Audio driver for Realtek HD Audio Hardware? [Testing thread]

[deomsh]

According to info in HDACFG.INI your chipset is Intel C200. Not seen in this thread before.

I suggest first set pcipatchB=$7900 in HDACFG.INI.

Please run two tests:

1) without HDAICOUT.HDA (rename).

2) with HDAICOUT.HDA

Always reboot in between.

BTW verbinterface=$0, so use HDAICOUT.HDA with all lines starting with $0

Listen with headphones to *any* sound/ crackle.

Please report results and upload HDAICIN.TXT only (rename accordingly, so first test HDAICIN1.TXT etc.).

Edited March 29, 2023 by deomsh
Addition

[Danielx]

1 hour ago, deomsh said:

According to info in HDACFG.INI your chipset is Intel C200. Not seen in this thread before.

I suggest first set pcipatchB=$7900 in HDACFG.INI.

Please run two tests:

1) without HDAICOUT.HDA (rename).

2) with HDAICOUT.HDA

Always reboot in between.

BTW verbinterface=$0, so use HDAICOUT.HDA with all lines starting with $0

Listen with headphones to *any* sound/ crackle.

Please report results and upload HDAICIN.TXT only (rename accordingly, so first test HDAICIN1.TXT etc.).

Oh my god, it worked! 

PcipatchB=$7900 did the trick!

At first it clicked very loud. Then the windows 98 boot sound came but a bit too extreme loud 😅 it was a bit crackling thou. Btw no hdaicout.hda used, only the ini with pcipatchB.

 

Still i wonder why it doesnt write an hdaicin.txt file without hdaicout.hda

 

With the hdaicout.hda without any other changes besides the change in the ini file, still no sound. That was the try without removing anything.

But removing everything besides beginning with $0 is leaving not much over. And still doesnt really work. I added a zip file with the hdaicout.hda in the specific condition plus the hdaicin.txt.

 

Hdaicout-data.zip

Edited March 29, 2023 by Danielx

[deomsh]

Congratulations!

If I understand well you said you tried different versions of HDAICOUT.HDA from the thread. I ment another version, lines STARTING with $0, not changing this file.

Nevermind, maybe HDAICOUT.HDA is not needed. So for now rename.

I found in the datasheet of Realtek codec ALC272 that the DAC and the Volume widgets should have the same number.

OutputWidget in your HDACFG.INI is already $2, so try VolumeWidget=$2 too. Then reboot.

Next open WAVEOUT.EXE with Start => Run => WAVEOUT and try with your mouse if volume sliders work while playing some sound.

Addition:

About HDAICIN.TXT: this file logs all HDA verbs sent by the driver to the codec (verbs are a sort of 'commands' to GET/ SET registers in the codec) . Return values are only specific in case of GET verbs (read-out of specific actual bit-values in registers of widgets/ nodes, and converted to verbs in HEX - note all registers are described in bits!).

In case of using HDAICOUT.HDA you can sent additional GET / SET verbs to the codec(s) - in practice about max 200 additional verbs in HDAICOUT.HDA.

'Structure of verbs and registers' are all described in Intel's HD-Audio Specification.

Edited March 30, 2023 by deomsh
Additions, typo, corrections

[Danielx]

22 hours ago, deomsh said:

Congratulations!

If I understand well you said you tried different versions of HDAICOUT.HDA from the thread. I ment another version, lines STARTING with $0, not changing this file.

Nevermind, maybe HDAICOUT.HDA is not needed. So for now rename.

I found in the datasheet of Realtek codec ALC272 that the DAC and the Volume widgets should have the same number.

OutputWidget in your HDACFG.INI is already $2, so try VolumeWidget=$2 too. Then reboot.

Next open WAVEOUT.EXE with Start => Run => WAVEOUT and try with your mouse if volume sliders work while playing some sound.

