KNOWN ISSUE: [Windows 7 with] 2gb or more RAM

**Eugene Gavrilov** · Aug 19, 2009, 11:55 PM

btw, http://www.microsoft.com/whdc/archive/highperf-drv.mspx

Quote:

or an example of a driver that implements its own DMA-channel object, see the SB16 sample audio adapter in the Windows DDK. If the constant OVERRIDE_DMA_CHANNEL is defined to be TRUE, the conditional compilation statements in the source code enable the implementation of a proprietary IDmaChannel object, which the driver uses in place of the default IDmaChannel object from the IPortWaveCyclic::NewXxxDmaChannel call.

I will have a look at this code later. Please review it, too.

Quote:

It is becoming increasingly common for a PC to contain more than 4 gigabytes of RAM, and on those systems it is possible to have mappings that are located higher than 4 GB in physical memory (in both the 64-bit CPU case and the x86 PAE case). To achieve the best performance on these machines, IHVs should create devices that support 64-bit addressing, otherwise data copying in software is required. This has historically been the case with devices that require 24-bit addressing on systems with more than 16 MB of RAM. Devices should use WaveCyclic instead of WavePci if they cannot read from or write to anywhere in physical memory. It is acceptable for a driver to make this decision at StartDevice time.

(Ibid.)

E.

Russ · Aug 20, 2009, 01:59 AM

The SB16 sample's IDMAChannel implementation appears to just call the base class implementation of each function and does not really implement any of the functionality itself. In any case it should prove useful to show how to get an initial IDMAChannel object to work with, and then you should only need to implement some of the functions (like (I assume) AllocateBuffer) and call the base class implementation for the rest.

The MSVAD sample also appears to have a partial implementation of a custom IDMAChannel.

thrillhouse82 · Aug 20, 2009, 10:48 AM

Quote:

Originally Posted by Russ

The SB16 sample's IDMAChannel implementation appears to just call the base class implementation of each function and does not really implement any of the functionality itself. In any case it should prove useful to show how to get an initial IDMAChannel object to work with, and then you should only need to implement some of the functions (like (I assume) AllocateBuffer) and call the base class implementation for the rest.

Yes, this is what it does. It is merely an example of how to provide a custom IDmaChannel implementation. BTW: AllocateBuffer doesn't need to be implemented since NewStream doesn't require this. You will see that in my sample code I'm currently preparing; it is done but I want it to at least load once and run a few simple tests before I send it to you guys. Is C++ fine or do you need C wrapper stuff?

thrillhouse82 · Aug 21, 2009, 11:11 AM

I uploaded the first version of the sample implementation. You can review it at http://home.arcor.de/infolchinger/CDmaChannel.cpp and http://home.arcor.de/infolchinger/CDmaChannel.h

I did some quick tests on a WinXP 1GB machine and it appears to do what it should. Tonight I will test it on a 4GB Win 7 machine. This will be the moment of truth since on this system, the current driver version will always fail with random noise. If the physical address of CDmaChannel is below 2G there, chances are that we soon have a driver that will work on all systems regardless of available memory.

Some notes about the source: It is written in C++ and brings its own new/delete operator. If your driver already defines operator new and operator delete, you can just delete the ones of the CDmaChannel class. The class itself can be instantiated by calling CDmaChannel::Create or (if you use C) by calling CreateCDmaChannel. Parameters are size in bytes and highest acceptable physical address. Therefore, it should be possible to replace NewMasterDmaChannel by a call to CDmaChannel::Create and return this object instead of the one created by NewMasterDmaChannel.

Results of the Win 7 Tests: Returned physical address was good on several tries.

RPGWiZaRD · Aug 22, 2009, 12:43 AM

It's actually working?

If so you've just extended kX driver's life at least another 10+ years (I hope). ^^

thrillhouse82 · Aug 22, 2009, 08:15 AM

Quote:

Originally Posted by RPGWiZaRD

It's actually working?

If so you've just extended kX driver's life at least another 10+ years (I hope). ^^

I don't know if it will work in the driver's context, but the sample object behaves as expected and returns usable addresses, so chances are quite good

But usually it is a longer way from a working prototype to a production level driver than most people would expect.

Therefore, please be patient and give Eugene some time to integrate and test this approach before celebrating

Lex Nahumury · Aug 22, 2009, 07:16 PM

Hi thrillhouse82,

If you haven't done so already, it's best to contact Eugene directly through email
kxproject@yandex.ru in regards to development details such as these.
Thank you.

RPGWiZaRD · Aug 29, 2009, 03:29 PM

Any updates? I'm so curious to know if this bug gets fixed or not because I'm planning on upgrading comp soon so need to know if I'll have to buy a 2GB ram kit instead of 4GB.

