MattRocks wrote on 2026-01-13, 13:30:

Doom3 will query OpenAL because Doom3 is using OpenAL for mixing the sounds that it has already rendered. When Doom3 queries OpenAL, the Doom3 log is blindly repeating whatever OpenAL driver printed. The driver said it has 127 voices - that's true, but Doom3 doesn't use them.

I had some time today, so I fired up Doom 3 and used Falcosoft's OpenAL test tool to monitor its use of hardware voices. Here's a screenshot:

On my X-Fi card, Doom 3 used 127-30=97 hardware voices simultaneously during a firefight with multiple opponents. This can be easily replicated in Delta Labs, where the player is frequently attacked by several infested human enemies who use machine guns. I can provide the relevant savegame if someone else wants to test this.

Falcosoft wrote on 2026-01-28, 11:56:

I have updated my OpenALTest program to version 1.2. The post of the original download is modified.

If you're still working on it, can I request a few features?

1. Change the name of the "Actual" field to "Current" since that better describes it
2. Add a "Maximum" field which would record the highest number of hardware voices available when monitoring was started (basically the opposite of Minimum)
3. Add an "X-RAM Total/Free" field, similar to what AIDA64 has, if that's even possible to check in real time without hindering the game from using it

Thanks again for making this!

Joseph_Joestar wrote on 2026-01-29, 11:57:

If you're still working on it, can I request a few features? […]
Show full quote

Falcosoft wrote on 2026-01-28, 11:56:

I have updated my OpenALTest program to version 1.2. The post of the original download is modified.

If you're still working on it, can I request a few features?

1. Change the name of the "Actual" field to "Current" since that better describes it
2. Add a "Maximum" field which would record the highest number of hardware voices available when monitoring was started (basically the opposite of Minimum)
3. Add an "X-RAM Total/Free" field, similar to what AIDA64 has, if that's even possible to check in real time without hindering the game from using it

Thanks again for making this!

Hi,
OK, I have added the requested features.
Fortunately with the help of AL_EAX_RAM_SIZE and AL_EAX_RAM_FREE enumerations you can get the full/free X-RAM sizes without actually reserving X-RAM chunks.
Unfortunately I had to work half blind since I have no X-Fi for testing. I could only use OpenAL Soft that in newer versions emulates X-RAM.
But in case of OpenAL Soft the scope of both buffers and X-RAM is the running process so OpenAL Soft cannot be used for real testing (It always reports maximum buffers and X-RAM regardless what other OpenAL processes use).
So you should test it and report back if it is working or not on real hardware:

Falcosoft wrote on 2026-01-29, 15:25:

So you should test it and report back if it is working or not on real hardware:

It works fine, except for one detail. On my X-Fi Titanium Fatal1ty it detects 63360 X-RAM Maximum instead of 65536. I'm guessing 2MB are used for the DSP or something? In any case, here's a direct comparison between your tool and what Prey sees in-game after using the listSounds console command per Creative's guide.

BTW, your tool lists Available X-RAM Maximum as 63360 while monitoring is on, even when no game is running. But the X-RAM Size field below that correctly shows 65536 if I run your EAX test. Again, I remember reading that X-Fi cards use 2MB X-RAM for DSP cache, so that might explain it.

Joseph_Joestar wrote on 2026-01-29, 15:46:

It works fine, except for one detail. On my X-Fi Titanium Fatal1ty it detects 63360 X-RAM Maximum instead of 65536. I'm guessing […]
Show full quote

Falcosoft wrote on 2026-01-29, 15:25:

So you should test it and report back if it is working or not on real hardware:

It works fine, except for one detail. On my X-Fi Titanium Fatal1ty it detects 63360 X-RAM Maximum instead of 65536. I'm guessing 2MB are used for the DSP or something? In any case, here's a direct comparison between your tool and what Prey sees in-game after using the listSounds console command per Creative's guide.

The attachment Prey_Test02.jpg is no longer available

BTW, your tool lists Available X-RAM Maximum as 63360 while monitoring is on, even when no game is running. But the X-RAM Size field below that correctly shows 65536 if I run your EAX test. Again, I remember reading that X-Fi cards use 2MB X-RAM for DSP cache, so that might explain it.

Hi,
I think this works as expected. The X-RAM 'Monitoring' always asks how many bytes are 'Available'. In this respect it reflects the working of the buffer section. So the 'Maximum' part shows the maximum result of the 'get available X-RAM' calls.
If you want to get the full size of the X-RAM you can get it by pressing the 'Start' button instead. That part asks the full size regardless if it's available or not.

