Jump to content

Help: Fixing Lenovo Y510p Nvidia GT 755M on OS X Yosemite 10.10.2


169 posts in this topic

Recommended Posts

Sorry for being away for so long. Quite busy i was.

 

Hello!

 

Til now I could not find it by myself. I have got only the one of MBP 11,2. But this is a bit useless, because it has only IGPU. I must not get it from you. Perhaps anybody can help.

 

Have fun.

 

Can you upload the ioreg of MBP 11,2? I would be glad to take a look for IGPU.

 

EDIT : Nevermind. Found it.

 

....The attached ACPI tables file is gone. Any info on how to generate my own?....

 

I have made a script to do all the DSDT/SSDT patching automatically, but i've yet not implemented extracting your own tables (not enough time). If you want i can upload it for you. But you'll have to extract acpi tables from linux, thats the only way the script is going to work.

 

It is configured to patch tables only from linux.

 

I'll upload it to github as soon as possible. (but if you want i can send you the script)

 

FYI: Its very easy to extract acpi tables from linux just 1 line of command! (and if you dont have it installed you can use your usb to boot into linux without installing. Try it.

 

OT:

 

@Ahmed: I flashed mod BIOS and played with DVMT pre-alloc (64, 96 MB) to get Intel working but still nothing yet. (also custom EDID doesnt seem to work, i'll try again when i'll have time).


.....Does it need to be on SMBIOS MBP 11,1? I have it on iMac 14,2 or whatever the recommended is at this point in time.....

 

Nothing is recommended. You can use MBP 11,1/11,2 and iMac 14,2. Also feel free to try others. I personally use MBP 11,2

Can you upload your config.plist minus the SMBIOS as well? I tried modifying mine as per your initial post, but I still can't get it to boot. As soon as it goes to switch to graphics mode, it reboots. I'm willing to try messing with the NVIDIA drivers and try messing with FakePCIIID to get the second card to detect incorrectly. I think this might be the way to get the first card to initialize properly. My reasoning is that OS X initializes the UltraBay card correctly and skips the internal because of this.

Sure, they are attached. The initial post have some things to try but not necessarily a must. Actually, what you will see in the config.plist(s) is that I did not inject Nvidia and did not use any FakeID. I also dropped injecting Intel and setting the ig-platform-id because I believe they are irrelevant. For boot-args I use nv_disable=1 which does not affect the built-in Nvidia card by anything I can notice other than allowing 7MB of VRAM instead of 0MB so the system is barely useable to get any data out of it. I also find using nv_disable=1 resulted in ultrabay card VRAM disappear which is something to consider. Later on, we should install the appropriate drivers and use nvda_drv=1 instead but that's of course assuming the card is working as supposed to.

 

Can you upload the ioreg of MBP 11,2? I would be glad to take a look for IGPU.

 

@Ahmed: I flashed mod BIOS and played with DVMT pre-alloc (64, 96 MB) to get Intel working but still nothing yet. (also custom EDID doesnt seem to work, i'll try again when i'll have time).

That's a good line to go with, I hope you find something useful. Let me know if you need any help there.

 

Nothing is recommended. You can use MBP 11,1/11,2 and iMac 14,2. Also feel free to try others. I personally use MBP 11,2

I would recommend a mac model that originally have Nvidia card so things would be smoother I guess. This includes some iMac 14,2 models, some MBP 11,1/11,2/11,3 and maybe others as well. As long it is for a model that have Haswell processor and Nvidia card it should be a good start. I would like to add that most MBP models that have Nvidia use GT 750m and some iMac 14,2 use GT 755m. If the card supported by the selected SMBIOS is different than what's inside our Y510p then I think we may need to use a FakeID. Mine have GT 755m so I will stick with iMac 14,2 at the moment and let's see where this would go.

SMBIOS.zip

...That's a good line to go with, I hope you find something useful. Let me know if you need any help there.....

 

Something weird happened when i changed DVMT pre-alloc memory. It seems like the gradients are increasing with allocated memory. With default 32 MB you know how the gradients look, with 64 MB they are bigger (more space between two bands) and with 96 MB they are even more bigger. They keep increasing with the increase in memory allocation. What's weird is the debug logs are the same. No difference.

 

I wonder what would happen if i set it to 0 MB.

 

......I would recommend a mac model that originally have Nvidia card so things would be smoother I guess. This includes some iMac 14,2 models, some MBP 11,1/11,2/11,3 and maybe others as well. As long it is for a model that have Haswell processor and Nvidia card it should be a good start. I would like to add that most MBP models that have Nvidia use GT 750m and some iMac 14,2 use GT 755m. If the card supported by the selected SMBIOS is different than what's inside our Y510p then I think we may need to use a FakeID. Mine have GT 755m so I will stick with iMac 14,2 at the moment and let's see where this would go.....

