Intel Core i9-13900K

[KarymidoN]

89ºC at gaming (stock) bruh... NH-U14S got stomped.

 

[Hossein Almet]

EDITOR CHOICE, really?

 

[fevgatos]

Oh FFS these are TDP levels, do you think setting 253W or 241W TDP will get you the exact same power draw

YES. It will. I'ts not TDP, it's a power limit. If you set it to 250w, it will draw 250w, unless you do something wrong with the loadlines. It literally cannot pull more than the power limit you set.

 

[Crackong]

I'm referring to the results at reduced TDP. The i9-12900K at a power limit of 125W scores about 24000 points at 125W.

'Reduced TDP' isn't equal to 'Power Limit'
In HWB's review They specifically mentioned it is 'Power Limited'
So the 13900k can't boost beyond the limit.

We all know TDP means really nothing for modern CPUs

 

[watzupken]

Hard pass for both Zen 4 and Raptor Lake for me. I don't need more heat in my home. Intel once again takes the crown with regards to performance (mostly), with 2 additional awards in high power and temperature.

 

[watzupken]

Editor's Choice, really?

No award, no review units ever again. Jokes aside, it is a fast chip if you are willing to ignore the ridiculous power draw and temperature. I don't believe you can get this to run cooler without delid.

 

[Tomgang]

Oh my God. Meteor lake and zen 4 is to hot, to power hungry and not so fast that i find it worth upgrade from Zen 3.

Hornestly I am glad I just chose zen 3. Yes it's not as fast. But it also less power hungry and runs much cooler. I am keeping my 5600X and 5950X for a while more. They are at least easy to aircool.

Bofh cpu and gpu these days just seems to be pushed to the max and way out of there efficiency range. Not the direction I like. Sure it's cool with fast hardware, but it is uncool to run 90C+ and eat 200 watt+. That is just nuts.

 

[Solid State Brain]

'Reduced TDP' isn't equal to 'Power Limit'
In HWB's review They specifically mentioned it is 'Power Limited'
So the 13900k can't boost beyond the limit.

We all know TDP means really nothing for modern CPUs

I changed that to "power limit" just before you replied because I expected something along these lines as an answer.

To clarify: I meant power limits. On Intel CPUs you change the power limits and the CPU draw no more than what has been set unless, as fevgatos mentioned above, the so-called DC Loadline has been incorrectly set.

 

[Toss]

what is this 117C ?
what is throttle temps in those garbage

 

[neatfeatguy]

EDITOR CHOICE, really?

At least it didn't say "Highly Recommended". I would have scoffed at him if it said that. The CPU is faster than it's predecessor, but I personally am not impressed with it, nor with AMD's offerings.

Perhaps I'm just biased since I moved to a 5900x and a 3080 and the performance gains compared to what I have and need them for is kind of lack luster. I see no reason to upgrade to anything new that's come out in terms of CPUs or GPUs.

 

[64K]

Strictly from a gamers perspective. Very small FPS gains over an i5 13600k in games and costs $270 more. I don't pay any attention to Synthetics scores because I don't play Synthetics. I play PC games.

 

[oxrufiioxo]

Both flagships are meh this generation. They both can be impressive at times but even with my 360 CLC I'd cap either of them to 125w-180w.

 

[fevgatos]

It's true I'm afraid. Power limit = The maximum power allowed. That's just a fact.

 

[HD64G]

Efficiency scaling 13900K vs 7950X

 

[xorbe]

That 5800X3D though, 99.1% at 4K gaming and #2 efficiency, while 13900K #1 power usage for 100%.

 

[fevgatos]

Efficiency scaling 13900K vs 7950X

View attachment 266328

The graph is a failure though

 

[W1zzard]

Why are you testing on windows 11 21H2 when Intel specifically said that 22H2 has the new core scheduler for 13th Gen along with all the improvements to thread director and then complain in your pros cons that cores get put on wrong workloads. wtf.

