Intel Core i9-12900K, i7-12700K, i5-12600K Specs, Pricing & New TDP!

legcramp

[H]F Junkie
Joined
Aug 16, 2004
Messages
12,396




HUNB_Specs.png
 
Was happy to hear 12900k multicore isn't better than 5950x (supposedly) - been encoding ~200 UHDs non-stop since June, no buyer's remorse of 'if I just held off a little longer maybe this could have gone quicker'. :p
 
I have to say I'm surprised that not even the i9 can muster more than 8 performance cores. That means the top-end i9-12900K has the same number of performance cores as the i7-12700KF that is $205 cheaper. I doubt the efficiency cores are going to have much impact on gaming. It would seem to me that performance is going to tank quickly beyond 8 cores. I guess they are really banking on 8 real cores being all that is needed for a long time - perhaps because current consoles only have 8 cores.

Good luck overclockers! 😈

Intel and AMD both squeeze almost all available headroom out of the chips via boost clocks now. I'm not sure we're going to see anything exciting that respect. About the only chance we have of seeing a chip with actual overclocking headroom again is if Intel gets it's fab issues worked out to the point of actually being ahead of the game again.
 
I have to say I'm surprised that not even the i9 can muster more than 8 performance cores. That means the top-end i9-12900K has the same number of performance cores as the i7-12700KF that is $205 cheaper. I doubt the efficiency cores are going to have much impact on gaming. It would seem to me that performance is going to tank quickly beyond 8 cores. I guess they are really banking on 8 real cores being all that is needed for a long time - perhaps because current consoles only have 8 cores.
I was just asking myself this question since NE just put 12700K up for preorder $449 - wondering how much diff the 4 less E-cores is going to make in gaming/productivity etc.

Maybe 12700K is going to be the sweetspot in this lineup.
 
In gaming, it will be zero difference, at the same clocks (12900k will still be better in gaming, due to higher clocks). In highly threaded work---it will absolutely make a difference.
True that 12900K will run higher frequencies and turbo higher at stock settings. But what madman buys one of these to run stock?

I'd be surprised if 12900K is legitimately binned higher than 12700K rather than merely certified higher. And if past Intel precedent is anything to go by, 12900K or 12700K it will be the same lottery ticket's chance for a high OC on the 8 P-Cores.
 
True that 12900K will run higher frequencies and turbo higher at stock settings. But what madman buys one of these to run stock?

I'd be surprised if 12900K is legitimately binned higher than 12700K rather than merely certified higher. And if past Intel precedent is anything to go by, 12900K or 12700K it will be the same lottery ticket's chance for a high OC on the 8 P-Cores.
The 12900k does have 5MB more L3 cache. Which doesn't sound like much but-----the 10900k's extra 4MB of cache made a noteable difference in gaming. So there could be a similar uplift here, as well. But it would depend on if 25mb still isn't "enough" for 30MB to continue scaling more performance.
*12900k also has 2MB more L2 cache
 
Last edited:
Hmmm, interesting. That 12900KF looks to be a winner in my book. Don't need onboard graphics, but the 8+8 and the 30mb smart cache looks promising.
 
I want to see how the thread director thing impacts performance and how it works on win10.
 
I want to see how the thread director thing impacts performance and how it works on win10.
It's going to be slower with Windows 10 since that OS doesn't have the updated scheduler and will most likely never get it.
 
I don't get the point still for this e cores crap on a desktop. I would prefer if they used the die space for two more cores personally.
This is why the chip has a hardware scheduler. Intel has patented a bunch of interesting stuff for thread / core selection the last few years. I know it sounds crazy... but it is very very possible a chip with a mix of high and lower performing cores will actually be much faster in all scenarios accept for when a chip actually NEEDS more high performing cores. Its possible Intel will do very well vs AMDs chips and even spank them hard in single thread... and only really loose in things that can actually use a ton of HP. So I don't expect them to be winning any rendering benchmarks... but I expect they are going to win gaming bench marks again.