BTW, your X-RAM situation is analogous to hardware buffers:
As we all know neither the Audigy has 63 nor the X-fi has 127 buffers. The proper sizes are 64 and 128 respectively. But you never get the full amount of buffers as available since 1 is always reserved for Windows even when absolutely no games/audio applications are running.
(In case of SB Live! under WinXP the available buffers are never more than 30, while SB Live! really has 32).

Regarding your screenshot:
It says that '65076 KB total onboard memory is used'. If you look at the currently available X-RAM value on my program then you can see 460 KB. 65536 - 65076 = 460 exactly. So the results of the 2 calculations correlate perfectly.

Falcosoft wrote on 2026-01-29, 16:22:

Regarding your screenshot:
It says that '65076 KB total onboard memory is used'. If you look at the currently available X-RAM value on my program then you can see 460 KB. 65536 - 65076 = 460 exactly. So the results of the 2 calculations correlate perfectly.

Heh, I made that connection too.

So I'm guessing 2MB is always used for the X-Fi's DSP cache, even if nothing is actually running at that time. Creative's guide that I linked to earlier makes some references to those 2MB as well. Anyway, nice work on the testing tool!

I wanted to test a game that isn't based on IDTech4, so I fired up Unreal Tournament 3. Here's a screenshot:

It uses 127-37=90 simultaneous hardware voices in a deathmatch game with 32 bots. Doesn't seem to utilize X-RAM a whole lot though, unlike Prey.

Joseph_Joestar wrote on 2026-01-29, 22:04:

I wanted to test a game that isn't based on IDTech4, so I fired up Unreal Tournament 3. Here's a screenshot:

The attachment UT3_Test01.jpg is no longer available

It uses 127-37=90 simultaneous hardware voices in a deathmatch game with 32 bots. Doesn't seem to utilize X-RAM a whole lot though, unlike Prey.

Hmm, it's interesting since I have thought that X-RAM is automatically utilized by default even in case of missing explicit support. It's in the OpenAL programming guide where the X-RAM extension is mentioned:

X-RAM Modes The X-RAM extension to Open AL has two modes of operation – an ‘automatic’ mode (the default) and a ‘managed’ mode. […]
Show full quote

X-RAM Modes
The X-RAM extension to Open AL has two modes of operation – an ‘automatic’ mode (the
default) and a ‘managed’ mode. In automatic mode an application does not need to make any
function calls, or even query for any extensions, and Open AL buffers will automatically be loaded
into X-RAM if it is found and has enough storage space. In managed mode the application
developer has complete control over which Open AL Buffers are uploaded to X-RAM or not.

https://www.openal.org/documentation/OpenAL_P … mmers_Guide.pdf

Would you do a short test?
1. Start 2 instances of my test app.
2. Press 'Start Monitoring' on the 1st instance.
3. Press 'Start' and then 'Play' on the 2nd instance.
4. Watch if the Current X-RAM value changes on the 1st instance or not.

My test program neither enables nor disables X-RAM explicitly. I wonder what happens in such a case.
(The played footstep sample is about 276 KB so approximately this amount of X-RAM should be reserved).

Falcosoft wrote on 2026-01-29, 22:18:

Would you do a short test? 1. Start 2 instances of my test app. 2. Press 'Start Monitoring' on the 1st instance. 3. Press 'Star […]
Show full quote

Would you do a short test?
1. Start 2 instances of my test app.
2. Press 'Start Monitoring' on the 1st instance.
3. Press 'Start' and then 'Play' on the 2nd instance.
4. Watch if the Current X-RAM value changes on the 1st instance or not.

My test program neither enables nor disables X-RAM explicitly. I wonder what happens in such a case.

Sure thing, here you go:

X-RAM utilization did increase slightly while the test was running.

Joseph_Joestar wrote on 2026-01-29, 22:33:

Sure thing, here you go: […]
Show full quote

Falcosoft wrote on 2026-01-29, 22:18:

Would you do a short test? 1. Start 2 instances of my test app. 2. Press 'Start Monitoring' on the 1st instance. 3. Press 'Star […]
Show full quote

Would you do a short test?
1. Start 2 instances of my test app.
2. Press 'Start Monitoring' on the 1st instance.
3. Press 'Start' and then 'Play' on the 2nd instance.
4. Watch if the Current X-RAM value changes on the 1st instance or not.

My test program neither enables nor disables X-RAM explicitly. I wonder what happens in such a case.

Sure thing, here you go:

The attachment OpenAL_Test02.jpg is no longer available

X-RAM utilization did increase slightly while the test was running.

