Contribute
Register

GPU performance issues on intel HD5300 (hp envy x2 13) SOLVED !

Status
Not open for further replies.
NEW files generated from mac osx 10.12.6

Your ioreg shows:
- IGPU QE/CE working with IGPU PM
- CPU PM working
- your kextcache output shows kexts not installed correctly (kext injection is only for installer/recovery)

You're using ig-platform-id 0x16260006, but 0x161e0000, 0x161e0001, 0x161e0101 are other possibilities (and seem to match your device-id).

Note that your display is connected to the 0204 port (a bit strange that it is not LVDS).

Your files in ACPI/origin are too old to compare against ACPI/patched, so I didn't. You forgot to press F4 as per "Problem Reporting".
 
at the first I want to thank you for helping.
What you are doing for hack community is priceless.

You're using ig-platform-id 0x16260006, but 0x161e0000, 0x161e0001, 0x161e0101 are other possibilities (and seem to match your device-id).

I tried

ig-platform:

0x16260006 -works / 1536mb vmem / low gpu performance

0x16160006-works / 1536mb vmem / low gpu performance
0x16160000-works / 1024mb vmem / low gpu performance

0x161e0000-works / 1024mb vmem / low gpu performance

0x161e0001- black screen
Screen Shot 2017-08-24 at 2.07.37 PM.jpg

0x161e0101- boot / 7mb vmem / no qe/ce

smbios:
macbook8,1 -works
macbookair7,2 - works
macbookpro12,1 -works

I patched via hexeditor AppleIntelBDWGraphicsFramebuffer binary for DVMT

Find : 8945C839C7764F
Replace : 8945C839C7EB4F
success

I decided to do patch connector AppleIntelBDWGraphicsFramebuffer for 0x161e0001 to output via HDMI ,

I found the table:

ig-platform-id = 0x161e0001
0100 1e16 0102 0202 0000 2002 0000 5001 // 34MB BIOS-allocated memory, 21MB frame buffer memory
0000 0060 6c05 0000 6c05 0000 0000 0000 // 1536MB VRAM
0000 0000 0000 0000|0000 0800 0200 0000 = LVDS connector
3002 0000 0105 0900 0004 0000 0710 0000 = DisplayPort connector
ff00 0000 0100 0000 4000 0000 0000 0000 = Disabled / unused VGA connector
0000 0000 0000 0000 0204 0000 0000 0000
0200 0000 0300 0000 00ef 1c0d 0000 0000

but data: 3002 0000 0105 0900 0004 0000 0710 0000 for DisplayPort is just missing in AppleIntelBDWGraphicsFramebuffer binary file, this is my problem.

Note that your display is connected to the 0204 port (a bit strange that it is not LVDS).
it because of this:
File_000.jpeg


Your files in ACPI/origin are too old to compare against ACPI/patched, so I didn't. You forgot to press F4 as per "Problem Reporting".
it isn't old, it because no battery connected to motherboard and default year in bios/uefi was 2015
 
at the first I want to thank you for helping.
What you are doing for hack community is priceless.



I tried

ig-platform:

0x16260006 -works / 1536mb vmem / low gpu performance

What do you mean by "low gpu performance".
No "Problem Reporting" files attached.

it isn't old, it because no battery connected to motherboard and default year in bios/uefi was 2015

They are too old. You forgot to press F4. Please read "Problem Reporting" again. Carefully.
 
What do you mean by "low gpu performance".

there are two pics from one computer one of them from windows, another one from macos

from windows
Screen Shot 2017-08-24 at 5.32.24 PM.jpg

from macOS
Screen Shot 2017-08-24 at 5.32.05 PM.jpg

and of course I see that GPU is 5-6 time slower then it must to be

They are too old. You forgot to press F4. Please read "Problem Reporting" again. Carefully.

see. origin.zip
 

Attachments

  • origin.zip
    56.4 KB · Views: 57
I came back

I did a little research with cinebench + intel power gadgets

I captured data about CPU during cinebench gpu testing from Win and Mac and something strange happened there, please look, it can be curious

but for now I ordered usb ethernet and m2 ssd for more smooth experiments

Orange graph is MAC
BLue one is Windows
 

Attachments

  • Screen Shot 2017-09-14 at 11.32.36 AM.jpg
    Screen Shot 2017-09-14 at 11.32.36 AM.jpg
    451.8 KB · Views: 74
Last edited:
Success

Screen Shot 2016-09-18 at 8.53.53 PM.png


finally I found the way how to have intel hd5300 with expected performance.

Its very strange method but it works.

1) boot with efi from m-5y70-works-EFI.zip

Delete all cache data :
/System/Library/Caches/*.*
/private/var/db/BootCaches/*.*

reboot

2) NOW I can see not working QE/CE but working Audio
then rename /EFI/CLOVER/ACPI/patched__ to /EFI/CLOVER/ACPI/patched
reboot again
in this case most likely you will have freeze with kernel panic

3) reboot but in this case choose Clover config: config_hd5300.plist
and now everything works like it must to be

If you will find more clever and smooth way please write here.

But for now it's enough of experimentations


some helpful links:

https://github.com/Piker-Alpha/ssdtPRGen.sh/issues/279

https://github.com/gfoury/UX305FA-DSDT-Patch/issues/7

https://www.bountysource.com/issues/38188904-invalid-character-and-syntax-error

https://github.com/Piker-Alpha/ssdtPRGen.sh/issues/254
 

Attachments

  • m-5y70-works-EFI.zip
    2.1 MB · Views: 72
Status
Not open for further replies.
Back
Top