Intel didn’t just detail its new Ice Lake architecture at Computex this week, it also announced an upcoming CPU — the Core i9-9900KS. This chip is an eight-core spin on the 9900K, but with an even higher all-core base clock of 5GHz under load. That’s a 300MHz boost over the standard Core i9-9900K, with a 400MHz higher base-clock.
We’ve charted how the CPU will drop into Intel’s overall stack, with comparative parts from AMD listed. We’ve also included the Ryzen 7 2700X, for reference. Prices marked with an ~ are taken from retail pricing as of 5/28. The asterisk near the 9900K’s TDP is a reference to the fact that this chip has been demonstrated to only be a 95W processor if confined to a 3.6GHz operating clock. At higher speeds, the CPU’s power consumption rises significantly. At full load, the existing version of the Core i9-9900K can draw 160W – 180W. Increasing clock speed to an all-core 5GHz, assuming the chip holds this frequency at all times (as it should), will likely break the 200W threshold — more, if the voltage taps required to hit that frequency are significant.
Reviewers who saw the Intel Core i9-9900KS demo reported that it used high-end, but ultimately standard cooling components, with a 240mm ROG Ryujin 240 cooler. That’s a high-end system, but not a crazy one. Those fearing a repeat of the 28-core CPU with a 1.2kW chiller attached to it under the table need have no worries.
This CPU will be available at retail, with fully functional integrated graphics, at an undisclosed price. The overall positioning here seems clear. Intel is going to push the 9900K right to the edge to give it a better chance of beating past the Ryzen 7 3800X in a core-to-core comparison. The chances of matching AMD’s 12-core are slim unless that chip proves to have truly terrible scaling due to memory bandwidth pressure (and to be clear, that’s not an outcome we expect). While we wouldn’t be surprised to see the Ryzen 7 12-core scaling less-effectively than its Threadripper counterpart, the jump from 8 cores to 12 increases overall resources by 1.5x. Even with weak scaling, that can still work out to a significant performance improvement.
So why does it behoove Intel to go this route? A little bit of marketing and a little bit of strategy. An all-core 5GHz is a pretty powerful marketing claim, and not all applications scale well past eight cores. In fact, the general rule of thumb you can rely on is that the higher the core count, the worse the scaling (outside of so-called “embarrassingly parallel” workloads that scale exceptionally well). Plenty of multi-threaded applications also top out at 4 or 8 threads, because until quite recently, 16-core CPUs were still the sole domain of high-end workstations or servers.
Price is an interesting question. To date, Intel has shown little interest in reducing its per-core pricing below certain thresholds. While the company’s pricing has absolutely improved since Ryzen’s initial launch — an eight-core Intel chip is now literally less than half the price it was in 2016 — the company has, thus far, allowed AMD to claim the price/performance crown, while it fights for the absolute performance crown. We’ll have to see how the 3800X and 9900KS compare in head-to-head competitions before we know which of them will lead overall.
More Information: extremetech.com