Addition:

About HDAICIN.TXT: this file logs all HDA 'commands' written by the driver. Return values are only specific in case of GET 'commands' (read-out some actual bit-values in widgets/ nodes, but converted to 'words' in HEX).

In case of using HDAICOUT.HDA you can sent additional 'GET / SET commands' to the codec(s) (in practice about max 200).

It is all described in Intel's HD-Audio Specification.

well thank you for your help, its amazing! it works prettty well, now even the volume slider of waveout is working!

i'm very glad to found this application, wouldnt know any other way to get sound working with intel hd codec in windows 98se. can only recommend it, this hdatsr app and ofc this board! =D

[deomsh]

You're welcome.

Be aware HDA2.DLL is a 16-bits (win3.x) driver, so there can be instabilities with multi-tasking. Especially if copying big files while listening to some music...

But it's all *somewhere* in this thread.

[Vigami]

Hello everyone,
I have been reading this thread more or less from the beginning to the end and I think the knowledge and the support, which @deomsh is providing here, is just incredible. A big thank you from my side! But anyhow, despite all this information I still have an issue, so I decided to sign up to this forum and ask ... 🙂

I was trying to get audio running in WIN98SE on a DELL Inspiron 6400 notebook. After the HPA installation I first got some bluescreens, but that was quicly fixed with the [vcache] limitation in system.ini (I first tried without that). Without hdaicout.cpa I had no sound, but this was also quickly fixed using the hdaicout.cpa from the driver package (HDADRV9J). However, the sound is far too loud ... in fact I never expected that the little speakers in this notebook could be that loud. And unfortunetely waveout.exe does not have any effect on this. 🙉

The Dell Inspiron 6400 uses an IDT (Sigmatel) STAC9200 codec. A corresponding datasheet can be found here (for example):

https://www.recomb-omsk.ru/published/SC/html/scripts/doc/IDT_STAC9200_DST_20120530.pdf

From the widget diagram I take that SleepingWidget and OutputWidget should be $02 and VolumeWidget shoud be $0B. I tried that (and many other combinations), but without any success. The volume level just cannot be changed, neither in waveout.exe, nor in Winamp, nor by setting a volume level in HDAcfg.ini. And now comes the strange thing: I had a look into Hdaicout.cda and (after some time to understand) I saw that it sets the master volume to full level at widget $0B (as I expected). And when I change the lines ...

$00B3901F
$00B3A01F

to (for example)

$00B39008
$00B3A008

... I get an acceptable volume level. However, this level is still fixed, the volume sliders don't have any effect. But from my understanding the VolumeWidget $0B seems to be correct. But even with this setting in HDAcfg.ini the Windows applications do not control volume correctly. 🙁

Do you have any idea what's going wrong here?

[deomsh]

Hello @Vigami , thanks for your nice words. Good to hear Watler's Win3x HDA-driver is giving sound on your system.

I took a look in the datasheet, but found nothing special. Widget 0B should be the Volume Widget and payload 1F the maximum volume.

If I remember well, I first used 7F in my quasi-universal HDAICOUT.HDA (because of my own Realtek codec). Later I changed this to 1F, because of specs of other codecs.

In my opinion your experiment confirms 0B is the Volume Widget for your codec. VolumeWidget in HDACFG.INI should indeed set accordingly. But you tried this already.

Leaves following questions:

1) Are you sure WAVEOUT.EXE is in your Windows-directory (just in case)?

2) If trying to use volume slider of WinAmp, is output of WinAmp set to DirectSound? Otherwise the volume slider won't work. Also try Windows MediaPlayer 6.4, or Foobar or Gom Player (Win9x compatible versions).

By the way, with Foobar you can change the output buffer which can raise performance of HDA2.DLL. Foobar can also convert 24/32 bits HD-Audio streams to 16 bit (HDA2.DLL is 16-bit only and max 48 kHz). Not sure about WinAmp, I haven't used it much.

