Intel Nuc 2.0 (Broadwell)

[ammulder]

Sure, it would be great if you could move it to a better home. Thanks!

 

[RehabMan]

Sure, it would be great if you could move it to a better home. Thanks!

Moved: http://www.tonymacx86.com/el-capita...42-guide-10-11-intel-broadwell-nuc-2-0-a.html

 

[ammulder]

Thanks.

FYI, I see that there are Broadwell NUCs available with an m.2 wireless slot that takes an OS X-compatible card. It looks like the NUC5i5MYHE or NUC5i3MYHE can take a Dell DW1560 card (Broadcom BCM94352Z) that is reported to be OS X-compatible.

 

[RehabMan]

Thanks.

FYI, I see that there are Broadwell NUCs available with an m.2 wireless slot that takes an OS X-compatible card. It looks like the NUC5i5MYHE or NUC5i3MYHE can take a Dell DW1560 card (Broadcom BCM94352Z) that is reported to be OS X-compatible.

Too bad they provide only the low-end graphics with the NUCs that have replaceable WiFi.

The HD6000 and Iris models all have soldered WiFi instead of replaceable. A strange/unfortunate decision by Intel.

May be best to wait and see if they make better decisions with the NUC lineup for Skylake.

 

[ammulder]

Yeah, though, from reviews it didn't seem like HD5500 was tremendously slower than HD6000 (despite having half the execution units). It seemed like the theory was that the memory bandwidth was limiting the performance of the HD6000, because the Iris Pro with eDRAM performed so much better.

 

[ammulder]

OK, I'm trying on a NUC5i3. I was able to install El Capitan without any SSDTs, but the GPU was pinned at 300 MHz. So I got Toleda's HD5500 SSDT, and that seems to have made a difference to the GPU -- at least it fluctuates some now. I'll need to try a game or something to see if it goes up to 900 MHz.

I'm not sure what to do about the CPU SSDT (i3-5010U). I haven't seen a CPU SSDT posted for that model. I ran the beta ssdtPRGen.sh, and it emitted 5 SSDTs -- looks like 1 for the CPU and 4 for the motherboard? If I use that generated SSDT.aml with toleda's GPU one, it seems to make things worse -- the CPU doesn't go below 1.3 GHz, even though ssdtPRGen and the SSDT it seems to have emitted say it should set it down as far as 500 MHz. The power usage also seems slightly higher in Power Gadget.

Any advice would be appreciated.

From F4: View attachment f4-dump.zip
From ssdtPRGen.sh: View attachment ssdts-generated.zip

Update: if I try toleda's SSDT-1.aml without any SSDT.aml, then it doesn't seem to work -- the GPU sticks at 300 MHz. But either that way or with no SSDTs at all the CPU ranges from 1 GHz to 2.1 GHz. If I use toleda's SSDT-1.aml with the generated SSDT.aml then the GPU frequency fluctuates between 300 MHZ and 900 MHz, but the CPU frequency only uses a smaller range from 1.3 GHz to 2.1 GHz. I guess I'm hoping to find or craft a CPU SSDT that has a bigger range -- it seems like 500 MHz to 2.1 GHz is the maximum possible.

 

[RehabMan]

OK, I'm trying on a NUC5i3. I was able to install El Capitan without any SSDTs, but the GPU was pinned at 300 MHz. So I got Toleda's HD5500 SSDT, and that seems to have made a difference to the GPU -- at least it fluctuates some now. I'll need to try a game or something to see if it goes up to 900 MHz.

Using toleda's SSDT accomplishes a rename of GFX0->IGPU, which activates IGPU PM.

I ran the beta ssdtPRGen.sh, and it emitted 5 SSDTs -- looks like 1 for the CPU

Only SSDT.aml is generated. The others are OEM SSDTs extracted -- side effects of the script, not output.

 

[ammulder]

OK. Should I expect to be able to alter the generated SSDT.aml to expose a larger range of CPU frequencies? What I see in Power Gadget is 1.3-2.1 GHz even though the script data and the emitted SSDT.aml seem to say that the full range should be 0.5-2.1 GHz.

 

[RehabMan]

OK. Should I expect to be able to alter the generated SSDT.aml to expose a larger range of CPU frequencies? What I see in Power Gadget is 1.3-2.1 GHz even though the script data and the emitted SSDT.aml seem to say that the full range should be 0.5-2.1 GHz.

Do you have FakeSMC plugins installed? If so, remove them.

Are your results different using SSDT.aml vs. not using it?

 

[ammulder]

Yes, different results. This is what I see in Power Gadget:

• No SSDTs: CPU 1.0-2.1 GHz, GPU 300 MHz (fixed)
• Only toleda's SSDT-1: same as above (maybe doesn't work by itself?)
• ssdtPRGen.sh SSDT with toleda's SSDT-1: CPU 1.3-2.1 GHz, GPU 300-900 MHz (or maybe 350-900, it's hard to tell because it fluctuates a lot)

So right now if I have the GPU working at its full frequency range, the CPU has a reduced range.

In some sense it doesn't matter a lot because the power usage is still super-low: Power Gadget shows e.g. 0.71W with IA 0.06W at idle. But I'm trying to get it "right".