[HowTo] Intel GMA 900 (for 915GM and 910)

[zorxd]

my 915GM (Asus Z71A) Allocate 16MB on boot, which is enough for me

[mastrchee]

my 915GM (Asus Z71A) Allocate 16MB on boot, which is enough for me

 

 

lucky you

 

so QE and CI works fine for u i'm assuming? guess 16MB is fairly enough to boot up. ASUS has always gone more than the official specs (i have experience with such effects with asus, with my desktop vga, lol which i dont really mind )

 

if this works fine for u, means it proves my theory...atleast a small step further to solve this problem.

 

grippa, very interesting theory and sounds very plausible.

 

I agree that the upcoming Intel notebooks will probably not be GMA900, however, it's possible that Apple have development notebooks in-house that use GMA900. If true, we may see working drivers in newer builds of OSx86. However of course we cannot rely on this.

 

I'm going to investigate the possibility of increasing the VRAM on system boot (or at least before OSX loads).

 

thanx! that would be really nice

 

and look here ->

my 915GM (Asus Z71A) Allocate 16MB on boot, which is enough for me

 

he has 16MB! yay, and it seems it works for him just fine (from what i understood)

[Tak]

The Intel spec states that the maximum pre-allocated memory is 8MB. It's possible that the Asus video BIOS (usually embedded in the system BIOS) is enabling the extra memory.

 

I've found a few things that may help.

 

Linux users have various hacks and workarounds for the 915 chipset to enable higher resolutions and something called "xvideo" and "DRI" (direct rendering), similar (in a way) to QE and CI.

 

Some links:

 

http://forums.gentoo.org/viewtopic-t-36193...t-dri+i915.html

http://forums.gentoo.org/viewtopic-t-364512.html

 

A tool called 915resolution:

http://www.geocities.com/stomljen/

 

This tool hacks the video BIOS on the fly to enable higher resolutions. When I get time I'm going to see how this works. Perhaps if I can enable a very high resolution to force the GMA900 to allocate more VRAM, then load up OSX afterwards.

[mastrchee]

grippa, very interesting theory and sounds very plausible.

 

I agree that the upcoming Intel notebooks will probably not be GMA900, however, it's possible that Apple have development notebooks in-house that use GMA900. If true, we may see working drivers in newer builds of OSx86. However of course we cannot rely on this.

 

I'm going to investigate the possibility of increasing the VRAM on system boot (or at least before OSX loads).

 

 

ohh...and since, ASUS has increased memory to 16MB on boot (on certain ASUS laptops), probably Intel-PowerBooks (if exists & yeah most probably, being sneaky is apple's 'thing' ) have more VRAM on boot. I hope this doesnt end being up BIOS problem.

 

but then, any bios-mods?

 

The Intel spec states that the maximum pre-allocated memory is 8MB. It's possible that the Asus video BIOS (usually embedded in the system BIOS) is enabling the extra memory.

 

I've found a few things that may help.

 

Linux users have various hacks and workarounds for the 915 chipset to enable higher resolutions and something called "xvideo" and "DRI" (direct rendering), similar (in a way) to QE and CI.

 

Some links:

 

http://forums.gentoo.org/viewtopic-t-36193...t-dri+i915.html

http://forums.gentoo.org/viewtopic-t-364512.html

 

A tool called 915resolution:

http://www.geocities.com/stomljen/

 

This tool hacks the video BIOS on the fly to enable higher resolutions. When I get time I'm going to see how this works. Perhaps if I can enable a very high resolution to force the GMA900 to allocate more VRAM, then load up OSX afterwards.

 

 

thank you very much for this update...i'm also lookin into what information i can find on this, but i dunno much about the coding side of macs, i'm also willing to learn about macs using x86...since there's a big opotunity rite now.

 

btw, i was just about sleep...lol guess i wont be doing that now

[Tak]

but then, any bios-mods?

 

See above We may not need to mod the BIOS directly with that tool

 

my 915GM (Asus Z71A) Allocate 16MB on boot, which is enough for me

 

zorxd, do you have 16MB showing on this screen?

 

 

And does QE and CI work? Also can you change the video memory in the BIOS?

[mastrchee]

i hope the other pro coding guys are lookin upon us right now...pls help.