If there is a choice in "Sounds and Audio Devices" / "Multimedia" (as part of Control Panel), set the 'Output' of Playback Device to DirectSound (I am not sure about this at the moment, maybe I am wrong - no active Windows 98SE installation available to check).

I don't know how WAVEOUT.EXE is working. In the source code there seems to be no GET/SET commands, but some Microsoft Wave-API. That's all my wisdom about this subject.

About the Volume Widget of your codec: it's mentioned as 'Master' Volume Widget in the datasheet. Maybe this is not a regular volume widget?

I believe Intel's High Definition Audio Specification mentions a 'Volume Knob' with 'master' function (or something like that).

There is also a Volume Widget for mono-out mentioned in the datasheet. Did you try VolumeWidget=$11 ?

 

Edited April 19, 2023 by deomsh
Addition

[Vigami]

Hello @deomsh, thank you for your quick reply and thank you for sharing your opinion, that I am not completely on the wrong way ... 🙂

So went through your questions and remarks:

Quote

1) Are you sure WAVEOUT.EXE is in your Windows-directory (just in case)?

Yes, it's in the Windows directory (I just checked again 😉). And it seems to work for my bluetooth audio (device 0). But for for HDA (device 1) it doesn't have any effect. When I change the sliders and than press "waveOutGetDevCaps" again, both chanels jump back to 32256. When I do the same thing for bluetooth audio, the sliders will stay in place.

Quote

2) If trying to use volume slider of WinAmp, is output of WinAmp set to DirectSound? Otherwise the volume slider won't work. Also try Windows MediaPlayer 6.4, or Foobar or Gom Player (Win9x compatible versions).

By the way, with Foobar you can change the output buffer which can raise performance of HDA2.DLL. Foobar can also convert 24/32 bits HD-Audio streams to 16 bit (HDA2.DLL is 16-bit only and max 48 kHz). Not sure about WinAmp, I haven't used it much.

I wasn't aware of this, but you can configure it in the WinAmp settings. And when I choose the DirectSound output plugin there, the WinAmp slider can actually control the volume! 😀 However, the bad thing is, that the maximum volume is whatever I set in the HDAICOUT.HDA. It doesn't change the volume of the system sounds and when I set the volume too high in HDAICOUT.HDA, all those system sounds will be too loud (and if possible, I would like too keep them activated).

With Windows Media Player 9 it's the same (I hadn't tried before), but I have not checked any othe players.

And Winamp can change the buffer size as well (currently it's set to 2000ms). But I have not (yet) checked if this setting has any effect, because currently the audio quality is not my highest priority. But I might try this when everything else ist set up properly ...

Quote

If there is a choice in "Sounds and Audio Devices" / "Multimedia" (as part of Control Panel), set the 'Output' of Playback Device to DirectSound (I am not sure about this at the moment, maybe I am wrong - no active Windows 98SE installation available to check).

Sounds like a good idea, but unfortunately I can only choose between "Bluetooth Hands-free Audio" and "HDA Sound" there, so no possibly to configure DirectSound there.  🙁

Quote

About the Volume Widget of your codec: it's mentioned as 'Master' Volume Widget in the datasheet. Maybe this is not a regular volume widget?

I believe Intel's High Definition Audio Specification mentions a 'Volume Knob' with 'master' function (or something like that).

There is also a Volume Widget for mono-out mentioned in the datasheet. Did you try VolumeWidget=$11 ?

Of course I cannot be sure, but according to the specification the widgets $0B and $11 are the only output widgets with volume control. And yes, I think I tried it before, but to be sure I just tried again: $11 does not work at all. Even adjusting the volume in HDAICOUT.HDA does not have any effect on the speaker output.

It really looks like I'm stuck with my fixed volume setting in HDAICOUT.HDA. This is a little disappointing, but on the other hand it's much much better than no sound at all! If you have any further ideas, please let me know. But if not, I say thank you anyway!