 

 

True. I forgot about the Nvidia card coz i was not using it. For me, i focus only on Haswell models coz Intel HD4600 is not supported in any mac notebooks only in iMac 14 models.

Something weird happened when i changed DVMT pre-alloc memory. It seems like the gradients are increasing with allocated memory. With default 32 MB you know how the gradients look, with 64 MB they are bigger (more space between two bands) and with 96 MB they are even more bigger. They keep increasing with the increase in memory allocation. What's weird is the debug logs are the same. No difference.

 

I wonder what would happen if i set it to 0 MB.

It is weird but I would advise otherwise as ased on my readings online you should set it to Max.

Hello!

 

Is the NVDA working with QE/CI? Vram size is perhaps cosmetic. (Can be set in Clover to correct value.)

 

Try to set SMBios to MBP 11,1 and inject this NVCAP with Clover: 05 01 00 00 00 00 01 00 06 00 00 00 00 00 00 0e 00 00 00 00

 

By the way, if you want to use only the intel gfx, SMBios should be set to MBP 11,2. It is in this case the best and nearest your hardware  matching one.

 

Have fun.

Hello!

 

Is the NVDA working with QE/CI? Vram size is perhaps cosmetic. (Can be set in Clover to correct value.)

 

Try to set SMBios to MBP 11,1 and inject this NVCAP with Clover: 05 01 00 00 00 00 01 00 06 00 00 00 00 00 00 0e 00 00 00 00

 

By the way, if you want to use only the intel gfx, SMBios should be set to MBP 11,2. It is in this case the best and nearest your hardware  matching one.

 

Have fun.

Nvidia is not working with QE/CI though full resolution is there but I think the Vram issue here is not cosmetic since it is visible how the UI is laggy. I will try your suggestion and regarding using Intel gfx only this is not my aim at the moment because we have a serious problem with Intel graphics on this machine which you can take a look on here.

Hello!

 

I know this, but some other ones are on the Intel and don´t use this SMBios. Could be a good part of not working Intel.

 

The NVCAP is for test purpose. If you can boot with it, look in IORegistryExplorer and look, if it shows up there completely in the same way as injected.

 

Have fun.

Hi, thanks for this suggestion. I tried this NVCAP and was able to boot with it though it didn't make and visual difference. I don't know however how to check with IORegistryExplorer because I could not find something related to NVCAP.

 

I also was able to set VRAM in clover which worked fine but it was only cosmetic as the graphics still without QE/CI (though the performance is a bit better).

Screen_Shot_2015_03_16_at_19_05_27.png

Screen_Shot_2015_03_16_at_19_06_44.png

Screen_Shot_2015_03_16_at_19_06_55.png

 

PS: do you know of any source to learn more about IOREG rather than reading the documentation by Apple?

Hello!

 

In IORegistyExplorer tree the NVCAP under Device GFX0 should show up the same way as injected.

 

Does your display show up in System Profiler as an internal one (built in)?

 

I don´t know any source about IOREG.

 

Have fun.

In IORegistyExplorer tree the NVCAP under Device GFX0 should show up the same way as injected.

Yea I found it under PEG0 instead, the same NVCAP:

Screen_Shot_2015_03_16_at_18_25_37.png

I uploaded the IOREG if you wanted to take a look: https://mega.co.nz/#!IF1VyIzZ!HriFGXH9SoxmsHMB1Ls9RWelahDHtTT-raCrJSNHjiE

 

 

Does your display show up in System Profiler as an internal one (built in)?

No it does not (a good point you made). It sees the display as external I believe:

Screen_Shot_2015_03_16_at_18_17_16.png

 Thanks for your help

 

Hello!

 

Thanks,this one is known to me. The most interesting one is the one of MBP 11,1.

 

Here one can find many other hardware dumps: http://tdev.me/2010/12/apple-hardware-dumps/

 

Have fun.

If this will really advance the progress on Y510P w/dual NVidia 755m, I'll be willing to buy one for under $1000. What are the odds that getting this dump would solve this issue?

If this will really advance the progress on Y510P w/dual NVidia 755m, I'll be willing to buy one for under $1000. What are the odds that getting this dump would solve this issue?

 

I would cool down a bit because it seems we have a problem with our hardware configuration regarding dual-Nvidia chips. If you saw the earlier posts you would find that OS X detected the built-in Nvidia as connected to an artificial external display but not to the real existing built-in display. If I'm not wrong, I would think that the built-in Nvidia card is not directly connected to the display but it sends the data through another thing that is directly connected to display (hint: Intel  :wallbash: ). This means both Nvidia cards are only capable of doing graphic processing but cannot work alone and the Intel card will still be the main for display even if it doesn't process any graphic by its own. If this is the case, then both the nvidia cards works somehow using Optimus Technology as Allan said before and there is no way to get this any further. I'm not sure by any degree about what I just wrote but I'm trying to analyse the results so far an I deeply hope I'm wrong here.

