News Overclocked Core i9-12900K Hits 5.2 GHz At 330W, Challenges Ryzen 9 5950X

[TerryLaze]

Holy cow that is a lot of power if true...

At that power level may as well just use ThreadRipper for the extra wideness and more memory channels which should still edge out DDR5 at launch. Maybe that's why Intel can't price this thing higher? Hm... Also, TR 5K around the corner, supposedly with the 3D-cache is going to make Alder Lake sweat profusely during it's prime time. That's not even counting Ryzen with the 3D-cache refresh, which I think it's going to be low volume. More like a "we're still here" from AMD, haha.

Day 1 reviews are going to be saucy.

Regards.

It is but leaving the CPU at stock will only lose you a couple of hundred Mhz so performance won't be much lower but power will be extremely lower.
From the 11700k which is the same with the 11900k except for the special turbos that are irrelevant if you run productivity for extended amounts of time.
120W average at stock is turned to 283W because people only look at peak W.
https://www.tomshardware.com/reviews/intel-core-i7-11700k-cpu-review/2

 

[-Fran-]

It is but leaving the CPU at stock will only lose you a couple of hundred Mhz so performance won't be much lower but power will be extremely lower.
From the 11700k which is the same with the 11900k except for the special turbos that are irrelevant if you run productivity for extended amounts of time.
120W average at stock is turned to 283W because people only look at peak W.
https://www.tomshardware.com/reviews/intel-core-i7-11700k-cpu-review/2

It is still important to know both though. Peak Power consumption means the CPU will use that much energy/power for an amount of time if it is allowed to do so, so you need to cool it and power it accordingly. I can agree with the "diverting attention" as they are the bigger numbers, but doesn't make them less relevant.

Testing the CPUs with some friends, we found the typical usage of a 10700K OC'ed without power saving disabled is around 61W when gaming (very un-scientific) and my 3800XT was around 52W with very similar performance using different GPUs; they have a 3080 and I have a Vega64, yet our numbers were basically the same in games where there was a CPU bottleneck. We tested MMOs, so it was a relevant test we don't often see and we were interested.

Anyway, to your point, on average and as long as the CPUs keep their power saving mechanisms enabled (even under OC), they are not going to use so much power. Under torture testing though, it will become noticeable. Hence, most numbers in reviews are effectively worst case scenarios for 90% of people. People doesn't really play "benchmarks". This being said, I do remember Tom's testing WoW at some point to get some numbers and offer some interesting perspective to MMOs performance. I wish they could do such testing again as 99% of people does play MP games. I wonder which other games would be interesting to test though... Hm... Well, I digress here. Point is, power is still important to measure and context is important as always. Still, for Alder Lake, 300W+ is still a lot of power for the amount of performance it is showing; I can't deny it is a bit worrying.

Regards.

 

[Dr3ams]

If your current AMD CPU and GPU run all your games at optimal FPS with ultra settings, then who cares what a next gen. Intel CPU can supposedly do.

 

[Deleted member 14196]

There is no challenge the Intel loses

 

[VforV]

Another OC leak at 400W

Link > Intel Core i9-12900K Overclocked To 5.3 GHz Across All Cores With 1.44V Voltage, Hits Insane 400W Power Consumption In AIDA64 Stress Test

 

[TerryLaze]

Another OC leak at 400W

Link > Intel Core i9-12900K Overclocked To 5.3 GHz Across All Cores With 1.44V Voltage, Hits Insane 400W Power Consumption In AIDA64 Stress Test

The CPU was overclocked to 5.3 GHz across all 8 performance cores. Looking at the benchmarks reported within the CPU-z benchmark, the single-core score saw a 2% increase while the multi-threaded score saw a 1% increase over the 5.2 GHz overclock that leaked out earlier. Compared to Intel's stock Core i9-12900K configuration, the score is a 5% increase in single and 6% increase in multi-threaded performance.

Versus the AMD Ryzen 9 5950X, the overclocked Intel Core i9-12900K at 5.3 GHz scored a 34% lead in single-core and a 4% lead in the multi-core test. The Ryzen 9 5950X offers 33% higher threads than the Alder Lake chip but the main thing to note is the power consumption.

" Compared to Intel's stock Core i9-12900K configuration, the score is a 5% increase in single and 6% increase in multi-threaded performance."
Yeah at that point you are just doing the OC for the lols.
Nobody needs 5% better performance at 300% more power.

But 34% (29% at stock 125W )higher singlethread compared to 5950x and the same multithread well that's something that at least some people will find atractive.

 

[Phaaze88]

Yeah at that point you are just doing the OC for the lols.
Nobody needs 5% better performance at 300% more power.

