by Ilya Gavrichenkov
05/23/2000 | 12:00 AM
Aureal company needs no presentation. It is famous all over the world for its technologies and soundprocessors. In 1999 Aureal started production of its own brand-name sound cards based on its own sound chips.Today we'll talk about SQ2500 sound card from Aureal and compare it with a very popular MX300 sound card fromDiamond. S3, which swallowed Diamond Multimedia, is known to have given up the manufacturing of sound cards basedon Aureal chips and switched to ESS production. Nevertheless, Vortex2 based MX300 sound cards are still on the marketand are very popular even without driver support from S3. In our review we'll try to give an answer to the question: isSQ2500 a worthy replacement to MX300?
SQ2500 sound cards are supplied in two variations: OEM and Retail. A $60 OEM package contains the sound card, a CD withdrivers and several applications and a pretty poor user's manual. The $90 Retail version of the product apart from a nice boxincludes two full games (Heretic II and Drakan: Order of the Flame) and an OEM version of Slave Zero. It's up to you to decideon one or another package. SQ2500 sound cards can differ only by the type of S/PDIF digital output: it can be an optical TosLINKor an electrical RCA . Furthermore, the electric S/PDIF output can be the first from the top or the second just below the line-injack on the card's bracket. As in all supply variations SQ2500 has the same Vortex2 sound chip, which determines the performanceand sound quality, hereinafter we will talk about SQ2500 in general.
The heart of the card is a Vortex2 (AU8830) sound chip from Aureal. Vortex2 represents a second generation of Aureal'ssound processors and contains 3 million transistors. Aureal claims that Vortex2 performance is 600MIPS (Million InstructionsPer Second). Basically many sound processor manufacturers mention very high performance figures in MIPS compared to someordinary DSPs by TI or Motorola. Moreover, sometimes these sound processors are called DSPs, which is completely wrong.Hereinafter we'll refer to sound processors as to "sound DSPs" in order to accentuate their difference from ordinaryDSPs. In fact, sound DSPs used in sound cards have their specific goals, e.g. HRTF-filtering of a particular memoryblock or rerouting a sound stream to a DAC. Thus, sound DSPs operate sound streams, which is controlled by a specialinstructions system. When programming sound DSPs there's no need in a signal processing program and you just have tocall a built-in or external function like "transfer a signal with the following parameters from some input point tosome block for further processing". This instruction system is of a higher level than ordinary DSPs, which is the mainrestriction imposed over Vortex2 or EMU10K1 sound DSPs. As a result, sound DSPs can be used only for their specificpurposes. They can't perform operations like addition or multiplication and although their performance looks impressiveit can't be compared to ordinary DSPs' performance even if they're both measured in MIPS. Moreover, high MIPS values canbe obtained due to performing several similar operations in parallel, e.g. simultaneous HRTF-filtering of 16 sound streams.
The card is equipped with the following connectors:
We started our SQ2500 & MX300 comparison with the examination of their appearance and found the following differences:
|On the left is a Vortex2 chip on a MX300, on the right - a Vortex2 on a SQ2500. Note the differencein marking.|
However, these are just the external differences, which do not influence the sound quality. Of course, many sound cardsmay sound similarly but there are still some nuances between them. Sound coloring depends on many factors, such as cardlayout quality, discrete component filter quality, the spread of component parameters, etc. The problem is that soundvariations aren't easy to notice, as it in many respects depends on the ear, acoustics and type of music being played.As for us, we can't boast a cool musical ear, it's just the average one, and acoustics is also quite mediocre that's whyin our subjective opinion there's no difference in sounding between MX300 and SQ2500 although there may be a huge differencein performance. Don't forget that the main goal of our test is to find out all these differences. Or not to find any.
Before getting down to practice let's take a look at Aureal's new A3D 3.0 interface, especially since the drivers forVortex2 based sound cards already support it partially and there are even several new games supporting this new API (forinstance, Soldier of Fortune).
So, A3D 3.0 API is first of all a powerful 3D sound modeling tool, capable of reproducing all three major sound types:
Moreover, A3D 3.0 includes a universal sound engine that allows reproducing 3D sound with any hardware available, either withsound DSPs (for example, from Creative or ESS) or with sound HSPs (for example, from Yamaha).
It is for the first time throughout the entire history of sound industry that A3D 3.0 supports the following:
It's easy to notice that A3D 3.0 API is a universal and ultimate solution today. This gives us every right to hope thatAureal's ideas will be developed further on.