The truth is the vast majority of things our Machines (even crazy hardcore 3090 running gamers) don't really need 5+ghz high performance cores. However the threads that do under Intels setup... will execute on that one fast performance core and clock where it needs to be. This will make for chips that boost those performance cores higher for longer as one easy to understand example. If the OS/IO and your torrent program running in the back ground are executing on 2ghz E cores... they aren't creating waste heat. Which means that super single threaded game your running can enjoy 2 or 3 Performance cores boosting and eating the thermal space rather then your torrents and windows background intelligent transfer. ;) For better threaded games... again the cores that need performance to feed the GPU will get higher boosts on the P cores... while the 3D audio calculation can happen on those slower cores. Having 4 slower cores to do your multi threaded game audio ect is going to be much better then one more P core.

Long term the second and third gen of this tech... assuming this isn't Intels Pentium 5. Intel has patents that further expand on potential hardware scheduling that would include custom bit... and GPU compute. That seems much more complicated... but hey Apple is basically already doing just that. The Apple scheduler makes it so Apple devs don't have to think too hard about how their stuff is accelerated. Intel wants to get there.

We'll see how this works in practice however... that this still needs special support from the windows scheduler is disturbing and doesn't bode well imo. I would have assumed the internal chip power management and thread choice wouldn't have needed anything real special from windows. Anyway looking forward to actual numbers.
 
I don't get the point still for this e cores crap on a desktop. I would prefer if they used the die space for two more cores personally.

Whole industry is heavily leaving this way now with on chip/die specialty processors/clusters.

We got used to it with video decoders/encoders, now you'll see it for everything else

This is just (the start of) basically a PC background task co-processor/cluster
 
125w eh? Insert FX-12000K space heater joke here.

It will be interesting to see the difference the cache makes this gen. For a lot of gaming the i5 and i7 made far more sense than the i9 this past gen. If that i5 12600k stays below $300 retail for the most part, it could be a winner. Why buy even the i7 for gaming if the i5 is right on its heels?
 
I don't get the point still for this e cores crap on a desktop. I would prefer if they used the die space for two more cores personally.
In Enterprise the efficiency cores are going to be a big deal, the 11th gen 4/8's will be churning along happy then toss in some basic background tasks like an AV scan or a windows update or god forbid both then suddenly even excel gets choppy then we get calls and angry emails. a 4/8+4 configuration would completely remove that at way less power usage than a 6/8 configuration of similar dye size with less heat and ultimately better battery or power curves.
I can't speak for all Enterprise but cutting back power requirements is a huge deal for us, we are under constant pressure to reduce our energy usage and 12VO has helped a lot there and these should help more. I am hoping Intel can deliver here because the current champ of the price, performance, energy usage chart is Apple and I don't want to go full apple, but the M1-based AiO systems and mini's are super compelling for management because of a variety of reasons.
 
This is just (the start of) basically a PC background task co-processor/cluster

I'll laugh if we go back to multi-socket/slot coprocessor setups. We'll have "tall" gaming rigs and "wide" media machines...
 
From Anandtech:

The thing with memory support is that it usually quoted for a specific number of modules installed into the system. In this case, Intel is quoting these numbers using one module per channel (technically a 64-bit channel, but more on that later), meaning that these are the supported speeds when two memory modules are supported. The official supported speed changes if you have more memory, double-sided memory, or dual rank memory.

We’ve seen this before – server processors are notorious for having slower support when more memory modules are installed. It turns out the more bandwidth you need, the harder it is to keep that speed with higher capacity memory. It was only until Intel’s 11th Gen Core products that the memory design supported DDR4-3200 regardless of configuration, because sometimes that’s how long it takes to optimize a memory controller. For Alder Lake, DDR4-3200 is also supported in any configuration, but DDR5 changes depending on the memory.

Intel shared this table with us.

If the motherboard has two memory slots total, then the maximum support is DDR5-4800 in any configuration.
If the motherboard has four memory slots total, then the maximum support is DDR5-4400 when two slots are filled with any memory.
If all four memory slots are filled, single rank memory will support up to DDR5-4000.
If all four memory slots are filled, dual-rank memory will support up to DDR5-3600.


