Intel’s Rocket Lake is expected to provide a 20 percent improvement in IPC, but new rumors suggest that the high-end desktop version of the chip will still consume a lot of power at peak load. The CPU is supposed to run up to 98C with a sustained power consumption of ~ 250W.
According to reports from Chiphell, the Core i9-11900KF was tested with an entry-level 360 mm cooler while performing the AIDA64 FPU stress test. One reason you should consider these rumors with caution is that the AIDA64 FPU stress test is specifically designed to hammer the chip. This is great if your goal is to test equipment and isolate errors. It is less important if your goal is to measure the energy consumption of the normal workload. The AIDA64 FPU test also contains AVX-512 instructions. If you are using this code path for the Core i9-11900KF while the CPU has a frequency of 4.8 GHz, it will generate a lot of heat.
Chiphell image
Allowing the CPU to run as hot as 98C is less than ideal from an enthusiast’s perspective, but it can also demonstrate that Intel is comfortable with allowing its silicon to work at that temperature. Years ago, I had a conversation with an Intel engineer, who told me that the company expected its current Nehalem and Westmere CPUs to work perfectly at 95C in 24/7 operation, and that the chips should be able to withstand these types of temperatures for up to a decade at least. This does not mean that enthusiasts should throw away their expensive coolers and let their chips roast under a $ 15 + fan heatsink, but the fact that the CPU can run up to 98C is not automatically a problem.
It’s not surprising (assuming this rumor is accurate) to see the 11900KF running so hot. As we discussed recently, motherboard OEMs often program enthusiastic cards with system standards that run the CPU at full boost clock for a much longer period of time than specified by Intel. Intel is aware of this and its own guidelines are recommendations, not requirements. Normally, we would expect a CPU running the AVX-512 code to slow down the clock. The fact that it didn’t happen could mean that Rocket Lake keeps its clocks on AVX-512 workloads where other Intel CPUs slow. It can also indicate that the motherboard is configured not to drop the multiplier during AVX-512 when executing the AVX-512 code. The fact that this chip reaches 98C in an unknown configuration does not mean that this will be normal behavior for all processors.
Intel has chosen to take Rocket Lake as far as possible while remaining compatible with previous versions of Comet Lake, which includes defined TDP ranges. As such, it is not much different from AMD’s decision to keep the maximum TDPs of the Ryzen 5000 series identical to their Ryzen 3000 equivalents. Intel can still claim an energy efficiency gain, as long as it completes workloads in less time than 10th generation chips with the same energy envelope.
Rocket Lake is expected to offer strong performance gains in single-threaded workloads and a slightly smaller increase in multi-threaded workloads. Comet Lake has reached the top with 10 cores, while Rocket Lake offers only eight, but the CPI gains are likely to make up mostly or fully. Support for the AVX-512 may even give the eight-core Core i9-11900K some victories over the 10-core Comet Lake, although that depends on benchmark support.
We’ll know more when Rocket Lake launches in late March.
Now read: