Cinebench 20 and 23 Results

[iGPU]

Thanks for your hard work with this kext and the patches. I started using them on my Z690 build (ASUS Max Extreme / i9 12900KF / 64GB DDR / CPU iCue 360mm LCD AIO / Vega 56 for Mojave and 6900XT for Big Sur & Monterey).

The Cinebench tests were done in Big Sur (11.6.5). I'd run scores before switching from CPUFriend and CPUFriendDataProvider to your CPUTopologyRebuild. The boot-arg -ctrsmt was used and the ProvideCurrentCpuInfo Quirk was set to true for all boots. Also, I had active all of your patches, as appropriate for Big Sur and my i9 for both sets of tests as described in the next paragraph.

Single core scores went from almost 1900 (as shown below) when CPUFriend and CPUFriendDataProvider were used, to 985 when CpuTopologyRebuild was substituted for those other two. I found similar 50% reductions in Cinebench 20 multicore scores (not shown).

I'm wondering my experience is so different from what you described on your Performance page.

[b00t0x]

@iGPU CpuTopologyRebuild.kext is not a replacement for CPUFriend / DataProvider, so I think these kexts can be used at the same time.
I didn't test mixed kexts configuration because I'm currently using iMacPro smbios that does not need CPUFriend, but MacPro smbios probably needs CPUFriend to get better performance.

[iGPU]

Thanks for reply. From the Readme.md file, it was not clear that all 3 were to be used simultaneously: I thought that CPUTopologyRebuild was instead of the other two.

When all 3 are run simultaneously, Single Core in Cinebench R23 gives an 1824 result, similar to the 1895 that I posted above, with Multi-Core coming in at 27543 (AI Optimization and XMP-1 are activated in BIOS) . And these results were without using "-ctrsmt".

GB5 now has a Single Core score of 2118 with a Multi-Core score of 19314, while reporting 16-cores with 24 threads. (And again, these results were without using "-ctrsmt".)

(My i9 12900KF was specially binned for all cores at 5.3GHz.)

---

Adding -ctrsmt did not significantly affect the scores, but reported 8-cores with 24 threads.

[aryanrajseo]

Thanks for your hard work with this kext and the patches. I started using them on my Z690 build (ASUS Max Extreme / i9 12900KF / 64GB DDR / CPU iCue 360mm LCD AIO / Vega 56 for Mojave and 6900XT for Big Sur & Monterey).

The Cinebench tests were done in Big Sur (11.6.5). I'd run scores before switching from CPUFriend and CPUFriendDataProvider to your CPUTopologyRebuild. The boot-arg -ctrsmt was used and the ProvideCurrentCpuInfo Quirk was set to true for all boots. Also, I had active all of your patches, as appropriate for Big Sur and my i9 for both sets of tests as described in the next paragraph.

Single core scores went from almost 1900 (as shown below) when CPUFriend and CPUFriendDataProvider were used, to 985 when CpuTopologyRebuild was substituted for those other two. I found similar 50% reductions in Cinebench 20 multicore scores (not shown).

I'm wondering my experience is so different from what you described on your Performance page.

could you share your new cinebench result with multicore? your result is closer to my 12400f cpu, i assume you may need to recheck the SMBIOS.