Jump to content

fermi/gtx 480/gtx 470


cmf
 Share

1,729 posts in this topic

Recommended Posts

To the OP - it would be useful to include, in your first post, the overall status of this issue / project. Like, "big improvements recently but still not solved, not for general consumption yet".

 

thank you for your work on this.

Link to comment
Share on other sites

To the OP - it would be useful to include, in your first post, the overall status of this issue / project. Like, "big improvements recently but still not solved, not for general consumption yet".

 

thank you for your work on this.

what recent big improvements? nothing has changed since november.

and my gtx 480 is working quite nicely. probably all gf100 based cards are, except for opencl and video acceleration, obviously.

 

 

edit:

is anyone able to get temperature readout from their Fermi? I'm not at my machine atm, however I'd like to get an idea of how cool my 470 is. My understanding is that nvclock will not work with fermi. At least my card causes complete freeze. Are there alternatives? Also. should I be setting my fan-speed or power-mode at any time during boot or for on wake? DSDT related?

I also have display issues on wake from either sleep, computer and display. Scramble-vision.

Anyway. I'd like to find solutions to these items. Hope there is help. thanks

try fakesmc+nvclockport and istat menus.

Link to comment
Share on other sites

try fakesmc+nvclockport and istat menus.

 

I figure you may be talking about previous version of fakesmc ~ 2.7.2.

I started fresh build with fakesmc3 in /E/E, works for everything but graphics. Althought newer nvclockX had some error that caused me to remove it. Is support for GF100-based cards not included in this newer version?

Am I correct that I should look into the previous version 2.7.2 and its corresponding plugin? I can see that the devID for my 470 was added to that project back in November.

I have checked out the source, but am otherwise clueless on how to compile. I have installed Xcode3.25 on 10.6.6, 64 bit.

Is there a painless way to do this? Searching all google gives me no ideas, and most site searches are posts of people asking for someone else to do it. I thought I were to browse to the working directory, issue 'make' cmd and be done. This however doesn't work w/o makefile. Aye. Any pointers appreciated.

Or link to most "current" of fakesmc272

Thanks

Link to comment
Share on other sites

w00t, it's finally there :)

 

10.7 developer preview 2/11A419 has a NVDAGF100Hal.kext :)

 

			<key>IOPCIPrimaryMatch</key>
		<string>
			0x06c010de&0xffe0ffff
			0x0dc010de&0xffc0ffff
			0x0e2010de&0xffe0ffff
			0x0ee010de&0xffe0ffff
			0x0f0010de&0xffc0ffff
			0x104010de&0xffc0ffff
			0x124010de&0xffc0ffff
		</string>

 

and it's even the latest driver, r270 branch (270.02.00a15) :)

 

 

edit: there seems to be sm_20 support in PTX.dylib, so there is the possibility that opencl also works. but then there is no precompiled header for ptx20, so maybe not yet.

Link to comment
Share on other sites

			<key>IOPCIPrimaryMatch</key>
		<string>
			0x06c010de&0xffe0ffff
			0x0dc010de&0xffc0ffff
			0x0e2010de&0xffe0ffff
			0x0ee010de&0xffe0ffff
			0x0f0010de&0xffc0ffff
			0x104010de&0xffc0ffff
			0x124010de&0xffc0ffff
		</string>

 

and it's even the latest driver, r270 branch (270.02.00a15) :)

 

Seems like progress.

270 >> 256? new driver? guessing nvidia wouldn't release until Lion launch, but interesting note that it exists.

 

Also wondering about matching device ID of a specific card in plist. My card, 0x06cd10de, is not specifically listed. Does the specific device ID need to be included or does the "&0xffe0ffff" part cover a series or family of cards?

Also, what does the address of "&0xffe0ffff" correlate to?

Link to comment
Share on other sites

Seems like progress.

270 >> 256? new driver? guessing nvidia wouldn't release until Lion launch, but interesting note that it exists.

 

Also wondering about matching device ID of a specific card in plist. My card, 0x06cd10de, is not specifically listed. Does the specific device ID need to be included or does the "&0xffe0ffff" part cover a series or family of cards?

Also, what does the address of "&0xffe0ffff" correlate to?

i wouldn't be surprised if those supported geforce 500 series cards :)

 

specifically listing all device ids would be stupid, thats were the and-part comes in.

<your device id> & <mask> == <listed device id>, if this is a match, the kext will be loaded.

in your case: 0x06cd10de & 0xffe0ffff = 0x06c010de which is equal to the listed 0x06c010de.