22H2 was not available at the time of testing in summer. Also it's so new, I'm sure there's plenty of bugs, maybe some that hurt AMD, so I wanted to keep things fair.

Only one workload went to the wrong cores (virtualization)

I'll test 21H2 vs 22H2 soon in an in-depth article

 

[rv8000]

It’s actually horrifying to see the power usage and temps on this, I wasn’t expecting the efficiency to be that much worse

I see a lot of apologists in here who were lambasting AMD for 95c and increased power usage, 13900k is on another level.

I was thinking about a 13600k, but it looks like holding out for the 7800x3d or whatever models end up getting the 3d vcache treatment is the way to go

 

[W1zzard]

For some reason it appears that the temperature limit was increased for this review even though the default should be 100 °C.

Correct, so we can figure out the real temp these CPUs will run at without throttling

 

[Solid State Brain]

It's true I'm afraid. Power limit = The maximum power allowed. That's just a fact.

To be fair, an additional point of confusion here (besides that almost nobody appears to be aware that on Intel the reported Package Power can be accurate, if the DC loadline has been correctly configured) is that TechPowerUp's "CPU only" power draw figures are taken from the motherboard's 12V power connectors. So they are not just for the CPU chip only, but they include at the least also the VRMs, whose efficiency will vary with CPU load, VRM temperatures and of course motherboard design. Unless these variables have been factored out in some way.

 

[W1zzard]

To be fair, an additional point of confusion here (besides that almost nobody appears to be aware that on Intel the reported Package Power can be accurate, if the DC loadline has been correctly configured)

Just less wrong if you tweak loadline, and there's no way to know by how much unless you actually calibrate your measurements, which cannot be done unless you solder around on the motherboard

TechPowerUp's "CPU only" power draw figures are taken from the motherboard's 12V power connectors. So they are not just for the CPU chip only, but they include at the least also the VRMs, whose efficiency will vary with CPU load, VRM temperatures and of course motherboard design. Unless these variables have been factored out in some way.

Yup, we measure with high-precision, high-resolution test equipment on the 12V power connectors. Data collection happens on a separate machine, not the tested machine

 

[fevgatos]

To be fair, an additional point of confusion here (besides that almost nobody appears to be aware that on Intel the reported Package Power can be accurate, if the DC loadline has been correctly configured) is that TechPowerUp's "CPU only" power draw figures are taken from the motherboard's 12V power connectors. So they are not just for the CPU chip only, but they include at the least also the VRMs, whose efficiency will vary with CPU load, VRM temperatures and of course motherboard design. Unless these variables have been factored out in some way.

But that's easy to check with the VID, is it not? Most high end motherboards should be pretty accurate at auto LLC settings with matched VID and vcores.

 

[Dirt Chip]

wut? guru3d is showing power consumption of 368w at a multithreaded load, and a ~38k CB23 score, that's pretty much in line with HUB's findings

GURU3D use all system power, not just CPU: "Keep in mind that we measure the ENTIRE PC, not just the processor's power consumption."

 

[clopezi]

I would love to see how Noctua NHD15 performance with this CPU's, for many users it's our reference heatsink... and many times better than liquid cooling...

On the other hand, looks like DDR5 has no big advantage here. Looks like my DDR4 3200 that I would love to change, will live to see another day...

 

[Solid State Brain]

But that's easy to check with the VID, is it not? Most high end motherboards should be pretty accurate at auto LLC settings with matched VID and vcores.

Yes, the simplest way is tweaking the DC Loadline until VID (a calculated value) and vcore (actually measured CPU voltage) match as close as possible in a multithreaded workload, which can take some time. If not for more or less correct CPU chip-only power measurements, at least this would ensure that power limits are actually working as intended.

From what I've observed with recent middle-range motherboards, VID and vcore almost never match with stock settings, and some motherboards may set completely wrong DC Loadline values depending on other firmware settings.