 

[Vigami]

@deomsh
Oh wait, one more thing (😉):
Would you mind to briefly explain the steps, which typically need to be done in HDAICOUT.HDA? It is difficult to understand without explanation because (from my understanding) you are trying to cover different hardware in only one file. So far I have only seen that there is an initialization at the beginning followed by powering up all the nodes. And of course you are setting the volume and unmute the output. Inbetween I saw something about a Stream-ID (whatever that is) and a comment regarding the stream format (but the command seems to be wrong for my case, because it's not described in the spec). But there are lot more verbs in this file, for which I don't know what they are supposed to do. I have not yet finally decided if I will try any further, but in case I do, a short explanation of the (typical) necessary steps would help at lot.

Edited April 20, 2023 by Vigami

[deomsh]

1 hour ago, Vigami said:

Quote

1) Are you sure WAVEOUT.EXE is in your Windows-directory (just in case)?

Yes, it's in the Windows directory (I just checked again 😉). And it seems to work for my bluetooth audio (device 0). But for for HDA (device 1) it doesn't have any effect. When I change the sliders and than press "waveOutGetDevCaps" again, both chanels jump back to 32256. When I do the same thing for bluetooth audio, the sliders will stay in place.

My question was just to be sure.

Your answer on my first question contains new information you didn't share earlier (no offense of course). I suppose your bluetooth device is not a HDA-device???

Please post two action print-screens of WAVEOUT.EXE, one for your 'device 0' and one for 'device 1'. I also want to take a look at your HDACFG.INI. Also HDAICIN.TXT: two versions, one of your current setup, one without HDAICOUT.HDA (just rename). Always reboot your computer in between.

Good to hear redirecting output of your player to DirectSound activates the players' volume slider.

About the remaining volume-problem of system sounds: if you need them, we have to try solving this problem. I will think about it.

I will also (try to) address your question about my quasi-universal HDAICOUT.HDA. This will take some time however.

In between you can read the commands of (Output) Widget 02 of your codec in the datasheet (pages 39-42: NID = 0x02). 

[Vigami]

Maybe I should clarify one thing first: I don't really need the system sounds! I have only set up this WIN98SE installation for fun, so the system sounds are only important for the 'nostalgic' feeling. The solution now is much better than I thought it would be, so any further investigations only are from curiosity. Don't spend too much effort on my case.

About the behaviour of WAVEOUT.EXE I was (again) not fully precise. When I change the slider for the bluetooth device, the value is kept (also when I press the button again or restart waveout) and it is also kept when I then switch to the HDA device and try to do some change there (but it does not have any effect on the speaker volume). But when I change the sliders when HDA is selected, waveout will always jump back to its last setting (after pressing the button again). I hope, this expresses it clearly.

I don't know if the bluetooth audio is a HDA device, but I do not think so. It was installed together with the bluetooth driver and is already a bit older. And as far as I know this device is not mentioned in the HDAcfg.ini file.

I have attached all the files you asked for, but I don't get two versions of HDAICIN.TXT, because this file will not be generated without HDAICOUT.HDA (the windows folder then contains the HDAICIN.TXT from the last try with HDAICOUT.HDA). I don't know if it is useful, but I have also attached the HDALOG.TXT files (these files differ with or without HDAICOUT.HDA).

Again I say thank you for your help! Please note that I will not available over this weekend, so my next answer might take a bit longer.

 

HDAcfg.ini

HDAICIN_with_hdaicout.TXT HDALOG_without_hdaicout.TXT HDALOG_with_hdaicout.TXT

[deomsh]

On 4/21/2023 at 11:19 AM, Vigami said:

I don't know if the bluetooth audio is a HDA device, but I do not think so. It was installed together with the bluetooth driver and is already a bit older. And as far as I know this device is not mentioned in the HDAcfg.ini file.