.....I would think that the built-in Nvidia card is not directly connected to the display but it sends the data through another thing that is directly connected to display (hint: Intel  :wallbash: ).....

 

That is actually the case in every Optimus system. The graphics card is routed through intel somehow. All the framebuffer data is passed through intel graphics. So there's no way to disable intel. But I don't have any idea what happens in SLI system.

If this will really advance the progress on Y510P w/dual NVidia 755m, I'll be willing to buy one for under $1000. What are the odds that getting this dump would solve this issue?

 

I don't think that's a good idea. If there would have been a SLI system from Apple then maybe yes but now no, its useless.

That is actually the case in every Optimus system. The graphics card is routed through intel somehow. All the framebuffer data is passed through intel graphics. So there's no way to disable intel. But I don't have any idea what happens in SLI system.

 

Yep. The weird thing in SLI (or let's say the dual-GPU configuration) we have there is no trace of Intel at all in contrast with the usual Optimus configuration where Intel is present loud and clear. This is the case in all three OS's Windows, Linux, and OSX no trace of Intel while both nvidia GPUs are in charge. Can Intel be still behind the scene just receiving processed graphics (be either one or two nvidia GPUs) and route it to display? I don't know if the manual would have such deep anatomy of the hardware to answer my question. Also this configuration for laptop is somehow rare so the gurus don't bother checking and investigating assuming it's just one of those Optimus anyway which may not be the case.

 

By the way, you have modded bios so you can do us a favour here. Can you disable Intel from BIOS and boot to Windows using only nvidia? This may answer something.

....Can Intel be still behind the scene just receiving processed graphics (be either one or two nvidia GPUs) and route it to display? I don't know if the manual would have such deep anatomy of the hardware to answer my question.....

 

That is something hardware related. Searching the internet might lead somewhere.

 

....Also this configuration for laptop is somehow rare so the gurus don't bother checking and investigating assuming it's just one of those Optimus anyway which may not be the case.....

 

Try looking at forums with tags "alienware" and "sli". I have seen some Alienware SLI questions asked before. I'll have to search again.

 

 

....By the way, you have modded bios so you can do us a favour here. Can you disable Intel from BIOS and boot to Windows using only nvidia? This may answer something....

 

Ok i'll try that and report back.

 

EDIT: any bootflags that i can try with clover?

That is something hardware related. Searching the internet might lead somewhere.

Try looking at forums with tags "alienware" and "sli". I have seen some Alienware SLI questions asked before. I'll have to search again.

I'm searching but still can't find something close to my question. Meanwhile, I thought to take a look in Nvidia Control Panel with and without SLI mode. Without SLI (i.e. Intel + built-in nvidia (Optimus mode)), the graph shown that Intel is definitely connected to the display as expected:

Physyx2.png

 

But in SLI mode (i.e. no Intel + built-in nvidia + ultrabay nvidia) the graph suggests only one of the cards is connected to laptop display which I found it to be the built-in card:

Physyx.png

 

I also checked device manager in both cases. Without SLI, the Intel card is connected to the display:

Bus_relations2.png

 

But in SLI with both nvidia cards, the built in card can be seen connected to the same display Intel used to use before with ID: LGD0416:

all.png

 

No trace at all for Intel while in SLI mode but only one clue.

@ Ahmed Salah AIS

 

Hello!

 

Try the attached Efistring injected with clover device properties. Use a Clover trial boot stick and no other gfx injection.

 

Have fun.

 

Hi styrian, thanks for this. I tried it and there's a little improvements as the card is identified correctly though with only 1024MB of VRAM. However, that number is somehow cosmetic as there is no QE/CI and the system still feels like having 7MB VRAM. I also want to note that I am not able to boot and log in without nv_disable=1 or else system will restart before showing any graphics. I attached a copy of ioreg.

 

Screen_Shot_2015_03_24_at_21_48_13.png

Screen_Shot_2015_03_24_at_21_49_04.png

Also the system still not able to connect to built-in display.

ioreg.zip

Hello!

 

Try this.

 

Have fun.

 

 

Hi, thanks.

 

I tried it and it improves the value of VRAM shown to 2048MB but still only a value without effect as the graphics performance has not improved. I also tried with the web drivers but they gave me troubles booting. In all cases I'm not able to boot without nv_disable=1.

 

1_about.png

2_display.png

3_profiler.png

ioreg.ioreg.zip

×
×
  • Create New...