With the Core i9-12900KS, Intel is looking to one-up their Alder Lake offerings—no doubt to preempt AMD from trying to take the gaming performance crown with the Ryzen 7 5800X3D, which releases later this month. While the new Ryzen comes with a new technology in the form of the new 3D V-Cache, the 12900KS is fully based on the Core i9-12900K with the addition of some additional boost technologies we've seen before on earlier Intel processors. The most immediately visible change is the Core i9-12900KS now ticking at up to 5.5 GHz—an impressive 300 MHz higher than the 12900K and a whole gigahertz higher than the 5800X3D, though you shouldn't just compare MHz like that. In order to achieve such a high boost frequency, Intel is bringing Thermal Velocity Boost and Adaptive Boost to Alder Lake. In order to properly support these new technologies, all our testing was done with an updated BIOS that includes Intel's 0x1F microcode update designed specifically to support the Core i9-12900KS.

Overall, averaged over our 38 application benchmarks based on a mix of various workloads from all segments, single-threaded, mixed-threaded, and highly-threaded, we find the Core i9-12900KS 4% faster than the Core i9-12900K. To be honest, I expected more. Back when we tested the Core i9-9900KS vs. 9900K, the difference was 10%. These gains aren't totally unexpected, though. 5.5 GHz vs. 5.2 GHz is a 5.7% increase, and that's only looking at the "up to" maximum turbo boost rating that has several asterisks attached to it, more on that later. As meager as those gains sound, the 12900KS is without any doubt the fastest gaming processor available. It's also surprisingly fast in applications. In our aggregated performance ranking, it matches the AMD Ryzen 9 5950X flagship exactly, which is a full 16-core/32-thread design, whereas the 12900KS is a mix of 8 P-cores and 8 E-cores with a total of 24 threads only. These numbers make it clear that the new Golden Cove architecture is a big improvement in terms of IPC.

Not everything ran perfectly, though. In several of our tests, the workload got scheduled onto the wrong cores. We did use Windows 11 for all our testing, which has proper support for the big.LITTLE architecture of Alder Lake. Intel allocated extra silicon estate for "Thread Director," an AI-powered network in the CPU that's optimized to tell the OS where to place threads. However, some tests still showed very low performance. While wPrime as an old synthetic benchmark might not be a big deal, I'm puzzled by the highly popular MySQL database server not getting placed into the P-cores. Fixing this manually is a few clicks only, and a new technology needs time to mature, of course. I'm still quite positive that an expectation of "works 100%" is not unreasonable for this tech. Had those two tests worked better, the 12900KS could even claim the title "the fastest processor we ever tested."

As mentioned before, the 12900KS is the clear gaming performance leader, but the gains over the 12900K are even narrower than in applications. With 2% at the academically important resolution 720p, 1% at the 1080p Full HD people actually game at, and even less at higher resolutions, this is nothing you'd ever notice subjectively. The reason for smaller gains at higher resolutions is that the higher resolution shifts the load, and thus bottleneck, from the CPU to the GPU. Against AMD's offerings, this could still be what it takes to defend the throne. The Ryzen 5 5800X is 13% slower at 720p and 8% slower at 1080p. AMD talks about a 15% gaming performance boost; if we consider cherry-picking and the benchmark mix, it could end up being very close. For you as a gamer, such skirmishes really don't matter much; it's actually a good idea to save some money on the CPU and invest the savings into a faster graphics card, which will yield higher FPS in the end.

Out of the box, the Core i9-12900KS runs at the same PL1=PL2=241 W power limit setting as the Core i9-12900K. Intel does rate the processor higher in their "base power" rating, which seems to be a "guideline TDP" value—150 W vs. 125 W. The reality is that the 12900KS will pull 240 W of power all the time in demanding apps, like rendering, which loads all the cores. For gaming, especially at higher resolutions, heat output will be considerably lower because, like I explained before, gaming is GPU-limited. If you want to use the 12900KS at its full potential, you must pair it with very good cooling, and you'll still see temperatures reaching 90°C and above. The 240 W power limit does actually limit the processor a little bit. When we removed the power limit, we gained another 1% application performance on average. The 12900KS has a TjMax of 115°C, which is a much needed increase over the 100°C of the Core i9-12900K. The only way we could cool the beast in non-stock runs was by undervolting it manually; I picked an undervolt of -0.1 V, which helped avoid thermal throttling in those test configurations.

