all going liquid cooling. Even though it is "easier" to cool a 500W TDP server CPU than desktop CPU due to surface area (hot spots are more spread out and larger surface is easier than a smaller surface to remove heat the same level of heat) the end result is you still need to cool 500W. That is not easy to do.

 

[NinoPino]

The temperature is irrelevant since they are both thermal throttling.

Ops. My mistake, I will search a new proof.

The cooler can cool 300w on intel and 229 on and amd chip.

The cooler can probably cool more watt, as you said both CPUs are throttling.

Obviously intel is easier to cool. Your own numbers prove that, lok

No man, they both are throttling, so this prove only that the cooling solution is undersized.

 

[TheHerald]

Ops. My mistake, I will search a new proof.


The cooler can probably cool more watt, as you said both CPUs are throttling.


No man, they both are throttling, so this prove only that the cooling solution is undersized.

But one is throttling at a way lower power draw cause it has worse dissipation. How hard is it to get? I can't believe you are not getting this...

 

[NinoPino]

But one is throttling at a way lower power draw cause it has worse dissipation. How hard is it to get? I can't believe you are not getting this...

Worse dissipation is only a piece of the cake, the other piece is the maximum wattage to dissipate.
The better dissipation of Intel CPUs will never compensate the biiig difference in wattage that MUST be transferred out of the die.

 

[TheHerald]

Worse dissipation is only a piece of the cake, the other piece is the maximum wattage to dissipate.
The better dissipation of Intel CPUs will never compensate the biiig difference in wattage that MUST be transferred out of the die.

But they don't require more power. Thats your own biased view that has no basis in reality. There are 65 and 35w intel chips. Why are you ignoring this?

 

[jeremyj_83]

But they don't require more power. Thats your own biased view that has no basis in reality. There are 65 and 35w intel chips. Why are you ignoring this?

Intel's TDP is total bull. That 65W 14900 has a 219W PL2 and the 35W 14900T has a 106W PL2. That means that with the 65W CPU you could still need to remove over 200W. Compare that to a 65W AMD that has an 88W PPT. Which is going to be easier to cool? Thermodynamics says the AMD.

 

[JarredWaltonGPU]

Take this with a grain of salt, but L1Techs was saying they probed some sources at big PC OEMs who said it appeared to be in the range of 10% to 25%. I assume they mean the number of CPUs they expect to replace during the entire time those systems remain under warranty, but that wasn't specified. I find it stunning, if true. Also, not clear if they meant just i9's or both i7's and i9's.

Even replacements in the low single digit percentages would be terrible. 10% to 25% would be an insane number of failures! Even one quarter of that rate would be awful and cost Intel billions. Which would explain why it's being very cautious about saying anything regarding root cause and probably looking for alternative solutions that won't be as bad.

Like I said earlier, I wouldn't be surprised if Intel actually does have a really good idea of the root cause. It may simply be that a fix will be incredibly expensive — like replacing mounting mechanisms or chips or both. Problems like that, you don't admit to unless forced to issue a recall. I suspect we have not heard the end of this and we'll probably end up with a class action suit or similar.

 

[TheHerald]

Intel's TDP is total bull. That 65W 14900 has a 219W PL2 and the 35W 14900T has a 106W PL2. That means that with the 65W CPU you could still need to remove over 200W. Compare that to a 65W AMD that has an 88W PPT. Which is going to be easier to cool? Thermodynamics says the AMD.

PL2 lasts for less than a minute. Even with that 219w pl2 you are still drawing LESS power than the 7950x. Of course for just a minute, after that minute you are drawing 1/4th the power.

 

[jeremyj_83]

PL2 lasts for less than a minute. Even with that 219w pl2 you are still drawing LESS power than the 7950x. Of course for just a minute, after that minute you are drawing 1/4th the power.

Only if tau isn't set to unlimited which had been the default Intel wanted for multiple CPU generations. At the same time we know from testing that if you actually limited the Intel chip to only its TDP that it is quite a bit slower than the AMD 7950X at the same PPT.

 

[NinoPino]

But they don't require more power. Thats your own biased view that has no basis in reality. There are 65 and 35w intel chips. Why are you ignoring this?

Are you joking ? 65/35w CPUs ?
Please.

 

[TheHerald]

Only if tau isn't set to unlimited which had been the default Intel wanted for multiple CPU generations. At the same time we know from testing that if you actually limited the Intel chip to only its TDP that it is quite a bit slower than the AMD 7950X at the same PPT.

Unless you are buying a z mobo (which you shouldn't with a locked chip) it isn't. But even if it was 106w on the 14900k is still pretty low, no?

And no, we don't know that it's quite a bit slower

 

[TheHerald]

Are you joking ? 65/35w CPUs ?
Please.

Yes, I like my cpus efficient. If I wanted 200+ watt chips like the 7950x I'd buy a furnace instead.

 

[jeremyj_83]

Yes, I like my cpus efficient. If I wanted 200+ watt chips like the 7950x I'd buy a furnace instead.

I don't think you understand the word efficient. If you did you would realize that efficient and Intel are mutually exclusive terms. We have shown in older threads that the 13900k (same as the 14900k just lower clocks) needs to be at a much higher TDP to equal performance of the AMD chips. In fact we see that a 105W TDP seems to be the best spot of the 7950X as you get minimal gains at 170W TDP. The Intel chip does scale better with additional power, however, Intel needs a 125W TDP to be faster than the AMD at 65W usually. Hence the Intel chip isn't nearly as efficient.

 

[bit_user]

Intel cpus don't produce more heat.