Background music in MP3 and AC-3 format will be decoded and reproduced on the fly with the help of hardware or CPUresources. The so long-awaited SQ3500 sound card from Aureal will have an integrated Motorola hardware decoder on board.By the way, SQ2500 has two connectors that can be used for upgrade. The installation of a daughter card with Motorola DSPis also very likely to be possible, although there are no official comments about this yet.
Before we pass over to real practice take a look at our testing system, which was configured as follows:
We used two sets of reference drivers when testing sound cards: 2.041 + A3D 2.25 and 2.048 + A3D 3.12. Alldrivers can be installed and removed without any difficulties. The new 2.048 version has initial A3Dverbreverberation engine support that allows reproducing reverberation through EAX 1.0 interface only. Consequently,new options appeared in the sound card properties:
The best way to demonstrate A3Dverb features is the A3D 3.0 Player demo.
If you don't have this player, download an update for A3D 2.0 Player here: playerupdate.exe (130KB)
By the way, here http://saturn.spaceports.com/~a3dfan/ you can download a set of Aureal demo program updates(you can also find new beta-versions of reference drivers there), which will let you listen to them via A3D 3.0.For example, you can hear volumetric sound sources in the A3D Race demo. We think, the idea of volumetric soundsources modeling "sounds" much like the future ZoomFX interface from Sensaura company.
Another new feature in 2.048 drivers is the possibility to upload SF2 sound banks, but it's still very limitedbecause only 2MB, 4MB and 8MB SF2 2.0 banks can be uploaded. In fact, it's a long-awaited option because basicallyAureal ARL format is plainly the SF2 2.0 but with its own addressing system. Although Aureal drivers have only GeneralMIDI support, the progress made is hardly deniable.
To compare the sounding of MX300 and SQ2500 we launched several demo programs from Aureal and a few games with A3D2.0 and EAX 1.0 support but didn't hear any difference. As to the 3D sound perception in general, we got the bestimpression when listening via headphones. Then comes 4.1 acoustics, and only then comes a 2-speaker configuration.
Now let's get to tests of the sound cards hardware features, namely of their sound chips, and see how theyperform in some games.
First of all we launched Minerva 1.08 test package from Aureal on both sound cards with two driver sets. Wegot the following results. During playback through headphones or 2 speakers 76 DS3D hardware streams were involved,some of them being dynamically reserved for early reverberated sounds playback. During playback through 4 speakers 16DS3D hardware streams were involved and other 60 streams were strictly reserved for early reflection. But in case ofMX300 we witnessed an interesting situation: with 2.041 drivers and playback through headphones or 2 speakers all 76DS3D hardware streams were detected but with 2.048 drivers - only 16 DS3D hardware streams could be detected, unlikeSQ2500. As a matter of fact, the number of DS3D streams supported via hardware isn't very important because lackingsound streams can be emulated and controlled by system CPU. As to the CPU utilization increase during software controlof sound streams, tests showed that general performance in real applications remains almost unaffected.
Let's see how CPU utilization changes depending on the number of supported hardware DS3D streams and on the driverversion. For this purpose we will use ZD AudioWinbench99 testing utility. When testing CPU utilization we set headphonesand 4.1 acoustics (Quad) playback. This table shows CPU utilization in percents during playback and mixing of DS3D streamswith a 44.1KHz sampling frequency and 16bit precision:
|MX300 2041 (Quad)||SQ2500 2041 (Quad)||MX300 2048 (Quad)||SQ2500 2048 (Quad)||SBLive! Quad|
|Hardware DS3D streams||32 (16)||32 (16)||16 (16)||32 (16)||32|
|Streaming: Voice 8||1.09 (0.7)||1.03 (0.933)||0.697 (0.706)||1.42 (0.851)||1.1|
|Streaming: Voice 16||1.05 (1.54)||1.93 (1.65)||0.651 (1.24)||1.3 (1.6)||1.31|
|Streaming: Voice 32||1.57 (3.33)||1.62(3.46)||3.24 (3.48)||1.89 (3.8)||2.72|
The table shows results for Streaming Buffer only because Static and Streaming mean the same thing for modern PCI sound cards.The matter is that today's sound cards don't have local on-board sound memory, so, all the data is stored and processed in thecomputer's system memory. Besides, the data from Streaming Buffer can be drawn all at once, as from a Static buffer, and inportions, i.e. as a stream.