I studied your HDACFG.INI a bit, and it appears to me there are TWO function groups in your codec: an Audio Function Group (AFG) and a modem. First I thought you choose the wrong HDAICOUT.HDA file, but this is not true. The driver HDA2.DLL identifies the wrong function group as HD-audio and sets the 'CODEC Index' wrongly to $1 instead to $0.
Be aware I made a mistake lately - On March, 29: I mis-red 'Verbinterface=$1' for 'CODEC Index=$1'. Luckily Danielx's computer didn't explode, actually his 'CODEC Index' was $0 in his HDACFG.INI ).

Lets take a look at the first ten lines of your HDALOG.TXT, also to give you some help to understand HD-audio Verbs.

$000F0000=$83847690
$000F0004=$00000000
$000F0005=$00000000
$100F0000=$00000000
$100F0000=$00000000
$100F0004=$00000000
$100F0005=$00000000
$100F0000=$14F12BFA
$100F0002=$00090000
$100F0004=$00020001

The first Verb sent by the driver is $000F0000. This verb can be analyzed in four parts ($ is not part of this, internal to the driver!):

0  00  F00 00: first '0' stands for Codec Index 0, second two '00' for the root node. The 'F00' part is a code to obtain parameters of the codec function group, so to read-out. The last part is the payload, '00' at the and means parameter 'Vendor ID'. The full verb is also called : 'AC_PAR_VENDOR_ID'. Second part of the log is the codec response: $83847690

If you search on the internet for 8384:7690 you will find Sigmatel 9200 HD Audio Codec.

Verb $000F0004 asks Subordinate_Node_Count and $000F0005 for Function Group Type.

Maybe because there are no meaningful responses (???), the driver does not further check parameters of function group '0' starting with node '01', but starts instead with function group '1' (Codec Index 1) with node '00' of this function group again. This time some verbs are sent twice:

$100F0000=$14F12BFA ;; AC_PAR_VENDOR_ID => HDAudio Soft Data Fax Modem with SmartCP
$100F0002=$00090000 ;; AC_PAR_SUBSYSTEM_ID (not of interest to us)
$100F0004=$00020001 ;; Subordinate_Node_Count: according to the datasheet you found is the starting widget number '02' and there is only ONE node (which make no sense in case of an Audio Function Group response of verb $101F0004 should give meaningful information about number of widgets and starting number -NID- of the first widget). See further your HDACFG.INI.

So this explains why your Audio Function Group responses to the verbs in HDAICOUT.HDA because these are actually starting with Codec Index '0'.

But why WAVEOUT.EXE is 'seeing' two devices with 'Device 1' as HDA-device, instead as 'Device 0' is not fully clear to me. Maybe while with 'VolumeWidet=$0B' the 'good' node-number is addressed by the driver, but in the wrong function group? Earlier in this thread problems with a modem identified as a HDA-device by HDA2.DLL where mentioned, but not solved.

You can try to play a bit with the waiting timing in HDACFG.INI (now wait1=$100 and wait2=$100), so maybe doubling in steps up to $1000?). Still a miracle your Audio Function Group with 'Codec Index=$0' is fully driven by the verbs in HDAICOUT.HDA.

Some verbs you asked for (in the examples I am using the widget-numbers you already found for your codec):

$ 0 02 706 10 ; AC_VERB_SET_CHANNEL_STREAMID;01 => is related which digital audio stream the DAC will process.

$ 0 02 2 4011 ; AC_VERB_SET_STREAM_FORMAT;44.1kHz_16-bits => watch length of this code part of the verb: only '2', so one nibble (4 bits = a half byte). The payload is complex (and twe full bytes = 16 bits), for instance 0011 is an alternative to set 48kHz_16-bits. So also DAC related.

You understood the verb AC_VERB_SET_AMP_GAIN_MUTE already, so only left are:

$ 0 0B 701 00 ;AC_VERB_SET_CONNECT_SEL;Front? => means connect this widget to lowest connection number (to a widget earlier in the signal path) if this widget is a selector widget, or has selection properties.

