Monday, November 1st 2021

Intel Core i9-12900K 36% Faster Than Stock in Maximum Turbo Power Mode

The recently announced Intel Core i9-12900K is set to launch on November 4th alongside the i7-12700K, and i5-12600K which is the date when we will see reviews for the processors released. We have already seen numerous leaks regarding the processors performance leading up to the announcement and we have now received some new leaked performance figures for the processors when operating in Maximum Turbo Power (MTP). The MTP is defined by Intel as the maximum sustained power dissipation of a processor compared to TDP which is the base power draw. The performance difference between these two power modes has been revealed from Cinebench R20 multi-threaded results posted by Wofstame the Gaming Desktop Product Planning Manager for Lenovo China.

The Intel Core i9-12900K scores 7492 points when running at its TDP of 125 W and 10180 points or 36% faster when operating at the MTP of 241 W. This performance difference is less notable for the other processors with the Core i7-12700K seeing a 30% improvement between its 125 W and 190 W power modes while the Core i5-12600K sees a 10% improvement from the 125 W TDP to 150 W MTP. Intel appears to be extracting the maximum performance from their Core i9-12900K with diminishing returns from the increased power budget compared to the other processors.
Source: @9550pro
Add your own comment

120 Comments on Intel Core i9-12900K 36% Faster Than Stock in Maximum Turbo Power Mode

#26
docnorth
Vayra86Probably because any higher is not a good idea given the IHS solution they implement. Its thin. Also, there is the problem of heat concentration, even with the thinner IHS/paste/solder solution.

If they can't dissipate fast enough, you can run more power through the chip, but it'll just not work within safe parameters even in the typical 24/7 Intel test bed of water chiller plus nuclear plant.

Aaand... boy what a surprise, Intel chips still get hot on the fabled 10nm that is ALSO clocked way out of its comfort zone like 14nm was. You don't say.
This makes sense. Also performance gains might shrink further beyond this limit.
Posted on Reply
#27
Pumper
Bruno Vieirathe price says everything intel thinks about the 12900k. A bit faster than the 5900X and not as close to the 5950X.
36%+performance for 100%+ power? Sure no.
If the little cores were as good as they say, they would do a 4P + 24E design with the same area.
How is that any different from AMD? A 5950X with all core overclock needs +160W at full load (over the stock ~140W) and only gains +20% performance www.kitguru.net/components/cpu/luke-hill/amd-ryzen-9-5950x-zen-3-cpu-review/17/
Posted on Reply
#28
persondb
Reminder to all, this isn't stock behavior but basically when power limits are raised to maximum. This is very similar to what AMD does with PBO(precision boost overdrive) which raises the limit of PPT/TDC/EDC to whatever you want(within motherboard set limit).
The Quim ReaperAs someone who much prefers traditional CPU coolers over AIO's, you have to ask if there is any current tower cooler out there that can cope with a 12900K at full power..The (supposedly) best of the best, the Noctua DH15 has a TDP limit of 250w, so at 241w, its going to be right on the edge of thermal throttling territory, in 100% CPU load productivity work like rendering and encoding, at least Gaming will be no problem, I suppose, as that never pushes any CPU to all core 100% load for anything other than brief periods for stuff like shader compilation.
This is basically an overclocked 12900K though. Not really overclocked since it's still in specs, but the default behavior is PL1 = 125W and PL2 = 241W.
Which means that the 12900K will turbo for a couple of seconds and then go back to the sustained power draw(PL1). You shouldn't worry about that unless you plan to set PL1 to 241W.
You can see the behavior in this picture.

Posted on Reply
#29
Zubasa
persondbReminder to all, this isn't stock behavior but basically when power limits are raised to maximum. This is very similar to what AMD does with PBO(precision boost overdrive) which raises the limit of PPT/TDC/EDC to whatever you want(within motherboard set limit).



This is basically an overclocked 12900K though. Not really overclocked since it's still in specs, but the default behavior is PL1 = 125W and PL2 = 241W.
Which means that the 12900K will turbo for a couple of seconds and then go back to the sustained power draw(PL1). You shouldn't worry about that unless you plan to set PL1 to 241W.
You can see the behavior in this picture.

Actually the stock behavior for the 12-gen K-SKUs is PL2 with no time limit. That is what the yellow text implies.
What you say should be true for the 12900 non-K though.
Motherboard makers actually have the option to go even further than 241W for the K-SKUs.
Posted on Reply
#30
persondb
ZubasaActually the stock behavior for the 12-gen K-SKUs is PL2 with no time limit. That is what the yellow text implies.
What you say should be true for the 12900 non-K though.
Motherboard makers actually have the option to go even further than 241W for the K-SKUs.
Then my bad, I misinterpreted it but that's kind of odd in my view as it seemed to imply that PL1 wasn't equal to PL2 unless the user specifically set it for that.
Posted on Reply
#31
AusWolf
MusselsYeah, as expected they have power and heat issues like recent intel generations. Damn.
Power issues: yes, heat issues: not necessarily. Actually, Ryzen 3000 and 5000 CPUs run much hotter when configured with power limits to match Rocket Lake. Power doesn't equal heat, despite common misbelief.
Posted on Reply
#32
Zubasa
persondbThen my bad, I misinterpreted it but that's kind of odd in my view as it seemed to imply that PL1 wasn't equal to PL2 unless the user specifically set it for that.
IMO Intel being more transparent is a good thing. They are finally acknowledging how their CPUs will be run IRL.
AusWolfPower issues: yes, heat issues: not necessarily. Actually, Ryzen 3000 and 5000 CPUs run much hotter when configured with similar power limits to match Rocket Lake.
A lot of it has to do with the higher density of TSMC N7 vs Intel 14nm, resulting in smaller surface area for heat transfer.
Intel 7 / 10nm ESF is theoretically even denser than TSMC N7, so it remains to be seen.
There are reasons why Intel explicitly states that an effort has been made to aid heat transfer. They wouldn't do all this extra work if it is not necessary.
Posted on Reply
#33
wolf
Performance Enthusiast
Well I don't much sit around gunning for cinebench scores these days, I want to see gaming performance.
Posted on Reply
#34
The Quim Reaper
persondbYou shouldn't worry about that unless you plan to set PL1 to 241W.
Which is exactly what I would do!