So, both Vortex2 chips on MX300 and SQ2500 sound cards support the same number of hardware DS3D streams though their markingis different. The fact that test packages show different number of hardware DS3D streams during headphones and 4 speakers playbackdoesn't really influence the test results, which can be seen from the table. Let's also note that even during partial DS3D streamsemulation the CPU utilization doesn't exceed 4% which is very little.
All of these were synthetic tests, now what about CPU utilization in real applications? Unfortunately, today's one and onlyconvenient and universal gaming test is Half-Life, because you can set exactly which interface your sound card will use for soundplayback and modeling. Also this game has several demos that allow fps measuring.
We used Half-Life with a 184.108.40.206 (build 893) patch and an UpLink Lite add-on that you can download here: uplink_lite.exe(3.8MB)
For fps measuring we used Survive demo that can be downloaded from Reverend Pulpit web-site. This demo loads the CPU andgraphics accelerator pretty heavily. All the tests in all games were run with the sound quality set to High Quality. Testingwas performed with a graphics accelerator working in OpenGL API with a 16bit color depth. At 320?240 resolution we used onlysoftware graphics rendering.
At first we ran the tests using a 2.041 + 2.25 driver set and here's what we got:
|No Sound||MX300 A3D 2.0 (QUAD)||SQ2500 A3D 2.0 (QUAD)|
|Soft 320x240||50.6||35.2 (34.93)||40.58 (40.8)|
|640x480||70.9||44.98 (45.02)||50.55 (50.43)|
|1024x768||69.2||43.7 (44.5)||50.57 (50.2)|
|1280x960||63.8||44.02 (44.39)||50.05 (50.17)|
You can see that CPU utilization during playback through headphones and 4 speakers (QUAD) is almost the same,because fps rates are almost equal. But CPU utilization is different if you compare SQ2500 and MX300. On average,SQ2500 loads CPU 14% less, i.e. during sound playback via A3D 2.0 API on a SQ2500 sound card you'll get a 14% higherfps rate, which is a quite considerable performance increase. If compared to the fps rate measured with the sounddisabled, MX300 increases CPU utilization from 30% to 37% and SQ2500 - from 20% to 29%. It's interesting that at lowresolutions when CPU utilization is maximal, the sound card also loads the central processor more than at higherresolutions.
Then we installed a 2.048 + 3.12 driver set and made new measurements:
|No Sound||MX300||SQ2500||SB Live! Value|
|A3D 2.0 (QUAD)||EAX 1.0 (QUAD)||A3D 2.0 (QUAD)||EAX 1.0 (QUAD)||EAX 1.0|
|Soft 320x240||50.6||34.5 (35.04)||34.22 (34.88)||39.72 (39.99)||48.02 (47.68)||46.31|
|640x480||70.9||44.4 (45.3)||46.04 (46.27)||51.74 (53.33)||64.37 (63.77)||62.45|
|1024x768||69.2||42.9 (42.8)||45.13 (45.23)||52.04 (52.65)||63.26 (62.89)||61.13|
|1280x960||63.8||43.07 (44.3)||44.19 (44.32)||49.62 (51.44)||60.66 (60.24)||58.6|
If we take a general look at figures in the table, we will see that the situation with CPU utilization duringsound playback via A3D 2.0 in both cases, on MX300 and SQ2500, remains unchanged. There's a little progress bySQ2500 in 4 speakers (QUAD) playback mode. This is rather strange because apart from HRTF-filtering, a Cross-talkCancellation (CC) algorithms must be calculated. The fact that fps rate hasn't decreased means that CC algorithm iscalculated by Vortex2 and not by the CPU. The 1-3% fps increase we observed could be explained by measuring errors orby some unexpected reserves of the chip, which is quite unlikely to be true.
As 2.048 drivers support EAX 1.0, we also provide the fps rates for SB Live! Value sound card just for a bettercomparison.
Working via EAX 1.0, reverberation is calculated on the basis of acoustic environment presets. Sound sourcespositioning and calculation of straight sound directions is done via DS3D. In case of A3D 2.0, the playback ofsounds propagated along straight directions, sound sources positioning and early reverberated sounds calculationsare carried out via a special engine included into A3D 2.0 or via the game's own integrated sound engine. Earlyreverberated sounds calculations are effected basing on acoustic environment geometry info in real-time mode withWavetracing technology. It's obvious that operations like that require much more computing resources and more RAMfor more efficient data exchange. If the gaming sound is programmed to be reproduced via A3D 3.0 API, reverberationcan be calculated in two ways: on the basis of presets (via EAX API and I3Dl2) or it can be dynamically modeled on thebasis of acoustic environment geometry data with Wavetracing technology.