MegaByte · Sep 15, 2009, 05:13 AM

I have upgraded my computer with more 2GB RAM and therefore fallen prey of this bug. Windows Vista x64 does work, but Windows 7 x64 doesn't.

powerflux · Mar 6, 2010, 07:03 PM

hi, i run an audigy 1394 and an audigy 2 platinum in the same pc. worked perfectly under XP, but now with windows 7-64 (and 4GB of ram) this bug popped up. playback seems to be OK with all buffers set at max. but it still sounds sometimes as if something's wrong.

also, windows thinks i've got a 8-speaker surround setup, even if i put it on '2.0 stereo' in the kxmanager. can't change it in the windows control panel (greyed out) this gives some irritating issues with gaming.. but that's a problem for later. not very important.

but, i was just checking for news on the 2GB+ bug, is there any? the changelog for 3350 says it's solved, but it clearly isn't. i love the kx drivers, so please continue the good work!

Tyrsonswood · Mar 6, 2010, 10:11 PM

Which output device is selected in windows audio panel as default? Try using SPDIF/AC3 with buffers at standard settings. You shouldn't be having any 4GB issues as it is fixed with 3550 drivers.

powerflux · Mar 8, 2010, 09:28 AM

Quote:

Originally Posted by Tyrsonswood

Which output device is selected in windows audio panel as default? Try using SPDIF/AC3 with buffers at standard settings. You shouldn't be having any 4GB issues as it is fixed with 3550 drivers.

hi, thank you for your answer.

i have the following symptoms
- stuttering/jittering/popping/clicking during playback. i haven't tried recording yet
- asio isn't available in adobe audition and renoise. simply no devices. i miss that too.
- mulitple programs trying to use the same card at the same time makes the noise far more worse
- most tracks start ok, but nearing the end the problem becomes more and more dominant
- measurement bars like the VU meters in audition respond very slow. renoise is unusable without asio
- can't use waveoutHQ on the audigy 2. greyed out.

using S/PDIF is not really an option since i want to use the analog in- and outputs on the cards. AC3 files have the problem too.

i've tried the following things
- reinstalling the drivers multiple times, swapping cards between slots, using only one card, switching between the two, everytime reinstalling the driver.
- setting the buffers to max. solves the problem for 90% of the time, but it doesn't 'feel' right the way the music sounds.
- booting windows XP-32 with 2GB's installed, memory remap on (2GB available). no problemo
- booting windows XP-32with 2GB's installed, memory remap off (2GB available). no problemo
- booting windows XP-32 with 4GB's installed, memory remap on (3.12GB available). no problemo
- booting windows XP-32with 4GB's installed, memory remap off (2GB available). no problemo
- booting windows W7-64 with 4GB's installed, memory remap on (3.12 available). trouble
- booting windows W7-64with 4GB's installed, memory remap off (4 available). trouble
- trying wave-out, directsound and kernel streaming. neither solved the problem
- running the PC at stock speed and underclocked (normally it is overclocked) with all the options mentioned above
- using the onboard realtek and some old philips sound card. no problem in W7-64 and XP-32. but that's no alternative

and because of upgrades/testing
- tried 4 different video cards (nvidia 8600gt, ati 3850, ati 3870, ati 4850x2) since i've switched to W7, no changes
- new powersupply. didn't solve the problem

i will try having 2GB's installed with W7-64, but my memory isn't easily accessed.

my chipset is the 975XG on the asus p5wdh deluxe motherboard.

Tyrsonswood · Mar 8, 2010, 12:29 PM

Well, first of all, I wasn't saying to use a SPDIF device, just select that as your default output in Windows audio panel, keep using your same analog input and output hardware. If you tried it and it didn't get rid of the noises and stuttering then you have something else going on..... it's worked for a lot of guys.

Next, realize that no 32 bit applications are going to work with a 64bit OS and KX drivers.

powerflux · Mar 10, 2010, 05:58 PM

Quote:

Originally Posted by Tyrsonswood

Well, first of all, I wasn't saying to use a SPDIF device, just select that as your default output in Windows audio panel, keep using your same analog input and output hardware. If you tried it and it didn't get rid of the noises and stuttering then you have something else going on..... it's worked for a lot of guys.

Next, realize that no 32 bit applications are going to work with a 64bit OS and KX drivers.

i misunderstood both.. S/PDIF as default audio device _almost_ solves the problem. few clicks in an hour, but with the buffers at default! nice, thanks.

with 'no 32bit apps going to work' do you mean not working at all? because audition works. just badly and with no asio support..

Tyrsonswood · Mar 10, 2010, 06:26 PM

It's the ASIO support mostly, from what I've read.... I don't have 64bit myself.

Glad the other change helped some, It's what I was set at when I had the Win7 RTM installed and it worked great.

JessicaD · Mar 15, 2010, 09:43 PM

Powerflux,
Prior to migrating to Windows 7 from Windows XP did you run the Windows 7 Upgrade Advisor?
To determine if your current computer configuration will be compatible with Windows 7 please go to the following link: Windows 7 Upgrade Advisor
Another resource to verify compatibility is available at the Microsoft Windows 7 Compatibility Center located at the following link: Windows 7 Compatibility Center home
Jessica
Microsoft Windows Client Team

Aug 20, 2009, 01:59 AM	#32
Russ HardwareHeaven Extreme Member Join Date: Jan 2005 Posts: 5,561 Rep Power: 62	Re: KNOWN ISSUE: [Windows 7 with] 2gb or more RAM The SB16 sample's IDMAChannel implementation appears to just call the base class implementation of each function and does not really implement any of the functionality itself. In any case it should prove useful to show how to get an initial IDMAChannel object to work with, and then you should only need to implement some of the functions (like (I assume) AllocateBuffer) and call the base class implementation for the rest. The MSVAD sample also appears to have a partial implementation of a custom IDMAChannel.