$ 0 0B 707 40 ;AC_VERB_SET_PIN_WIDGET_CONTROL;out_enable => PIN-widgets are connected to jack's etcetera. But mostly multi-function: Input or Output. Often their default state is Input.....

Be aware maximum number of verbs sent to the codec is 255 or so. In practice about 200 lines left for HDAICOUT.HDA.

Further reading: see Intel's HD-audio Specification and your datasheet. Nice to read is AN20_Add_HDA_verb_table.pdf - found with Google, very short.

 

Edited April 26, 2023 by deomsh
Corrections

[Vigami]

Thanks a lot for this information! 👍

Unfortuantely I slowly get the feeling that there is not much one could do in this case. I played around with the wait settings, but that didn't have any positive effect (it only made the whole computer slower?!). Then I had a look into the driver source code. I'm not a pascal programmer and don't know much about drivers, but from my understanding the driver checks every codec and determines the number its nodes. It then steps though all theses nodes until it finds the (first) audio node ('AC_GRP_AUDIO_FUNCTION'). Unfortunately in my case the reported number of nodes for codec $0 is zero, so the driver does not look any further and continues at codec $1. I think it's a bit strange, that codeg $1 at first also reports zero nodes, but the driver then asks again and gets a different result. But I don't think it's worth the effort to dig even deeper and clarify this.

Do you know if the 'commands' in HDAICOUT.HDA are executed before or after this driver initialization?

[Vigami]

I was curious and I asked myself: If the modem is the issue, what would happen whithout it? So I opened the laptop and completely removed the modem card. The result was partially as expected, but partially surprising ...

As before again zero nodes are reported during initialization first, but now the driver tries again until it sees a proper vendor ID and at least one node:

$000F0000=$00000000
$000F0000=$83847690
$000F0004=$00000000
$000F0005=$00000000
$000F0000=$83847690
$000F0002=$00102201
$000F0004=$00010001
$001F0011=$C0000004
$001F2000=$102801BD
$001F000F=$0000000F
$001F000A=$00000000
$001F000B=$00000001
$00224011=$00000000
$00270610=$00000000
$00270500=$00000000
$00B3A01C=$00000000
$00B39033=$00000000

As expected I now see only one codec in HDAcfg.ini and Vendor ID and CODEC_AFG_xyz settings are populated correctly (?). But unfortunetly the volume control with waveout.exe still doesn't work. And what's really strange, my volume reduction in HDAICOUT.HDA also doesn't work anymore. So the speakers shout again with full volume. And I didn't manage to reduce the volume again. Therefore I decided to put the modem back in and told myself I should be happy with what I got: Audio output with fixed volume as set in HDAICOUT.HDA. 🙂

I have attached HDAcfg.ini and HDAICIN.TXT without modem for your information. If you ask me, I whould now keep it as it is (with modem back in and fixed volume). But if you would like to try something more, I will be happy to do further tests for you.

HDAcfg.ini HDAICIN.TXT

Edited April 24, 2023 by Vigami

[deomsh]

As such some changes without modem give interesting results, sorry to hear there are new problems involved.

First about the 'wait' states in HDACFG.INI making your computer slower: true if using system sounds or playing audio.

Second about 'if the 'commands' in HDAICOUT.HDA are executed before or after this driver initialization' - not sure. On the one hand I would say the driver must always first check the codec's properties before sending SET-verbs, but on the other hand finishing writing of HDALOG.TXT is a bit later than finishing writing of HDAICIN.TXT judged to file times.

However you have already a working system, without WAVEOUT.EXE, but with possibilities to maximize volume with your setting in HDAICOUT.HDA and a working volume slider if your player outputs to DirectSound.

But don't give up to fast, see your 'investments' in the HDA2.DLL driver as 'junk bonds': high possible payout, but also high risk to get nothing. Don't care about my efforts: to me trying to troubleshoot this driver is pure FUN.