Sadly, we live in a world chock full of sheep, so it's not just for the lols.
People will put crappy 120 and 240mm AIOs on their cpus because liquid is more better than air, and completely overlook all the shortcomings.
They will also try overclocking and not think twice about power consumption or how fast today's cpus already are without that crap.
Somebody told them, or they found out, that it's more performance, so they want that extra performance.

About 2 more weeks to wait...

 

[Finnegan1616]

At 10nm, Intel still has a lot of room to work with as AMD goes for 5nm and 3nm early.

those are literally just marketing strategies

10nm Intel ~= 7nm AMD
also, they already stated how they expect to be behind until 2025, sooooooo

 

[Finnegan1616]

Holy cow that is a lot of power if true...

At that power level may as well just use ThreadRipper for the extra wideness and more memory channels which should still edge out DDR5 at launch. Maybe that's why Intel can't price this thing higher? Hm... Also, TR 5K around the corner, supposedly with the 3D-cache is going to make Alder Lake sweat profusely during it's prime time. That's not even counting Ryzen with the 3D-cache refresh, which I think it's going to be low volume. More like a "we're still here" from AMD, haha.

Day 1 reviews are going to be saucy.

Regards.

Intel already stated how they expect to be behind until 2025

 

[-Fran-]

Intel already stated how they expect to be behind until 2025

https://www.tomshardware.com/uk/news/intel-ceo-declares-end-amd-gaining-market-share

I don't know what you're going on about, but I have the suspicion your understanding is incorrect. Manufacturing wise, they haven't said that explicitly either. In reality you can argue they may be, but then again, it all comes down to actual measured performance and not paper numbers. Having an inferior process, while disadvantageous, is* by no means the only deciding factor.

Regards.

 

[RodroX]

Sadly, we live in a world chock full of sheep, so it's not just for the lols.
People will put crappy 120 and 240mm AIOs on their cpus because liquid is more better than air, and completely overlook all the shortcomings.
They will also try overclocking and not think twice about power consumption or how fast today's cpus already are without that crap.
Somebody told them, or they found out, that it's more performance, so they want that extra performance.

About 2 more weeks to wait...

Still 2 more weeks indeed, and the suspense is killing me.

 

[artk2219]

How many years and generations did it take AMD to beat Sandy Bridge? Not sarcasm.

Sandy Bridge? Hell, nehalem. They didn't consistently outperform Nehalem per clock until Ryzen. Bulldozer while not a great chip was decent at multitasking, so the general experience wasnt bad when doing many things at once. But if you had a single threaded task you were working on, it would show its age.

 

[lazyabum]

those are literally just marketing strategies

10nm Intel ~= 7nm AMD
also, they already stated how they expect to be behind until 2025, sooooooo

Do long as they are not giving up there's still hope for them, but not us as this technology crunch is way past ridiculous.

 

[PCWarrior]

Even if we are to believe that the Chinese forum post is true there too many unknown variables. First of all, what 12900K? A review sample grade or an early potato engineering sample? Even if we assume it is the latest stepping we definitely don’t know what the quoted power figures really are and under which test they were obtained. While the reported benchmark result is for CPUz it is not necessary that the reported power consumption figure is taken while the cpu was running the CPUz benchmark. For all we know in cpuz it could only be pulling 150W and the 330W figure is the peak power consumption when he tested with small FFT Prime 95 with AVX. Also that 330W power figure could be total system power consumption as measured from the wall. Which makes this whole discussion null and void. Is this really the level of journalism we get now? Google-translated Chinese forum posts without even any evidence like a screenshot or a photo? Ridiculous.

 

[Finnegan1616]

i wouldnt be surprised if Intel charges 5950x price for this.😂

it is faster, at least until AMD releases their next CPU

so deal with it ig
it still has faster single-core at stock speeds, and around the same multi-core

https://www.tomshardware.com/uk/news/intel-ceo-declares-end-amd-gaining-market-share

I don't know what you're going on about, but I have the suspicion your understanding is incorrect. Manufacturing wise, they haven't said that explicitly either. In reality you can argue they may be, but then again, it all comes down to actual measured performance and not paper numbers. Having an inferior process, while disadvantageous, is* by no means the only deciding factor.

Regards.

Have you ever heard of something called trash talking? Here's the real expectation
https://www.theverge.com/2021/7/26/...ated-new-architecture-roadmap-naming-7nm-2025
also
"So this period of time when people could say, 'Hey, [AMD] is leading,' that's over. We are back with a very defined view of what it requires to be leadership in every dimension: leadership product, leadership [chip] packaging, leadership process, leadership software, unquestioned leadership on critical new workloads like AI, graphics, media, power-performance, enabling again the ecosystem. This is what we will be doing with aggressive actions and programs over the next couple of years."

Do long as they are not giving up there's still hope for them, but not us as this technology crunch is way past ridiculous.