[suleiman]

guys you might want to search the net for 915 resolution hack. This is a hack for laptop GMA 900 users with the EXACT SAME PROBLEM as we have, but for linux instead of darwin. From what I recall it worked on FreeBSD, which as you all know is what Darwin is based on.

 

Check it out!

 

edit: sorry about that Tak, and everyone else, I was only on page 10 of this thread and hadn't noticed you the double post. Ignore this message!

[vanja]

I've heard about 915 resolution and actually played with it but without success. Are you saying, suleiman, that you are using your 915 without any problems with x86?

[brighttie]

well guys...i've been trying everything on the net to get my 915GM (0x2592) to work...and so far here's my results.

........

 

Just FYI,

1. I got solid blue screen every time as long as I modified AppleIntelIntegratedGraphics.kext from 2582 to 2592 and put it in /System/Library/Extensions, doesn't matter whether other AppleIntel* kexts are there or not.

2. External monitor works every time ONLY when modified AppleIntelIntegratedGrahics.kext are in /System/Library/Extensions, seems that AppleIntelIntegratedGraphics.kext is required for dual monitor support and control.

3. My spcs is Dell Inspiron 2200, 915 GM/GMS 910GML express chipset family.

 

Hope we can work it out some day.

 

Sorry for the long quote before.

[mastrchee]

i got the s-video thing to work...i only used one resister @ 100ohms between 2-4 on the diagram shown earlier on this forum. luckily it works fine for me now. I hope the memory glitch can be fixed.

 

for the moment, if your really desperate you could just try the S-Video hack...its quite easy, it didn cost me a cent, i have plenty of resistors hangin around (may not be the case for u, but even so its cheap to buy a couple), so i fixed it to the svideo port, then covered it with some black tape. now osx boots with QE and CI. no problems attached

[Tak]

Just FYI, <snip>

Can you please edit your post and remove/snip the unnecessary huge quote - basic etiquette.

 

cheers

[suleiman]

no sorry vanja, I didn't mean to say that i had the 915 resolution hack working on x86. So far everything that I've tried has led to a no go.

 

The problem is that i can't even try to use the s-video hack, as the drivers pop up a "YOU MUST REBOOT" message for me after every reboot after the 10.4.3 update, and with or writhout an external monitor plugged in.

 

To be truthful i'm waiting for 10.4.3. full to be released, if it is, and i'm guessing graphics problems will be fixed/close to being fixed with that release. The files are changed around and the kext's that load for our cards have been changed to include everything from 2582 to 2592.

 

Cheers

[zorxd]

zorxd, do you have 16MB showing on this screen?

 

 

And does QE and CI work? Also can you change the video memory in the BIOS?

 

 

 

Yes (well 15826 or something else near 16MB)

 

but I still need the dummy s-video adapter

 

And I use 855resolution in Linux (the latest version of 855resolution includes the support of i915GM from 915resolution) and it works well. But with Xorg 6.9 or 7.0 it will be useless because the hack will be incorporated into the driver.

[Tak]

Yes (well 15826 or something else near 16MB)

 

but I still need the dummy s-video adapter

Hmm. So, assuming the memory theory is correct, we need more than 16Mb at boot for QE and CI without bluescreen.

[suleiman]

what i've always found odd about the memory theory, is that with an external monitor plugged in i still got a read out of only 8 mb. Strange.

 

On a side note, I just realized that we've been working to get a fix since August the 22nd. Which means for 3 months we haven't made any progress (if you cancel out the adapter hack).

 

Sorry for the glum news, but we gotta spread the word a little more about our problem. 3 monthts no go is pretty crummy.

[Tak]

what i've always found odd about the memory theory, is that with an external monitor plugged in i still got a read out of only 8 mb. Strange.

 

Probably because the extra memory is only allocated when the kexts are loaded, i.e after the Darwin boot screen. That's when the external monitor is detected by the kext and the kext requests the extra memory. The Darwin boot screen will only ever show the default initial allocation.

 

Edit - do you mean the read-out in OSX itself? If so, it may just be reporting the default allocation.

[mcdull]

the thoery not correct. at least from what I see.

okay, for those who you S-video hack, like me, you can plug in the s-video device "right after" the screen showing how many memory is allocated. and from what I understand you only need 4mb or less to get true color under 1024x768 resolution. And memory is only related to the resolution issue but not CI or QE.