In case of our Half-Life test game, reverberation can be reproduced via EAX 1.0 only, i.e. on the basis of presets.Our table shows a very interesting picture. When sound is played back by MX300 sound card via EAX 1.0, fps rate figuresare close to those occurring with A3D 2.0. At two other resolutions we can notice a 5% fps rate increase, which can beexplained both by measuring errors and by the fact that sound reproduction via EAX requires less computing resources thanA3D 2.0. But in case of SQ2500, we can observe fps rates almost equal (in fact they're even 3% higher) to those taken withSB Live! Value. In fact, SQ2500 loads the CPU at least as much as EMU10K1 DSP based SB Live! Value and at the most even alittle less. It would be unfair to say that these results are impressive: they are more than impressive. But let's put ouremotions away, there's a discussion board for sharing them. ;)
Let's compare the fps rate measured during sound playback via EAX + DS3D and obtained with game sound disabled. SB Live!Value sound card loads CPU during reverberation reproduction from 8% to 12%, and SQ2500 under the same conditions increasesthe central processor utilization from 5% to 10%.
Thus we can say that Vortex2 chips used on MX300 and SQ2500 sound cards differ from each other. A new version of Vortex2used on SQ2500 provides higher general performance in games, which show the fps values. In this case the performance increasecan be observed when sound is reproduced via both: A3D 2.0 API and DS3D API + EAX 1.0. It seems to us that these improvementsin the new Vortex2 chip version were achieved through optimization of data transfer to the system memory and possibly throughinternal chip architecture modifications. Maybe Aureal's engineers just fixed some bugs inside the chip. Anyway, the mostimportant thing is the result, which is far more than satisfactory.
Apart from plain gaming performance there's also one more thing, which can't be expressed numerically. We're talking aboutsounding quality in games. Sounding in games via A3D 2.0 has been discussed in and out already and there's no point in repeatingthe universal truths about early reflections dynamically calculated on the basis of environment geometry that allow locatingsound sources and estimating distance to them. This feature is available only on Vortex2 based sound cards. The sound propagationalong straight directions isn't an interesting subject either because any sound card with A3D or DS3D support easily copes with it.There's the third and last component left: late reverberation. Until recently only Aureal chip based sound cards didn't havethis feature. Now this situation is improved and they feature this support though on initial stage only. This means that it'sstill too early to expect the current Vortex2 based cards drivers to reproduce reverberation of the same quality as EMU10K1based sound cards can provide. For example, according to many listeners the quality of reverberation in Half-Life produced byVortex2 based sound cards is not in the least inferior to SB Live! sound cards sounding. On the whole the situation with thereverberation quality of Vortex2 based sound cards is the following: in some games reverberation quality is OK; in some gamesreverberation quality can be regarded as acceptable but different (sometimes quite significantly) from SB Live! sounding; insome games reverberation is nauseating or totally absent.
If you've never heard how reverberation sounds on SB Live! sound cards, it will be difficult for you to estimate thereverberation quality on Vortex2 based sound cards. If you have the possibility to compare, you'll be able to form yourown opinion. Moreover, there is every reason to expect the drivers for Vortex2 based sound cards with full reverberationsupport. That's why we won't give a list of games, in which we didn't like the reverberation quality. But reverberationwe heard in Half-Life, Need4Speed: High Stakes, Test Drive 6 and Baldur's Gate was very nice. In Torment reverberation wasnot very vivid but we can't totally disregard it.
By the way, currently the owners of Vortex2 based sound cards will have to choose between non-reverberated sound playbackvia A3D 2.0 or non-early reverberated playback via DS3D (DS) + EAX 1.0. Only in games with A3D 3.0 API sound modeling asimultaneous reproduction of all three 3D sound components will be possible. These components include: sounds, propagatedalong straight directions, early reverberated sounds and late reverberation. But in this case you'll need drivers withcomplete A3Dverb support.
If the game uses not only the A3D 3.0 interface but also Aureal's own sound engine (which is absolutely free), userswith any sound cards will be able to hear 3D sound and background soundtracks in MP3/AC-3 formats except dynamicallycalculated early reflections which require hardware support.