Link to comment
Share on other sites

i wouldn't be surprised if those supported geforce 500 series cards :)

 

specifically listing all device ids would be stupid, thats were the and-part comes in.

<your device id> & <mask> == <listed device id>, if this is a match, the kext will be loaded.

in your case: 0x06cd10de & 0xffe0ffff = 0x06c010de which is equal to the listed 0x06c010de.

 

Thanks for the prompt response. I had a feeling that those & meant to include a series of cards.

 

Any ideas on finding a compiled build of the latest pre-v3 build of fakesmc and plugins? I'm very interested in seeing temps of my card respond to loads.

 

Thanks

Link to comment
Share on other sites

what recent big improvements? nothing has changed since november.

 

and my gtx 480 is working quite nicely. probably all gf100 based cards are, except for opencl and video acceleration, obviously.

 

You missed the point of my comment. I was suggesting a quick digest of the current status of the thread. "Big improvements" was part of an example of the type of comment that would help new readers to monster, though somewhat curated, threads like this to understand what's going on.

 

A case in point from my perspective, your detail about opencl and video acceleration is hardly "obvious" to someone new to the thread and topic, hence my suggestion.

Link to comment
Share on other sites

I was suggesting a quick digest of the current status of the thread.

 

The following applies mainly to the GTX 460, see post #1549 by rominator below for more information on other fermi architecture video cards.

 

When booting install media or OS X installations that don't have the Quadro 4000 drivers installed, make sure to boot with GraphicsEnabler=n. Any injection method (gfx0 device in DSDT, device-properties string, injector.kext) must be disabled, uninstalled, obliterated. Otherwise you will not be able to boot.

 

Make sure you're running at least 10.6.5

Install Chameleon 2.0 RC5

Set GraphicsEnabler=y in /Extra/com.apple.Boot.plist

Install the latest nvidia Quadro 4000 drivers.

If you still don't get any hardware acceleration, you may need to add PCIRoot=1 to your /Extra/com.apple.Boot.plist

 

To get working HDMI audio, see 'how to patch AppleHDA' thread on the ProjectOS X forums.

This requires binary patching of two kernel extensions and adding some code to your DSDT.

 

OpenCL does not work, some users experience random freezes and "OpenGL Channel Exception" kernel panics.

OpenGL graphics performance (going by Cinebench and OpenGL Extensions viewer scores) is on par with a 9800 GTX+ or GTS 250.

 

Display rotation doesn't work, results in a rotated but black screen with mouse pointer.

 

At present, one way to postpone the "Fermi Freeze" issue is to leave VLC media player running in the background, looping a video clip. It doesn't have to be visible, it can stay hidden in the dock.

 

Netkas' partner-in-crime Rominator is working on a patched BIOS ROM to hopefully cure the freeze issue:

http://www.insanelymac.com/forum/index.php?showtopic=253523

 

I believe that sums it up.

Link to comment
Share on other sites

Thanks :P

 

 

My final suggestion, and please understand it is not a demand, is that this type of information be put back in the first thread. Moreover, the above example is great, but I'm even suggesting something more high-level. What is being solved in the thread and why? What have we accomplished so far? What challenges remain to be solved?

 

The perspective I bring is that of a user new to this issue, and in the end it's just a suggestion that in my opinion will help new users to figure out what's going on at a glance so they can decide if they want to dive in and learn more.

Link to comment
Share on other sites

I don't understand what you want, or why you want it.

 

The first post in this thread is old and full of workarounds and files that aren't necessary anymore.

As you can see, when the fermi cards first came out, they didn't work on Hackintoshes at all, or you had to jump through hoops to get them to work, only GF100 chip cards could be made to work but there were issues, and so on and so on.

 

Anyway, I'm not going to write up a calendar of events!

 

If it's a history lesson you want, just read the thread and you'll see every single event leading to the present state of affairs unfold before your eyes.

 

To get Fermi 4xx cards to work the procedure is the same as any other nvidia card on a Hackintosh - except that you must boot with graphics injection disabled until you install the Quadro 4000 drivers from nvidia.

 

I believe that all necessary information at this time is available in my post above. And if there is more, I'm sure someone will add it here.

 

If you'd like the first post to be changed you should get in contact with the original poster.

Link to comment
Share on other sites

I apologize. I did not mean to agitate anyone. But what you say there regarding the first post being old is a great reason for what I suggested. In my opinion the OP should update the original post in the thread with the current status of the thread, including high-level commentary and some current details.