I'll assume you meant to say "So" not "Do"

Ofc I still have hope for Intel, I expect them to take the lead back by 2025, but people act like this never happened, as if Intel expects to be completely ahead forever, sure, it's possible, but beyond unlikely
https://www.theverge.com/2021/7/26/...ated-new-architecture-roadmap-naming-7nm-2025

 

[-Fran-]

Have you ever heard of something called trash talking? Here's the real expectation
https://www.theverge.com/2021/7/26/...ated-new-architecture-roadmap-naming-7nm-2025
also
"So this period of time when people could say, 'Hey, [AMD] is leading,' that's over. We are back with a very defined view of what it requires to be leadership in every dimension: leadership product, leadership [chip] packaging, leadership process, leadership software, unquestioned leadership on critical new workloads like AI, graphics, media, power-performance, enabling again the ecosystem. This is what we will be doing with aggressive actions and programs over the next couple of years."

Your interpretation of that paragraph is not the same as mine, clearly. And I'll leave it at that, since I don't like the semantics rabbit hole.

Regards.

 

[Specter0420]

Sadly, we live in a world chock full of sheep, so it's not just for the lols.
People will put crappy 120 and 240mm AIOs on their cpus because liquid is more better than air, and completely overlook all the shortcomings.
They will also try overclocking and not think twice about power consumption or how fast today's cpus already are without that crap.
Somebody told them, or they found out, that it's more performance, so they want that extra performance.

About 2 more weeks to wait...

Power is extremely cheap in many places (free for some renters too) and, depending on latitude and season, the waste heat just helps offset their heating bills anyway. That excuse only really works if you're a poor person or in a developing country, but then what are you doing with one of these rigs?

I had a giant air cooler with two fans, couldn't get above 4.9Ghz all-core at comfortable temps... Upgraded to one of the better AIOs, now I run 5.2Ghz and hit a stability wall, not a temperature wall...

There are plenty of games\sims that need far more CPU performance than is currently available for max settings, even after overclocking. Virtually every single VR flight sim for example (DCS, P3D, MSFS, XP11, and IL2) will be severely limited by CPU when failing to reach 90FPS in VR, much less the 144 headsets can do these days... In VR an extra 5% in single core performance can be the difference between 45FPS and 90FPS or between 60FPS and 120FPS if you know what you're talking about.

When you don't understand why large groups of people do things and assume they are fools, it usually just means you're a fool. See "Dunning-Kruger" for more details.

 

[Phaaze88]

Power is extremely cheap in many places (free for some renters too) and, depending on latitude and season, the waste heat just helps offset their heating bills anyway. That excuse only really works if you're a poor person or in a developing country, but then what are you doing with one of these rigs?

I had a giant air cooler with two fans, couldn't get above 4.9Ghz all-core at comfortable temps... Upgraded to one of the better AIOs, now I run 5.2Ghz and hit a stability wall, not a temperature wall...

There are plenty of games\sims that need far more CPU performance than is currently available for max settings, even after overclocking. Virtually every single VR flight sim for example (DCS, P3D, MSFS, XP11, and IL2) will be severely limited by CPU when failing to reach 90FPS in VR, much less the 144 headsets can do these days... In VR an extra 5% in single core performance can be the difference between 45FPS and 90FPS or between 60FPS and 120FPS if you know what you're talking about.

When you don't understand why large groups of people do things and assume they are fools, it usually just means you're a fool. See "Dunning-Kruger" for more details.

~OWW...

 

[Specter0420]

~OWW...

Ooops Wrong Window?

BTW: I once built and overclocked two identical PCs, i7-920s, but had to go overkill on only one of the PSU due to limited availability. Bronze vs Gold IIRC. The lower efficiency Bronze PSU died violently after 6 years, almost sounded like a gun when off as something (capacitor?) popped. It took the GPU and the PCIe slot it was plugged into with it and the mobo had random issues before it died completely a year or so later.


The Gold PSU has been running the same i7-920 overclocked from 2.6GHz to 3.8Ghz for 14 years now, including a constant 460W at the wall since I started mining recently. It mines with a 1080 and a 1070ti when my son isn't gaming on it. FYI, the cost to load a PSU 24\7 for a couple years is only a couple hundred extra dollars. You undervalue your limbs.

Do you still think the extra $60 was a waste? I didn't just need to replace that PSU, it murder-suicided about $650 worth of components, out of warranty, all together. I'll never get another low efficiency PSU.

 

[Phaaze88]

Bronze vs Gold IIRC

Not just me, but someone would say that there's more to psus than that. There are bronzes that don't blow up catastrophically, and there are golds that do.
Do you remember what the exact models were?