My 9900k is overclocked to 5GHz all the time the PC is on. with power saving C-states disabled.
Posted on Reply
#35
Vayra86
persondbThen my bad, I misinterpreted it but that's kind of odd in my view as it seemed to imply that PL1 wasn't equal to PL2 unless the user specifically set it for that.
Hey you don't win benches playing fair now, you need to get the masses into 'euh wut? Don't get it, but I'm sure its better' territory.
Social media will do the rest for you, with a slew of under 100 IQ-Youtubers alongside them.
Posted on Reply
#36
Turmania
Funnily just couple years ago I was pointing finger at AMD for not having efficient cpu and gpu regards to power usage. Now tables have turned.
Posted on Reply
#37
Guwapo77
Vayra86Probably because any higher is not a good idea given the IHS solution they implement. Its thin. Also, there is the problem of heat concentration, even with the thinner IHS/paste/solder solution.

If they can't dissipate fast enough, you can run more power through the chip, but it'll just not work within safe parameters even in the typical 24/7 Intel test bed of water chiller plus nuclear plant.

Aaand... boy what a surprise, Intel chips still get hot on the fabled 10nm that is ALSO clocked way out of its comfort zone like 14nm was. You don't say.
What's 9 watts in the grand scheme of things? It will probably slap 250w anyways. It almost reminds me of silly American prices, a donut costs $1.99 vs saying it costs $2.00. You know, the 250W might be reserved for a future Black model. You know, the model that is guaranteed to best AMD's V-Cache processors.
Posted on Reply
#38
Dristun
The Quim ReaperWhich is exactly what I would do!

My 9900k is overclocked to 5GHz all the time the PC is on. with power saving C-states disabled.
Hopefully nobody from UN Environment Assembly is reading this forum or TPU might be in trouble
Posted on Reply
#39
The Quim Reaper
DristunHopefully nobody from UN Environment Assembly is reading this forum or TPU might be in trouble
Not really, It maybe clocked to the max, but its still just mostly idling, barely using any power (30w-40w) when doing the normal day-day stuff.
Posted on Reply
#40
AusWolf
The Quim ReaperNot really, It maybe clocked to the max, but its still just mostly idling, barely using any power (30w-40w) when doing the normal day-day stuff.
30-40 W at idle is a lot, imo. You could get it down to the lower teens by enabling C-states. What's the benefit of disabling C-states anyway? :confused:
Posted on Reply
#41
Marshal_90
Holy Moly 241 W for 11000 score? Is this just for the CPU or the whole system?

R9 5950x's whole system under load consumes less power than only 12900K CPU!

I thought this was supposed to be a Ryzen moment!
Posted on Reply
#42
Richards
Electricity is cheap... intel wants the performance crown
Posted on Reply
#43
The Quim Reaper
AusWolf30-40 W at idle is a lot, imo. You could get it down to the lower teens by enabling C-states. What's the benefit of disabling C-states anyway? :confused:
That 30-40w was for the entire PC.

CPU is less than 20w
Posted on Reply
#44
AusWolf
RichardsElectricity is cheap... intel wants the performance crown
www.bbc.co.uk/news/business-58746953
The Quim ReaperThat 30-40w was for the entire PC.

CPU is less than 20w
Ah! That's different. My mistake. :ohwell:
Posted on Reply
#45
ViperXTR
Any leaked reviews yet? :)
Wanna see that 12600K go
Posted on Reply
#46
Lord_Soth
I think that the title is misleading, the correct title is:

"Intel Core i9-12900K 36% Slower at Stock without Maximum Turbo Power Mode"

since all benchmark until now are with maximum turbo power ON at 241w, to nearly match 5950X stock 105w TDP.
Posted on Reply
#47
Cobain
Here we go again... 241 watts at what? Prime95? 3d Mark? Cinebench?

Do you even play games?

Boring
Posted on Reply
#49
Crackong
AusWolfPower doesn't equal heat, despite common misbelief.
You mean temperature right ?

Since there is no physical work done within a CPU, all energy input is converting to heat.
So power = heat inside a CPU.
Posted on Reply
#50
Darmok N Jalad
Yes, peak power consumption might not matter in many users’ day-to-day use, but I suspect that even in random workloads, the boosting to one core is going to put similar strain on the power delivery. Basically, motherboards still must be designed to accommodate the CPU’s boosting behavior to achieve that very top performance. That adds complexity, which adds cost. You get to pay for it either way, and I think the real take home is that this looks like the future state of high-end desktop computing. This power boosting can deliver more performance, but we’re already running into a thermal density-cooling limit—one that will likely only get worse with smaller nodes. This is going to be a one-time grab for that extra performance that will require some huge architectural gains to be able to walk back from. I think this is just how CPUs are going to behave from now on.
Posted on Reply
Add your own comment