 

This is an example of what I consider to be a thread with a helpful, current first post:

http://www.insanelymac.com/forum/index.php...95248&st=60

 

I appreciate that you jumped in here, but I was not asking you or anyone to create a calendar of events. I simply suggesting occasional updating of the original post. I don't suggest that for every thread out there, but for very large threads with a lot of people looking for answers, it is very helpful to people to be able to see at a glance from the first post what is going on.

 

For example, the freeze issue is something that would prevent many people from buying this card. Having that detail clearly spelled out as an unsolved issue in the first post of the thread would be very useful. The user can then decide 1) am I willing to buy this card and have potential issues? 2) am I willing to help come up with a solution if I buy this card anyway?

 

Again, these are all suggestions, and you and the OP are clearly welcome to ignore. If suggestions geared towards clarity and ease of consuming a very large thread like this are unwelcome, I am sorry.

Link to comment
Share on other sites

1. Stop quoting entire posts when you are replying directly below them! It's all garbage and destroys readability and I have edited the post you are quoting, your quote is obsolete, etc etc - I beg you please edit your posts and delete your unnecessary quotes.

 

/EDIT *thank you*, readability is up by 200%!

 

2. Send a private message to the original poster and voice your concerns. I cannot help you because I can't edit the first post, I only have mod privileges in the "Genius Bar". And even if I could I would not do it, out of respect for the author.

 

3. I am not agitated, I am stating my beliefs in a calm, polite and composed manner. :angel: Your suggestions and ideas are not unwelcome, I never said that and I don't know how you can read that into my posts. There is no need to apologize. The only issue I have with your posts is the unnecessary quoting.

 

The first post in this thread was documentation of the journey made by the original poster in order to get his video card to work, it was never meant to be a friendly guide. But I agree it's a complete mess and it would be nice if it was updated with current information.

 

If you can't get hold of the original poster and you think there should be a newbie friendly step-by-step guide feel free to write it yourself, you now have all the necessary information to do so. Don't wait for someone else to do it!

Link to comment
Share on other sites

1. Stop quoting entire posts when you are replying directly below them! It's all garbage and destroys readability and I have edited the post you are quoting, your quote is obsolete, etc etc - I beg you please edit your posts and delete your unnecessary quotes.

OK. Though it was my intention to retain the entirety of the quote as I sensed some anxiety and defensiveness from the OP, and then you, and a quote is a good record of what was actually responded to before edits are made. Re-reading the posts, I think for the most part I imagined it, and I'm sorry for that.

 

The first post in this thread was documentation of the journey made by the original poster in order to get his video card to work, it was never meant to be a friendly guide.

 

This is how these things usually start, but now this thread is one of THE references on this issue, which is why I made the suggestion. But you're right, perhaps it should have been done in a private message.

 

Anyway... after all this (and "this" was really trying to be constructive criticism), and especially your helpful summary for me, which I really do appreciate, I'll probably not upgrade to one of these cards, especially in light of of the new ATI cards in the new MBPs.

 

 

EDIT:

 

See, this is partially why I quote. Before this edit, it looks like I ignored your suggestion to create my own guide, but that was something you added in an undocumented edit.

 

Anyway, at this point in time I am exploring options for video card upgrades. I am not at this time interested in writing a guide on an issue I have very little understanding on, and will not likely have the need to learn as I will likely not be purchasing the card. But I hear you. As a general rule of thumb, instead of complaining about something (not that I think I am complaining, I am offering constructive criticism), people should try to add value themselves.

Link to comment
Share on other sites

I certainly did not mean to trip you by editing my post. I don't go editing my posts in order to twist someones words or place anybody in a bad light. I think you're the one who's being defensive here :angel: I wasn't asking you to delete quotes because I want you to hide something "bad" that I said!

Thankfully you can edit your post too and respond to the added content, which you did, and (ideally!) everybody's happy.

 

What I do, is I often edit my posts to add more information and/or make changes to state my opinion in a better way. English is not my first language, so it happens quite often that I want to improve my own wording.

 

Yes, there are really only two viable solutions - write your own guide or contact the OP and ask him to do it.

 

As a general rule of thumb, instead of complaining about something (not that I think I am complaining, I am offering constructive criticism), people should try to add value themselves.

 

Bravo, exactly!

Link to comment
Share on other sites

I certainly did not mean to trip you by editing my post. I don't go editing my posts in order to twist someones words or place anybody in a bad light. I think you're the one who's being defensive here :( I wasn't asking you to delete quotes because I want you to hide something "bad" that I said!

