Presenting Cascade Lake-X
With Intel’s 10 nm products not yet being ready for the desktop market, they’ve had to get creative with their product releases. In a third processor update for the X299 platform they present us with Cascade Lake-X. It arrives with promises of increased performance through frequency bumps, courtesy of another advancement in 14 nm process technology.
Higher boost clocks
The specification of the Core i9-10980XE is not much changed from its predecessor, the Core i9-9980XE. CPU-Z 1.90.1 even reports it as being the same chip. It keeps the same 3.0 GHz base speed but increases the peak boost clocks by 100 MHz to 4.8 GHz. There are still 18 cores and 36 threads at a 165W TDP. The monolithic die stays soldered to its IHS. With Intel typically being conservative about their advertised frequencies, there is potential for better overclocking results.
- Intel Core i9-10980XE (QS)
- ASUS ROG Rampage VI Extreme Encore (BIOS 0010)
- G.SKILL F4-4000C18Q-32GTZKW (4x8GB)
- EK-MLC Phoenix 360 CLC
- Windows 10 1903, High Performance power plan
The primary reason “Effective Clock” was added to HWInfo from v6.14 was an issue with clock reporting on Cascade Lake-X. When initially performing these tests, it looked like the new processor got nowhere near its rated boost clocks. After investigation, the boost specification seemed to be met based on performance. This meant the clock reporting had an issue. An alternative was found by relying on CPU clock counters instead of the reported clock ratio. An added benefit is that this method reflects all frequency changes during the measurement period instead of reporting time-discrete readings. This includes clock gating, clock stretching and throttling. Using this new method returned results much closer to the expected values and is used for frequency reporting in this article.
Further information: https://www.hwinfo.com/forum/threads...te-clock.5958/
Any benchmark results are the average of three runs. In the single threaded tests, the thread affinity for the benchmark was manually set to the highest ranked core. HWInfo 6.14 was used to record the monitoring information during the run. The average frequency was measured using the “Effective Clock” item. The average power was measured directly from the VRM controller and reports the CPU input power.
Results and Analysis
In the single-threaded test, the clock frequency is very close to its advertised boost during the entire benchmark. When looking at the multi-threaded test the frequency is not as impressive and drops all the way to 3.4 GHz in order to stay within the default power limit. What is impressive is the highest temperature reported during this test, highlighting significant overclocking headroom through the use of soldered TIM (Thermal Interface Material).
Through the overclocked results we can verify that there is indeed plenty of headroom. Reaching a full 5 GHz on all 18 cores is very impressive, even considering the insane power consumption of 512 W during this benchmark. Comparing to the stock results translates in 45% more performance at 218% higher power consumption.
Continuing with Prime95 tests reveals which frequencies could be expected for long term overclocks. This tells a similar story of impressive frequencies at the cost of much increased power consumption. Up to 4.7 GHz when not utilizing AVX instructions is about 200 MHz higher than what could be achieved on a decent Skylake-X processor. When enabling full AVX 512-bit instructions, the frequency had to be lowered to 4.2 GHz to keep stability and prevent overheating. The high performance liquid cooling used just barely prevented the processor from throttling at just under 500 W of processor power consumption.
Original article: https://www.elmorlabs.com/index.php/...z-on-18-cores/