[Tak]

the thoery not correct. at least from what I see.

okay, for those who you S-video hack, like me, you can plug in the s-video device "right after" the screen showing how many memory is allocated.

The screen showing how much memory is allocated is only showing the default initial amount set by the video BIOS on boot. This never changes, even if you plug in an external monitor or s-video "dongle".

 

As I say in my previous post, the additional memory is allocated after the Darwin boot selection screen, once the kext detects the second display device. It's just a function of the kext to allocate more memory, and as a result of this, it provides enough to allow QE and CI to work.

 

and from what I understand you only need 4mb or less to get true color under 1024x768 resolution.

Correct, but the extra memory is being allocated because of the kext detecting the additional display device. When the kext is loaded, it allocates enough memory for the resolution/colours and QE/CI for both displays. However, possibly due to a bug, it doesn't allocate enough memory when there's only one display.

 

And memory is only related to the resolution issue but not CI or QE.

I disagree, QE and CI use the acceleration features of the video card, similar to how games use it. More memory is required for these features. If video memory was only needed for resolution/colours then we'd all have 32MB ATi X850 or Nvidia 7800 cards

 

I'm not saying this theory is correct, after all it's just a theory. However, I've yet to see anything proving it wrong.

[mortis]

Yes (well 15826 or something else near 16MB)

 

but I still need the dummy s-video adapter

 

And I use 855resolution in Linux (the latest version of 855resolution includes the support of i915GM from 915resolution) and it works well. But with Xorg 6.9 or 7.0 it will be useless because the hack will be incorporated into the driver.

 

 

ZORXD:

 

Correct me if i am wrong:

 

Are you saying that alhtogh your initial screen shows 16MB allocated to Video, if you boot without the dummy s-video adapter, you get a blue-screen??

 

Or are you capable of booting directly into OsX without any adaptors or external monitors??????

[mastrchee]

well, what i think about the memory theory

 

* when i mean DVMT on the following text, i mean the mobile version of DVMT, not the desktop version.

 

1. mac os x 10.4.1 kexts dont supports DVMT in their kexts on boot up. but DVMT works fine after booting to osx if u finally manage to boot in to it. but this process happens within the hardware without notifying the OS in runtime.

 

2. DVMT wont assign more than 8MB (except for some notebooks like asus) on boot up for a couple of text strings like the darwin boot up.

 

3. GMA900 Desktop Version does not include the Mobile DVMT feature, rather it switches between 64MB and 128MB according the total ram. So mac os x 10.4.1 does not have native support for DVMT.

 

4. and i dont think mac os x supports showing dynamic allocation of video/system memory in the operating system. since everything is initialized during boot.

 

5. notice in mac os x dashboard, when u open a widget, u get the water effect, but the 1st time when open a widget the animation is a bit sluggish most of the time, but after the 2nd time everything's smooth. from what i think, the sluggish performance comes due to allocation of memory, and 2nd time its faster & smooth because memory allocation has completed.

 

6. if the existing kexts have support for DVMT, i think everything should work fine. intel probably have all of this lined up already...

 

i'm not saying all this is correct, but the memory theory is the closest thing we've come to ( from what i see...) everything is based on simple facts u can observe with your eyes...nuthin very technical, so yeah, who knows whats really goin on...BTW MAXXUSS....u seem to be a coding pro (thanks ), what u think u can do about this? and what do u think about the whole thing?

 

u might find this interresting, i qoute from http://www.apple.com/macosx/features/quartzextreme/

 

"Quartz Extreme Requirements

 

Quartz Extreme functionality is supported by the following video GPUs: NVIDIA GeForce2 MX and later, or any AGP-based ATI RADEON GPU. A minimum of 16MB VRAM is required."

[Urbz]

OK, first, i think that this theory has a lot of merit.

I have a dv4000, and i can't set anything in the bios about DVMT, so i haven't rly gotten very far. Still, i'm testing out a few things with the video memory.

Does anyone have or has found a little utility for mac os x that can tell you how much video memory has been used, and etc?

