Intel announces the Core i7-12900KS: higher clocked Alder Lake

Quizzical

With AMD's Ryzen 7 5800X3D on the way, Intel apparently felt like they had to respond.  The outcome is the Core i9-12900KS, which officially launches on April 5.  Basically, if you take a Core i9-12900K, increase the nominal base TDP and stock clocks, and add 300 MHZ to the top turbo, you get the Core i7-12900KS.

I don't doubt that they an increase the base TDP and base clocks.  That's easy enough to do, though it now means that the Core i7-12900KS will officially burn more power without turbo than the Ryzen 7 5800X3D or Ryzen 9 5950X do at max turbo.  As with the 12900K, the Intel's turbo power will be 99 W greater than AMD's, at 241 W as compared to 142 W for most of AMD's desktop lineup.

But no one is paying an extra $150 for higher base clocks from a higher nominal TDP for parts that are going to be run with infinite turbo enabled anyway.  Adding 100 MHz to the max E-core turbo also merits a yawn, as the point of E-cores is precisely for the situations when you don't need high end performance.  No, the real point of this part is increasing the max turbo from 5.2 GHz to an enormous 5.5 GHz.

And that's where I'm skeptical that they can make it work well.  Intel has done this before, you know.  They called it the Core i7-8086K.  That was basically a Core i7-8700K with the max turbo increased from 4.7 GHz to 5.0 GHz.  For the most part, it performed identically to the cheaper 8700K.  It turns out that taking an aggressively clocked part and increasing the clock speed by another 300 MHz just by binning isn't very easy to do.

And remember that Intel wasn't feeling nearly as much pressure from AMD back then.  AMD's then top of the line, the Ryzen 7 2700X, had a max turbo of 4.3 MHz, as well as significantly lower performance per clock cycle.  It did have eight cores to Intel's six, but if you want faster cores, it hadn't even caught Intel's previous generation Core i7-7700K.  So Intel could clock the 8700K how they wanted, and still didn't leave themselves a comfortable 300 MHz of headroom.

Now though?  Intel's mainstream desktop CPUs were clearly trailing AMD's in any efficiency metrics you like before the launch of Alder Lake.  That led to pressure to clock the Core i9-12900K very aggressively in an effort at reclaiming the crown of fastest mainstream desktop CPU, which they more or less did in many but far from all workloads, and at the cost of runaway power consumption.  With lower bins available to eat up lesser performing chips, they could afford to set the 12900K specs fairly aggressively, and they did.  So do I really believe that they left themselves 300 MHz of headroom on anything in meaningfully higher volume than a press edition part?  No, no I don't.

It's also notable that nominal max turbo means less than it used to.  With some AMD part (possibly the Ryzen 9 3950X, though I'm not sure of that), there were a lot of complaints that a lot of the CPUs weren't reaching AMD's listed max turbo.  Of course, some where going above the nominal max turbo, even at stock settings.  AMD generally listed max turbo in multiples of 100 MHz, but the CPU clock speeds would run in multiples of 25 MHz, and different nominally identical chips in different motherboards would commonly top out at different max clock speeds.  I haven't seen reports that Intel breaks up their turbo that finely as opposed to 100 MHz bins.

I could certainly believe that Intel will bin out the best Alder Lake dies for use in the 12900KS.  I could even believe that it will commonly clock a little higher than the 12900K, even while staying inside of the same 241 W PL2 power cap.  But 300 MHz higher, as they want to imply?  Nope.  Not going to happen, or at least, not in typical use.  Maybe if you use exotic cooling, but not if you just stick it in a nice motherboard with a nice air or water cooler.

And so I find it interesting that Intel felt the need to launch this in response to the Ryzen 9 5800X3D.  Unlike Intel's launch, AMD's is a genuinely new part, not just a higher clock speed.  For that matter, the 5800X3D has a lower clock speed than the 5800X, and isn't overclockable.  Rather, AMD's chip has 96 MB of L3 cache rather than the standard 32 MB.

I've previously expressed skepticism that the 5800X3D will be as good of a part as AMD is claiming.  32 MB of L3 cache is already quite a lot, after all, and it's more than the Core i9-12900KS has.  For that matter, the previous generation Core i9-11900K had 16 MB, and we're not that far removed from the Core i7-7700K being a fast CPU with 8 MB of L3 cache.  Once you've got plenty of cache, the gains from adding even more are very limited.

It's not that the 5800X3D will be worthless.  Far from it.  There will be quite a few workloads out there where it's the fastest CPU that money can buy.  I just don't think that will be as common as AMD wants gamers to think it is.  But it's apparently a real enough thing that Intel felt the need to respond with what sure looks like a marketing stunt to shove something out the door a little before the 5800X3D arrives later in April.