Yesterday I revitalized my Virtual Box 6 Windows 98SE (setup /p i;s=detectbus) installation to test the capabilities of VBOX 7 (recent update). So I can do some experiments too.

During installation I noted my HDALOG.TXT seems fully different to yours. Print-screen with first few lines:

As you can see in my HDALOG.TXT there are comments inserted.

I am using HDA2.DLL from version HDADRV9L, 229.576 bytes (earlier I used HDA2.DLL from version HDADRV9J, 223.960 bytes. So check yours please.

Version 'L' of HDA2.DLL can be found on http://turkeys4me.byethost4.com/programs/index.htm

There is a chance on better 'behavior' of the driver, but of course 'worse' too.

Further you can try setting 'Verbinterface=$0' in HDACFG.INI, although this never solved anything is this thread so far. For some more details see: https://msfn.org/board/topic/178295-audio-driver-for-realtek-hd-audio-hardware-testing-thread/?do=findComment&comment=1162209

Another possibility is the HD-Audio Controller of your chipset is behaving differently without your modem, so maybe a candidate for setting 'pcipatchB' in HDACFG.INI. Your HD-Audio Controller is logged in HDACFG.INI:

[HDA]
TSR=TSR Found
PCI_VID=$8086
PCI_DID=$27D8

But, if I am right, in this thread HD-Audio Controller '27D8' (Intel) is not mentioned so far.

If I search for 8086 27D8 "quirk" there seems to be controller related problems, but no specific info found (Linux sources). Looking at the chipset of your laptop, your integrated chipset Intel 945PM seems to have Intel ICH7-U as Southbridge.

The Intel® High Definition Audio PCI Register from ICH7-U mentions that the so called 'No Snoop bit' is enabled by default. IF I am reading right.

Address Offset: 78h–79h Attribute: R/W, RO
Default Value: 0800h Size: 16 bits

Bit Description
15 Reserved
14:12 Max Read Request Size — RO. Hardwired to 0 enabling 128B maximum read request
size.
11
No Snoop Enable (NSNPEN) — R/W.
0 = The Intel® High Definition Audio controller will not set the No Snoop bit. In this
case, isochronous transfers will not use VC1 (VCi) even if it is enabled since VC1 is
not snooped. Isochronous transfers will use VC0.
1 = The Intel High Definition Audio controller is permitted to set the No Snoop bit in the
Requester Attributes of a bus master transaction. In this case, VC0 or VC1 may be
used for isochronous transfers.
Note: This bit is not reset on D3HOT to D0 transition; however, it is reset by PLTRST#.
10 Auxiliary Power Enable — RO. Hardwired to 0, indicating that Intel High Definition Audio
device does not draw AUX power
9 Phantom Function Enable — RO. Hardwired to 0 disabling phantom functions.
8 Extended Tag Field Enable — RO. Hardwired to 0 enabling 5-bit tag.
7:5 Max Payload Size — RO. Hardwired to 0 indicating 128B.
4 Enable Relaxed Ordering — RO. Hardwired to 0 disabling relaxed ordering.
3 Unsupported Request Reporting Enable — RO. Not implemented. Hardwired to 0.
2 Fatal Error Reporting Enable — RO. Not implemented. Hardwired to 0.
1 Non-Fatal Error Reporting Enable — RO. Not implemented. Hardwired to 0.
0 Correctable Error Reporting Enable — RO. Not implemented. Hardwired to 0.

So you can try 'pcipatchB=$7900' too, with AND without modem!

A small warning to readers of this thread: 'pcipatchB=$7900' is mentioned earlier, but not necessary a general value. Always check first in the datasheet of the chipset in use, or ask in this thread for help. Also about liability: writing to chipset registers is always at your own risk (if any).

Edited April 26, 2023 by deomsh
Corrections, addition