A slightly different approach to using the Core i9-12900KS is to just let it run into the thermal limit with the heaviest workloads, which are not games anyway. Doing so ensures you're getting the maximum performance your cooling solution allows because Intel has a highly effective, very fine-grained thermal throttling algorithm that only throttles overheating cores, and only by as little as is required to keep them from overheating, checked several times a second. To further reduce temperatures, you could manually reduce the thermal limit or adjust the PL1 and PL2 power limits to fine-tune them to your system's cooling capabilities, but you'd lose a little bit of performance in the process. It is not an unreasonable approach if you don't run rendering applications that load all the cores 100% all the time. Clocks will still peak at 5.5 GHz in low-threaded workloads and games.

In our "Average Clock Frequency vs. Thread Count" test, we got results that confirm Intel's claims of 5.5 GHz with up to two cores active. On my sample, only cores #5 and #6 can reach those clocks, the others will top out at 5.2 GHz even when the only loaded core. Of course, this is exactly how Turbo Boost 3.0 is designed to work. Problematic is that even on Windows 11 with all communication protocols between the OS and CPU active, there are still cases where threads don't end up on these two cores, so a bit of perfomance is lost. Once you go beyond two loaded cores, up to all eight P-cores active, the 12900KS will always run at 5.2 GHz, which is very impressive. Of course, this only works as long as the processor doesn't hit the 240 W power or its thermal limit. If those limits are broken, it will downclock slightly to stay within those constraints, which is a good thing.

I also did a round of testing with a manual all-core overclock pushing things to the max. Compared to other CPUs, my overclocking approach was a bit different as the limiting factor is our cooling system, not the CPU's maximum MHz or voltage. I first picked the highest voltage my cooling could handle and then searched for the highest clock frequency that was stable without overheating the processor. I reached 5.3 GHz "almost stable," which of course isn't good enough, but couldn't add more voltage without the CPU throttling due to overheating, which would lower the clocks below 5.3 GHz. So I had to settle for 5.2 GHz, which allowed me to bring down the voltage a bit while staying stable, leading to a load temperature of 104°C, 10°C below what I originally expected. Using a decent watercooling solution, the results were similar—the limit wasn't the heat capacity of the watercooler either, but rather the heat transfer from the tiny die pushing 350 W or more into the heatspreader and block. With custom watercooling and 1.4 V+, I'm sure 5.4 GHz all-core is in reach. The Core i9-12900K reached 5.0 GHz all-core in our review, and this 12900KS reaches 5.2 GHz, which is strong evidence that Intel is binning these chips and only uses the best ones for the KS processor. All-core manual overclocking isn't ideal for general usage, though. You'll miss out on the 5.5 GHz boost on up to two cores/four threads active.

Intel wants $750 for the Core i9-12900KS, which is a hefty $150 premium over the Core i9-12900K. There's simply no way this can be justified with the gains over the 12900K. It looks to me like Intel is pricing the 12900KS similarly to the Ryzen 9 5950X, which used to cost $750 a while ago, but even the 5950X has come down in price and sells for $670 these days; it was even $600 in early March. If you are in the market for such an expensive high-end CPU, definitely consider your workloads. If you're rendering all day or do something else that fully loads all cores, the Ryzen 5950X will be the better choice, also because it offers better energy efficiency, which lowers your power bill and the heat dumped into the room. Should your workloads be mostly single-threaded or scaling up to several threads but not beyond, with occasional spikes that fully load all cores, Alder Lake is a fantastic choice thanks to the IPC muscle of the Golden Cove architecture. Unless you absolutely need the fastest execution times, it'll be worth considering cheaper models from the Core i5 and Core i7 family. Spend the savings on a memory upgrade, bigger SSD, or more powerful graphics card. If you're building a super-high-end gaming powered by the RTX 3090 Ti and have a thousand dollars left, the Core i9-12900KS is probably what you should add next, and with that done, check out watercooling options.