The played footstep sample is about 276 KB so approximately this amount of X-RAM should be reserved.
And apparently this is the case (280 KB is reserved).
So it seems X-RAM is explicitly disabled in UT3 for some reasons. Or maybe the following happens ?

If a future alBufferData call is made on a buffer in automatic mode, the driver will assume that the
application is using the AL Buffer for streaming (requiring regular updates to the audio data in the
buffer), and the sample data will be moved from X-RAM to host memory.

Apparently sound through Creative's OpenAL still plays with disabled sound card in Windows Sound control panel.

I just tested Battlefield 2142. Verdun map, 16 player botmatch. Here's a screenshot:

It uses 127-13=114 simultaneous hardware voices. This really kicks off when there are multiple enemies and vehicles on the screen. X-RAM utilization is pretty good too.

Tested Quake 4. Here's a screenshot from the boss fight in the Nexus:

The game was using 127-47=80 simultaneous hardware voices at that time. X-RAM was almost fully utilized as well.

Joseph_Joestar wrote on 2026-01-30, 13:28:

Tested Quake 4. Here's a screenshot from the boss fight in the Nexus:

The attachment Quake4_Test01.jpg is no longer available

The game was using 127-47=80 simultaneous hardware voices at that time. X-RAM was almost fully utilized as well.

Hi,
It seems the test program works correctly with most OpenAL games. Would you add the download link of version 1.3 to the 1st post of this topcic? So if someone wants to try it for testing then one does not have to find the link 49 pages ahead.
(BTW, the test program is a 32-bt one and according to my tests works anything from Win98 SE to Win 11)
I think it's also proven that through OpenAL Creative's legacy HW accelerated cards can be used to their full potential even on modern (Vista+) Windows versions.

Falcosoft wrote on 2026-01-30, 13:49:

Hi,
It seems the test program works correctly with most OpenAL games. Would you add the download link of version 1.3 to the 1st post of this topcic? So if someone wants to try it for testing then one does not have to find the link 49 pages ahead.

Done. And thanks again for making this!

Falcosoft wrote on 2026-01-30, 13:49:

(BTW, the test program is a 32-bt one and according to my tests works anything from Win98 SE to Win 11)

Nice! I'll give it a try under WinXP at some point.

Unreal Tournament 2004 + Creative's EAX 5.0 patch for that game. Deathmatch with 32 bots:

It's using 127-44=83 simultaneous hardware voices. X-RAM is nearly maxed out.

Did another test with Doom 3 to see how much X-RAM it uses:

This seems to vary depending on which map is loaded, but it's still a pretty sizeable chunk that gets utilized. Also, I got to 127-28=99 simultaneous hardware voices by leaving more enemies alive during the firefight. Delta Labs Sector 3 is one of the best levels to test this due to all the ambushes there.

And here's Battlefield 2. Operation Harvest map, 16 player botmatch.

It uses 127-25=102 simultaneous hardware voices. As usual, having several vehicles and enemies on screen makes this number go up significantly. Also, X-RAM is nearly maxed out here.

P.S.

Battlefield 2 only allows 16 players for its botmatch mode. In real online multiplayer, you could have up to 64 players, which would probably increase the hardware voice count even further.

To all people with any Audigy card lurking here and running Win10 or Win11...

Just curious whether they can also reproduce the issue I describe here.

And those with X-Fi cards, the proper ones with HW EAX I mean, can they do the same when opening whatever X-Fi cards come with to tweak EAX?

Interestingly, you can break the control panel and not necessarily OpenAL EAX along with it. But possible to break both. I wish there was a driver reset that doesn't require reboot.

UCyborg wrote on 2026-01-31, 22:46:

To all people with any Audigy card lurking here and running Win10 or Win11... […]
Show full quote

To all people with any Audigy card lurking here and running Win10 or Win11...

Just curious whether they can also reproduce the issue I describe here.

And those with X-Fi cards, the proper ones with HW EAX I mean, can they do the same when opening whatever X-Fi cards come with to tweak EAX?

Interestingly, you can break the control panel and not necessarily OpenAL EAX along with it. But possible to break both. I wish there was a driver reset that doesn't require reboot.

Hi,
I could reproduce the part that when OpenAL + EAX is running then all Creative control panel items show missing EAX effects. But EAX effects are still working in programs/games after opening control panel items. And when the program using OpenAL + EAX is closed then everything is back to normal, control panel items show the previously selected/active effect. Tested with Audigy1 on Windows 10 using the last but one DanielK drivers.