I wore THIS psu down from overclocking a 7820X + 1080Ti - mainly stress testing. I admit, I took it a little too far compared to most with everything I did.
After a little more than 4 years, the system would soft freeze when running a game or gpu oriented benchmark.
I tried switching cables
taking the gpu out and blowing out the PCIe slot
DDU'ing drivers/installing older drivers
clean boot
removing XMP and other overclocks
temperatures were fine - I had a Kraken G12 + Celsius S36 on it
What worked was going into Afterburner and capping the power limit to 50%. I even tried raising it - the furthest I got was 60%.
Switched to a Corsair AX850 and haven't had a problem since.

You still think the extra $60 was a waste? I didn't just need to replace that PSU, it murder-suicided about $650 worth of components, out of warranty, all together. I'll never get another low efficiency PSU.

My earlier post wasn't even about psus, so how'd it get here? I don't claim to know anything about psus - I still have to use reviews and tier lists to get any info.
It was about:
-cooling, and how some people go straight to AIOs literally because they found out it was just better than air, and then they proceed to set themselves up for a subpar experience one way or another.

-today's cpus and gpus are already really fast and power efficient within the limits they're shipped with.
Overclocking does not do as much as it once did generations ago. Sometimes, all one needs to do with a cpu is raise the power limit a little, 'cause the base one was a little too low - like the 65w PL1 on a 11900(non-K), for example.

 

[$h0nuff]

Public relations gone wild! Does anyone else see an odd placement of power consumption commentary shortly before the release of Zen w/ 3d cache?

I've had fun with products from both vendors for many years now. Reading between the lines I'm interested to see if 3d cache + a layer of fill silicon results in higher wattage requirements, clock-for-clock, in AMD's next offering. Especially considering the node is identical save for stacked cache.

 

[Specter0420]

Not just me, but someone would say that there's more to psus than that. There are bronzes that don't blow up catastrophically, and there are golds that do.
Do you remember what the exact models were?

I wore THIS psu down from overclocking a 7820X + 1080Ti - mainly stress testing. I admit, I took it a little too far compared to most with everything I did.
After a little more than 4 years, the system would soft freeze when running a game or gpu oriented benchmark.
I tried switching cables
taking the gpu out and blowing out the PCIe slot
DDU'ing drivers/installing older drivers
clean boot
removing XMP and other overclocks
temperatures were fine - I had a Kraken G12 + Celsius S36 on it
What worked was going into Afterburner and capping the power limit to 50%. I even tried raising it - the furthest I got was 60%.
Switched to a Corsair AX850 and haven't had a problem since.



My earlier post wasn't even about psus, so how'd it get here? I don't claim to know anything about psus - I still have to use reviews and tier lists to get any info.
It was about:
-cooling, and how some people go straight to AIOs literally because they found out it was just better than air, and then they proceed to set themselves up for a subpar experience one way or another.

-today's cpus and gpus are already really fast and power efficient within the limits they're shipped with.
Overclocking does not do as much as it once did generations ago. Sometimes, all one needs to do with a cpu is raise the power limit a little, 'cause the base one was a little too low - like the 65w PL1 on a 11900(non-K), for example.

When you click that little "spoiler" link at the end of your comments, it's number 4. Sorry, I assumed you wrote that. It all adds up now that you tell me you don't know anything about PSUs.

 

[Phaaze88]

When you click that little "spoiler" link at the end of your comments, it's number 4. Sorry, I assumed you wrote that. It all adds up now that you tell me you don't know anything about PSUs.

OH, that thing.
I haven't updated it in a minute - need to do that or get rid of it.

 

[BX4096]

Sure but that doesn't make the 5950x any faster in that bench,
and if you compare the 11900k to the 5950x in game benches as V for V did a CPU that has twice the cores and is $150 more expensive then you can do this 12900k vs 5950x comparison in ashes as well.

Still, 11% worse than 5950X in multi-threaded performance is rather disappointing for a flagship CPU. I know it's priced somewhat cheaper, but it's still expensive as hell and I'm not sure many of us will see a reason to upgrade. I certainly put my upgrade ambitions on hold, either until the prices come down or something better than 5950x (not an AMD fan) comes out.

After all the growing "leak" press, the reality turns out to be a major disappointment...

 

[TerryLaze]

Still, 11% worse than 5950X in multi-threaded performance is rather disappointing for a flagship CPU.

11% worse in what? Show a link because in this article the 12900k is even faster than the 5950x in multi, at least when overclocked.
Also the percentage of people that actually need or even just care about multi is extremely small. It's only the semi-professionals because for anybody else they are never going to use the full multi.

But at 5.20 GHz, Intel's Core i9-12900K can score 851 points in the CPU-Z single-threaded benchmark as well as 11986.9 points in CPU-Z multi-thread benchmark. By contrast, AMD's flagship Ryzen 5950X CPU scores 648 and 11906, respectively.