It depends on the workload, but they certainly can. Heat == Joules (which is essentially the average Watts * duration of task).

At the end of the day you take a 7950x and a 14900k, you lock them both to 250w,

Ah, but you had to find a workload that could make them both reach that number, and then limit them to it. So, what you really mean is that you can create a scenario in which the Intel CPU isn't generating more heat, but runs at cooler temps (due to things like fewer bottlenecks at the junction between the die and IHS).

 

[bit_user]

This is why server CPUs with 500W TDPs are all going liquid cooling. Even though it is "easier" to cool a 500W TDP server CPU than desktop CPU due to surface area (hot spots are more spread out and larger surface is easier than a smaller surface to remove heat the same level of heat) the end result is you still need to cool 500W. That is not easy to do.

A few details...

• GPUs can reach 500W, so it can be done.
• server CPUs are typically needing to run in just a 2U chassis, which adds additional challenges.
• liquid-cooling a server potentially allows you to transfer the heat outside via water, rather than dumping it directly into the room air and then having to cool that.

 

[TheHerald]

I don't think you understand the word efficient. If you did you would realize that efficient and Intel are mutually exclusive terms. We have shown in older threads that the 13900k (same as the 14900k just lower clocks) needs to be at a much higher TDP to equal performance of the AMD chips. In fact we see that a 105W TDP seems to be the best spot of the 7950X as you get minimal gains at 170W TDP. The Intel chip does scale better with additional power, however, Intel needs a 125W TDP to be faster than the AMD at 65W usually. Hence the Intel chip isn't nearly as efficient.

I'm pretty confident that eg. a 14900t is vastly more efficient than the 7950x. VASTLY. Like t hey are not even close vastly. For example this test seems to point at that direction

 

[TheHerald]

Intel needs a 125W TDP to be faster than the AMD at 65W usually. Hence the Intel chip isn't nearly as efficient.

Yeah, that's the usual braindead take I keep seeing being spread around. Only problem is, it's just not true, lol. Even if you compare the 7950x to the 14900k, the only segment that AMD actually has somewhat of a lead, it's very not true. At 125w the 14900k scores around 31k in CBR23. The 7950x at 65w scores 22k. But whatever, as I've noticed, facts don't matter when it comes to hating on Intel so keep at it.

Now if you go further down the stack, oh boy. I'd like to see a 65w 7700x against a 125w 13700k. Wouldn't you? 😍

 

[bit_user]

Unless you are buying a z mobo (which you shouldn't with a locked chip) it isn't. But even if it was 106w on the 14900k is still pretty low, no?

And no, we don't know that it's quite a bit slower

As I've pointed out before, that "Applications" metric averages over both lightly-threaded and heavily-threaded apps! It won't be terribly good at predicting either one, which is why I think Toms and ComputerBase have the right idea by using two separate metrics.

 

[TheHerald]

It depends on the workload, but they certainly can. Heat == Joules (which is essentially the average Watts * duration of task).


Ah, but you had to find a workload that could make them both reach that number, and then limit them to it. So, what you really mean is that you can create a scenario in which the Intel CPU isn't generating more heat, but runs at cooler temps (due to things like fewer bottlenecks at the junction between the die and IHS).

Man it's not complicated. If my single tower air cooler u12a can dissipate 330 watts on a 14900k and only 200 watts on a 7950x, it's blatantly obvious that intel chips are easier to cool.

 

[TheHerald]

As I've pointed out before, that "Applications" metric averages over both lightly-threaded and heavily-threaded apps! It won't be terribly good at predicting either one, which is why I think Toms and ComputerBase have the right idea by using two separate metrics

Isn't that how people use their PC's?

 

[bit_user]

I'm pretty confident that eg. a 14900t is vastly more efficient than the 7950x.

But also slower. You cannot separate the the matter of performance and efficiency. The i9-14900T shouldn't be compared to the R9 7950X, because they differ so drastically in both performance and efficiency! Nobody who's in the market for one would also be considering the other.

I've talked about this before, as a "mismatched comparison" that you seem to like making.

 

[-Fran-]

Not sure if trolling or not understanding words.

Thermal Efficiency != Power Efficiency

A thermally efficient "thing" would be able to move a lot of heat with very little "something" (dissipation area, conductive material, etc; smaller cooler can move the same as a less thermally efficient thing with a bigger cooler, for example, and even then there's a few caveats).

A power efficient CPU will do more work (have better performance) at the same power target, which is independent from temperature achieved. Temperature is a direct consequence of the power used and how you're measuring said temperature. There's a whole interview with an Intel engineer that tries to explain how measuring the temperature in a modern CPU is really hard, so what you get is a rough approximation to what it should be.

AMD's IHS is shameful (hence they're improving it, allegedly) as it brings the thermal efficiency down a lot, but the power efficiency of AMD is still way better than Intel in most power targets that matter for consumers and data centers. This is not up for discussion anymore. It's a proven fact, no matter how colourful of an interpretation anyone wants to give to the data provided.

Regards.

 

[TheHerald]

But also slower. You cannot separate the the matter of performance and efficiency. The i9-14900T shouldn't be compared to the R9 7950X, because they differ so drastically in both performance and efficiency! Nobody who's in the market for one would also be considering the other.

I've talked about this before, as a "mismatched comparison" that you seem to like making.

But you don't seem to bother when the mismatched comparisons are done the other way, do you? I agree ISO power are the only relevant comparisons - the problem is when you do that Intel wins, in both MT and ST efficiency - in the vast majority of segments. That's just a fact.