DRAMADL_575px.png

So technically Intel listing memory support on Alder Lake as DDR5-4800 is a bit of a misdirection compared to previous launches. If we were to look at parity, two modules in a four-slot board, then really we’d be quoting DDR5-4400. Funnily enough, all of Intel’s benchmarks presented at this launch were run at DDR5-4400, as per specification. Kudos to the testing team to staying within those guidelines.
 
Last edited:
From Anandtech:

The thing with memory support is that it usually quoted for a specific number of modules installed into the system. In this case, Intel is quoting these numbers using one module per channel (technically a 64-bit channel, but more on that later), meaning that these are the supported speeds when two memory modules are supported. The official supported speed changes if you have more memory, double-sided memory, or dual rank memory.

We’ve seen this before – server processors are notorious for having slower support when more memory modules are installed. It turns out the more bandwidth you need, the harder it is to keep that speed with higher capacity memory. It was only until Intel’s 11th Gen Core products that the memory design supported DDR4-3200 regardless of configuration, because sometimes that’s how long it takes to optimize a memory controller. For Alder Lake, DDR4-3200 is also supported in any configuration, but DDR5 changes depending on the memory.

Intel shared this table with us.

If the motherboard has two memory slots total, then the maximum support is DDR5-4800 in any configuration.
If the motherboard has four memory slots total, then the maximum support is DDR5-4400 when two slots are filled with any memory.
If all four memory slots are filled, single rank memory will support up to DDR5-4000.
If all four memory slots are filled, dual-rank memory will support up to DDR5-3600.



So technically Intel listing memory support on Alder Lake as DDR5-4800 is a bit of a misdirection compared to previous launches. If we were to look at parity, two modules in a four-slot board, then really we’d be quoting DDR5-4400. Funnily enough, all of Intel’s benchmarks presented at this launch were run at DDR5-4400, as per specification. Kudos to the testing team to staying within those guidelines.
So that is why the EVGA Z690 DARK K|NGP|N has 2 memory slots while the EVGA Z690 CLASSIFIED has 4.
 
Here's a good video done by Linus tech tips about alder lake and ddr5 benefits. If you can stomach Linus that it.
Seem the answer, we have yet to know (if the first DDR-5 you can buy will make a relevant difference versus the good DDR4 you can buy)
 
I don't get the point still for this e cores crap on a desktop. I would prefer if they used the die space for two more cores personally.

If the E cores really only take 1/4 the space, and if the chart below is really accurate, then I think it makes sense.

ecore_pcore.jpg

https://www.anandtech.com/show/16959/intel-innovation-alder-lake-november-4th/2

It seems like Apple and AMD are clawing (and surpassing) intel in single threaded performance. Intel can make a faster single threaded core to compete, but the performance returns are diminishing. 4x the die size only gives about 28% extra performance. Note that the E-core is no slouch either, it competes with intel's 10th gen single core performance according to the chart.

For single threaded performance you can use the P-cores, but for anything that's parallelizable, E-cores make WAY more sense.
 
I don't get the point still for this e cores crap on a desktop. I would prefer if they used the die space for two more cores personally.
Two major reasons I can think of:

1) More cores, less die space. The efficiency cores are physically smaller as they have less transistors. So you can pack more of them in a given die space. That is why you saw things like ARM going on about 128 core server CPUs and such. Well, for some tasks more core is more better. When things are very parallel and involve a lot of context switching, having more cores can be better than faster cores. This kinda lets you have both. You pack on 8 e cores and maybe they only take the space of 2 p cores (I don't know that is the number just an example). Now you get more parallelism without getting a massive, and costly, die.

2) To help lower power usage. Various governments are getting more and more touchy about computer power usage, and this is a way to reduce it. You can have a system that idles lower and uses less power for mundane tasks which, face it, is what most of us do 90% of the time anyhow.

I also suspect as time goes on the numbers may vary. Right now they seem to be doing what most ARM chips do and having the same number of p and e cores. However once they've been out and used for awhile and programs get optimized, we might see cases where desktops have something like 2 e cores and 10 p cores, where laptops might be 8 e cores and 6 p cores or the like.

We'll just have to see how it goes.
 