 

Because for the mostpart, i agree with this theory of needing more than 16/8 (we can't be sure) mb of video memory at boot. One thing i would like to add is that my laptop's fan is always on, or goes on no matter what after a few minutes. I've analysed the processor performance and mac really DOES ajust my processor speed, so it seemed stupid. My graphs of processor activity, however, always peak (usually to 20 %) when just doing simple things like moving the mouse (albeit very fast and nonstop). Still, graphic intensive stuff has a very large impact on the processor. As in 100%. I know quark xpress is PPC, but some of it should still (no?) be given to the GPU. Same goes for playing video in an x86 safari.

So MY theory is more like an add-on to the previous one:

 

Most bioses locked to 7-8 MB video ram on boot. Mac needs 16 (traditionally, os x86 may be slightly different...) for a successful boot. Attaching an external tv or monitor tells mac to give itself the extra memory just before going into the actual OS. Once inside Mac, you may safely detect displays and get rid of the external monitor.

Which currently, is what we're trying to fix: making it assign that ram without the monitor.

But I think there is more to the story on the Mac side. As afformentioned by Grippa, DVMT works within the OS, and this may be the reason for the ripple effect in dashboard not working properly the first time. My addition: in general, DVMT is not working as it does in windows, or QE is not working as it does with other video cards because the processor is still doing most of the GPU work. Maybe someone with a real mac could check for me, but i dont think that moving the mouse quickly (even without an external monitor) should need 20% of the rocessor, even when it's running at the automatic setting.

 

I would like for others to test things like processor activity with me so we can hopefully get this theory as correct as possible, and then someone with some programming skill could help us out.

 

-Urby3

 

On a side note, i believe that the "final"drivers should serve us better than those we have now. And for the record, i do think that the future intel macs WILL include drivers for the gma900 and 915GM chipset.

[zorxd]

ZORXD:

 

Correct me if i am wrong:

 

Are you saying that alhtogh your initial screen shows 16MB allocated to Video, if you boot without the dummy s-video adapter, you get a blue-screen??

 

Or are you capable of booting directly into OsX without any adaptors or external monitors??????

 

yes my Darwin boot screen shows 15MB (I think it always show 1MB less than your total video memory) and I get the blue screen if I don't use my dummy s-video adapter

 

I don't beleive we get the blue screen because we don't have engough video memory. If it was only that, there should be a way to disable QE and CI get rid of the blue screen.

I think it's because our LCD is plugged on pipe B instead of pipe A and the driver isn't aware of that.

[mortis]

yes my Darwin boot screen shows 15MB (I think it always show 1MB less than your total video memory) and I get the blue screen if I don't use my dummy s-video adapter

 

I don't beleive we get the blue screen because we don't have engough video memory. If it was only that, there should be a way to disable QE and CI get rid of the blue screen.

I think it's because our LCD is plugged on pipe B instead of pipe A and the driver isn't aware of that.

 

Then the theory of allocating more video RAM wont work, (at least with only 16 MB)

 

I do agree with you that driver tends to recognize the screen in a second position (pipe , so it actually wont boot (only to blue screen ) if the primary display (pipe A for the driver) is not connected.

[suleiman]

To everyone concerned about testing out this theory: Why not go to the linux people and ask them why they needed the 915 resolution hack? My guess is that it's also a DVMT issue.

 

The question is, what do we do with this information?

 

The Dell Inspiron 6000 is what most people, and certainly the first poster, experienced success with. Why not get in touch with some of those people and ask them what their memory readouts are at boot? Futhermore, what's so special about the Dell Inspiron 6000 that the kext loads on the exact same GMA900 as many of us are using, but not on ours?

 

A while back there was someone who made note of the fact that Intel Celeron systems, as present in the Inspiron, seem to have more success then Intel Pentium M systems, as present in the D610 I have. Can any of you lend validity to this plausibility?

 

I'm pscyhed. Let's get this figured out.

[mastrchee]

OK, first, i think that this theory has a lot of merit.

I have a dv4000, ...

 

my processor also use to max out when running graphics/video/3d apps and opengl stuff like chess didnt even work. but after enabling QE and CI (i boot with the svideo hack), processor doesnt max out anymore. Animations seems to be original and smooth. also in my lap, mac os x is much faster than Windows

 

BTW, in my bios, there's NOTHING related to VGA settings , i mean nothing, no DVMT, no monitor, agp aperture settings...nothing..lol