 

I didn't mean to imply that you were doing anything like that, (Haha, what a terrible moderator that would be :) ) ; on the contrary, you were very helpful in your interaction with me.

 

Just in general, I like how quoting is a record of what was originally said and responded to, and it could be useful in a few situations. But yeah, it sure can add a lot of overhead to a page!

 

In terms of forums features, maybe the right approach is a revision history, where you always display the latest version, but the changes to a comment are recorded, which would prevent the need to overquote like I had been doing.

Link to comment
Share on other sites

Current Fermi Status (as I see it)

 

1. To use Fermi cards, you need to install the Nvidia drivers for Quadro. You will need to right-click and look into packages, run all 3.

 

2. You will need some form of injection, I can't speak to that so I won't. In a real Mac Pro, you use ATY_Init with Fermi in slot 1 and additional Nvidia EFI card in slot 3. Slot 3 needs at least 4 lanes for this to work.

 

3. Currently the only Fermi cards that are supported are GTx4xx cards. The most common are GTX460, GTX470, and GTX480. GTX5xx series cards can be made to render desktop in 2D but that is it. No 3D of any sort.

 

4. While the cards run better than when the drivers were first "leaked" last summer, they still are not living up to their potential. While this can be said for MOST Mac OSX cards, the Fermi cards have a hard time distancing themselves from the GTX260/275/285 cards of previous generation. What this means is that they aren't especially useful except for allowing one to feel like they have "newer, better" tech. The other big benefit is if you DUal Boot into Windows, you already have a much better card installed.

 

5. The GTS450 and GTX460 are currently plagued with a freeze problem that takes the fun out of using them. There are useable workarounds but the feeling that there is a tent flap blowing around in the wind is unsettling.

 

6. The GTX470 and GTX480 are immune from freezing. For most intents and purposes, they are solid stable cards. The GTX480 requires LOTS of power. I was reading a GTX590 test and saw that my 480 uses nearly as much power as a 590. Most of this power is being used to generate heat, not 3D. Consequently the metal plate on top gets HOT. (I am surprised that there isn't a time lapse somewhere of somebody cooking an egg on top..pretty sure it would do)

 

7. The current "exceptions" to 100% function is that FCP Preview windows come up empty and Embedded QT videos won't play.

 

8. People have found ways to enable Audio over HDMI, but it seems you need to use older Apple kexts and add the device id. I don't have an HDMI display with audio in my office so I won't be good source of info on this.

 

Everyone is EAGERLY awaiting next Quadro update in hopes that:

 

1. Current GTx4xx cards will be able to stretch legs and make all of this bother worth the wait.

 

2. The GTX5xx cards will have 3D, the GTX570 will become new the new Mac OSX darling.

 

3. With any luck, the freeze will be fixed if we don't kill it first. Since Q4000 doesn't have issue this would be result of benevolent programmer or general Code Tidying.

Link to comment
Share on other sites

okay, i'll update the first post.

 

The first post in this thread is old and full of workarounds and files that aren't necessary anymore.

what do you mean by that? everything i said there is still required, namely: bootloader, graphicsenabler, kexts. instead of the latter, one could also install the kexts directly using the nvidia installer, but the installer only works on macpro 3,1-5,1. so, just copying kexts to /S/L/E is probably easier for everyone.

 

CUDA works but not all CUDA cores are used

i'm repeating myself, but: they are working! if $program doesn't show all cores, it's the programs fault, not the fault of the cuda driver. even then, it might only be a display bug of that program.

 

(I am surprised that there isn't a time lapse somewhere of somebody cooking an egg on top..pretty sure it would do)

there is a video on youtube ... :D

 

btw. nice progress with fixing the freeze issue, hope you'll find a general fix!

 

 

and: did someone test 10.7 dp2 and fermi drivers yet? i'll install later/tomorrow, but i have to do some major partition/hdd/file shifting first ...

Link to comment
Share on other sites

Quick question. I have been reading through the past couple posts, and I kept seeing the words "Quadro drivers". Is that what is in the Nvidia update on tonymacx86's site? Or is that something different. I am currently using a 460 and I am able to boot in fine. I, like many others, have the random kernel panics. The other question I had is how do I know if I am getting "full acceleration"? If I can play Team fortress 2 does that mean I have full acceleration? I am still slightly new to this whole thing so I am trying to get a grip on what I need to know.

Link to comment
Share on other sites

 Share

×
×
  • Create New...