Intel Core i7-1065G7: A 65 W Ice Lake CPU?

[Redaktion]

The upcoming Ice Lake processor recently posted an impressive set of Geekbench scores, and now a set of SiSoftware benchmark results point to the CPU reaching up to 65 W. The quad-core processor will have a 3.9 GHz boost clock along with 2 MB of L2 cache and 8 MB of L3 cache too.

https://www.notebookcheck.net/Intel-Core-i7-1065G7-A-65-W-Ice-Lake-CPU.428176.0.html

[Peter Quiring]

8 threads simultaneously?  HT does NOT run the threads at the same time, it just improves performance when switching between 2 threads per core.  Intel uses SMP to expose HT to the OS though so it appears to have 8 threads (cores) but it's not true.

[GetSmart]

I think SiSoft Sandra is reading it wrong or not fully recognizing the new Intel SoC. Should be around 15W to 25W.

[Valantar]

What an odd article. First off, it's well documented that quad core Intel U-series CPUs turbo up to >50W during short term turbo. 65W is probably not outlandish for this. I've often seen the i7-8650u in my Latitude 2-in-1 (which seems capped at 12W for long-term power draw) pull more than 45W in short term turbo power. In other words, while the SiSoft number is high, it's not abnormally high - and not a number for sustained power draw, but rather for peak short term turbo power.

Secondly, the article speaks of "improved efficiency at lower clocks", which is exactly the opposite of what a lower base clock indicates. The lowered base clock indicates that it can't sustain the same clocks as the previous generation at 15W. In other words, efficiency at lower clocks is worse.

[Valantar]

8 threads simultaneously?  HT does NOT run the threads at the same time, it just improves performance when switching between 2 threads per core.  Intel uses SMP to expose HT to the OS though so it appears to have 8 threads (cores) but it's not true.

I guess that depends on your definition of "simultaneous", but no, HT does indeed run two concurrent threads per core. The entire point of HT is to allow resources not utilized by the main thread in a core to be utilized by a second thread. While the second thread often has to wait for the first to move on (unless the workload is specifically created to utilize leftover resources from a predefined primary thread), they are nonetheless concurrent.