Aug 21, 2009, 11:11 AM	#34
thrillhouse82 DriverHeaven Newbie Join Date: Aug 2009 Posts: 8 Rep Power: 0	Re: KNOWN ISSUE: [Windows 7 with] 2gb or more RAM I uploaded the first version of the sample implementation. You can review it at http://home.arcor.de/infolchinger/CDmaChannel.cpp and http://home.arcor.de/infolchinger/CDmaChannel.h I did some quick tests on a WinXP 1GB machine and it appears to do what it should. Tonight I will test it on a 4GB Win 7 machine. This will be the moment of truth since on this system, the current driver version will always fail with random noise. If the physical address of CDmaChannel is below 2G there, chances are that we soon have a driver that will work on all systems regardless of available memory. Some notes about the source: It is written in C++ and brings its own new/delete operator. If your driver already defines operator new and operator delete, you can just delete the ones of the CDmaChannel class. The class itself can be instantiated by calling CDmaChannel::Create or (if you use C) by calling CreateCDmaChannel. Parameters are size in bytes and highest acceptable physical address. Therefore, it should be possible to replace NewMasterDmaChannel by a call to CDmaChannel::Create and return this object instead of the one created by NewMasterDmaChannel. Results of the Win 7 Tests: Returned physical address was good on several tries. Last edited by thrillhouse82; Aug 21, 2009 at 10:06 PM. Reason: Notes about the source/Win 7 results

Aug 22, 2009, 12:43 AM	#35
RPGWiZaRD HardwareHeaven Lover Join Date: Dec 2007 Posts: 132 Rep Power: 0	Re: KNOWN ISSUE: [Windows 7 with] 2gb or more RAM It's actually working? If so you've just extended kX driver's life at least another 10+ years (I hope). ^^ Last edited by RPGWiZaRD; Aug 22, 2009 at 12:49 AM.

Aug 22, 2009, 07:16 PM	#37
Lex Nahumury DH Senior Member Join Date: Jan 2003 Location: The Netherlands Posts: 1,932 Rep Power: 64	Re: KNOWN ISSUE: [Windows 7 with] 2gb or more RAM Hi thrillhouse82, If you haven't done so already, it's best to contact Eugene directly through email kxproject@yandex.ru in regards to development details such as these. Thank you.

Aug 29, 2009, 03:29 PM	#38
RPGWiZaRD HardwareHeaven Lover Join Date: Dec 2007 Posts: 132 Rep Power: 0	Re: KNOWN ISSUE: [Windows 7 with] 2gb or more RAM Any updates? I'm so curious to know if this bug gets fixed or not because I'm planning on upgrading comp soon so need to know if I'll have to buy a 2GB ram kit instead of 4GB.

Sep 15, 2009, 05:13 AM	#39
MegaByte HardwareHeaven Junior Member Join Date: May 2007 Posts: 25 Rep Power: 0	Re: KNOWN ISSUE: [Windows 7 with] 2gb or more RAM I have upgraded my computer with more 2GB RAM and therefore fallen prey of this bug. Windows Vista x64 does work, but Windows 7 x64 doesn't.

Mar 6, 2010, 07:03 PM	#40
powerflux HardwareHeaven Newbie Join Date: Mar 2010 Posts: 3 Rep Power: 0	Re: KNOWN ISSUE: [Windows 7 with] 2gb or more RAM hi, i run an audigy 1394 and an audigy 2 platinum in the same pc. worked perfectly under XP, but now with windows 7-64 (and 4GB of ram) this bug popped up. playback seems to be OK with all buffers set at max. but it still sounds sometimes as if something's wrong. also, windows thinks i've got a 8-speaker surround setup, even if i put it on '2.0 stereo' in the kxmanager. can't change it in the windows control panel (greyed out) this gives some irritating issues with gaming.. but that's a problem for later. not very important. but, i was just checking for news on the 2GB+ bug, is there any? the changelog for 3350 says it's solved, but it clearly isn't. i love the kx drivers, so please continue the good work!

Mar 15, 2010, 09:43 PM	#46
JessicaD HardwareHeaven Newbie Join Date: Jun 2009 Posts: 3 Rep Power: 0	Re: KNOWN ISSUE: [Windows 7 with] 2gb or more RAM Powerflux, Prior to migrating to Windows 7 from Windows XP did you run the Windows 7 Upgrade Advisor? To determine if your current computer configuration will be compatible with Windows 7 please go to the following link: Windows 7 Upgrade Advisor Another resource to verify compatibility is available at the Microsoft Windows 7 Compatibility Center located at the following link: Windows 7 Compatibility Center home Jessica Microsoft Windows Client Team