Two major reasons I can think of:

1) More cores, less die space. The efficiency cores are physically smaller as they have less transistors. So you can pack more of them in a given die space. That is why you saw things like ARM going on about 128 core server CPUs and such. Well, for some tasks more core is more better. When things are very parallel and involve a lot of context switching, having more cores can be better than faster cores. This kinda lets you have both. You pack on 8 e cores and maybe they only take the space of 2 p cores (I don't know that is the number just an example). Now you get more parallelism without getting a massive, and costly, die.

2) To help lower power usage. Various governments are getting more and more touchy about computer power usage, and this is a way to reduce it. You can have a system that idles lower and uses less power for mundane tasks which, face it, is what most of us do 90% of the time anyhow.

I also suspect as time goes on the numbers may vary. Right now they seem to be doing what most ARM chips do and having the same number of p and e cores. However once they've been out and used for awhile and programs get optimized, we might see cases where desktops have something like 2 e cores and 10 p cores, where laptops might be 8 e cores and 6 p cores or the like.

We'll just have to see how it goes.
I believe your right about future chips changing up the core ratios. I would however suggest it will likely be the reverse. I believe we are much more likely to get chips more like the ARM setups. Expect e cores to outnumber the performance cores. The sell from someone like Intel would be the increased power envelope for the P cores. The e cores are only going to improve. I expect a second or third gen consumer part to have something more like 6/12. Which may make a ton of sense in the not to far off future. Desktop chips that can perform all the basic day to day stuff in a ultrabook power envelope.... and crank up the P cores only when needed, but have a ton of cooling head room. So they will be able to put a lot beefier P core in then they would if they where designing another chip.

What a lot of people don't consider is power envelope is a BIG consideration for everyone including AMD when designing a core. If the E cores can get 95% of the everyday stuff done... that allows them to but some real serious P cores in that may suck a ton of power per core... but may only fire up a few of them. Its going to be an interesting marketing sell. :)
 
For many of us this news boils down to this: "So what does it do for my 4k gaming experience? Nothing, you say!?!? Thank You Sir or Madam."

Now if Intel had announced they had shipped in quantity their new 3808 Ti Super Killer at $800, well, then I'd be replying to this thread even *[H]arder* ;)
 
This is why the chip has a hardware scheduler. Intel has patented a bunch of interesting stuff for thread / core selection the last few years. I know it sounds crazy... but it is very very possible a chip with a mix of high and lower performing cores will actually be much faster in all scenarios accept for when a chip actually NEEDS more high performing cores. Its possible Intel will do very well vs AMDs chips and even spank them hard in single thread... and only really loose in things that can actually use a ton of HP. So I don't expect them to be winning any rendering benchmarks... but I expect they are going to win gaming bench marks again.

The truth is the vast majority of things our Machines (even crazy hardcore 3090 running gamers) don't really need 5+ghz high performance cores. However the threads that do under Intels setup... will execute on that one fast performance core and clock where it needs to be. This will make for chips that boost those performance cores higher for longer as one easy to understand example. If the OS/IO and your torrent program running in the back ground are executing on 2ghz E cores... they aren't creating waste heat. Which means that super single threaded game your running can enjoy 2 or 3 Performance cores boosting and eating the thermal space rather then your torrents and windows background intelligent transfer. ;) For better threaded games... again the cores that need performance to feed the GPU will get higher boosts on the P cores... while the 3D audio calculation can happen on those slower cores. Having 4 slower cores to do your multi threaded game audio ect is going to be much better then one more P core.

Long term the second and third gen of this tech... assuming this isn't Intels Pentium 5. Intel has patents that further expand on potential hardware scheduling that would include custom bit... and GPU compute. That seems much more complicated... but hey Apple is basically already doing just that. The Apple scheduler makes it so Apple devs don't have to think too hard about how their stuff is accelerated. Intel wants to get there.

We'll see how this works in practice however... that this still needs special support from the windows scheduler is disturbing and doesn't bode well imo. I would have assumed the internal chip power management and thread choice wouldn't have needed anything real special from windows. Anyway looking forward to actual numbers.
I feel like at this level if you're going to torrent you should have a remote server or at least a server in a closet or something.
 
  • Like
Reactions: ChadD
like this
Back
Top