If you visit some 3D sound-related discussion boards you'll come across numerous discussions of notable performancedrops occurring with A3D 2.0 interface. We would like to mention that during any sound playback the performance drop isinevitable. First of all, it's caused by the sound rendering performed in the buffers formed in computer's system memory.Additional resources are required for special effects of different complexity that are used for 3D sound modeling. Ifsound cards had local sound memory like graphics accelerators do, gaming performance would increase together with thesound cards' prices. Moreover, the performance greatly depends on the gaming sound engine. It's not a secret that gamedevelopers can use their own sound engine or take some ready-to-use solutions from other companies including Aureal. Ofcourse, different engines have different optimization level. We took the fps rates in another two popular games: inQuake3 Arena, which uses Aureal sound engine and in Unreal Tournament, which uses its own sound engine. 2.048 + 3.12driver set was used.
Let's start with Quake3 Arena with 1.16n patch installed. We used a built-in Demo002 demo with settings set toNORMAL, only resolutions were changed. We took the fps rates with sound disabled, with sound played back via A3D2.0 and via DirectSound (DS):
|No Sound||MX300 (DS)||SQ2500 (DS)||SB Live!, DS|
|640x480||105.2||61.75 (85.2)||64.15 (88.45)||87.5|
|1024x768||87.05||60.45 (80.45)||62.95 (82.1)||81.5|
|1280x1024||56.4||53.15 (56.2)||54.25 (56.3)||56.3|
You can see that during playback via DS all three sound cards load the CPU almost equally. With the growth of theresolution the fps rate limiting factor turns the graphics accelerator performance (namely, its fillrate) and not theCPU resources used for sound rendering. A similar thing happens to sound playback via A3D 2.0 API: the higher is theresolution, the less noticeable is the CPU utilization by the sound card. Moreover, the table shows that Vortex2 chipon SQ2500 does its job better than his brother on MX300.
Now let's take a look at the situation with fps rates in Unreal Tournament. Before we started our tests, a v.405patch was installed and during the test Hardware 3D Sound was successively enabled or disabled. The video settingswere configured as follows:
For our test we used Wicked405 demo form the Reverend Pulpit web-site.This demo loads CPU heavily while performance in UT depends more on the graphics accelerator performance (namely on its fillrate).
|No Sound||MX300 (DS)||SQ2500 (DS)||SB Live! (DS)|
|640x480||31.53||27.35 (30.12)||26.72 (29.16)||30.49 (30.22)|
|1024x768||25.58||22.65 (24.74)||22.46 (24.3)||24.84 (24.8)|
|1280x1024||25.67||21.72 (23.73)||21.65 (24.33)||23.69 (23.74)|
Results show nothing new. The higher gets the resolution, the less is the fps rate difference. DS3D + EAX or DS only allowreducing the CPU utilization compared to A3D 2.0. The only interesting thing is that during playback via A3D 2.0 MX300 performedbetter than SQ2500. This is due to the fact that UT uses its own sound engine optimized for MX300. If we assume that Vortex2 chipof MX300 has some internal bugs, then this optimization won't help when another bugless Vortex2 chip is used.
Fps rates analysis allows us to make the following conclusion: if you play a single-player game, select higherresolutions. The performance drop caused by your sound card in this case will be minimal. But if you're playingdeathmatch, you'll have to make your own choice depending on your preferences.
Finally let's discuss one more great difference between MX300 and SQ2500 sound cards: the digital interface.In our opinion, any modern sound card should have digital interface. Why? Well, at least because DVD-video becomesone of the most popular ways of entertainment and multimedia kits of five speakers, a subwoofer, an amplifier and aAC-3 decoder cost just $170-250. All these AC-3 multimedia decoders have an RCA electrical digital S/PDIF In.
In order to get a digital out on your MX300 you can buy an MX-25 daughter card for approximately $40. The otherway of equipping your MX300 with a digital out will suit you if you are a RadioShack patron accustomed to soldering.If it's exactly what you are then pay a visit to the following web-site: http://www.algonet.se/~cyrano/mx2.html whereyou'll find useful advice on MX300 upgrade.
SQ2500 sound card won't be such a pain in the ass because it's equipped with a digital electrical RCA S/PDIF out.We tested the performance of this digital out by connecting SQ2500 to Cambridge DeskTop Theater 5.1 DTT2500 Digitalmultimedia kit with a built-in AC-3 decoder and listened to sounding during DVD video playback. The sound was wonderful.
However, portable MiniDisc recorders owners would appreciate an optical TosLINK digital S/PDIF out. Unfortunately,this is impossible. In theory SQ2500 sound card specifications allows optical digital out installation but we couldn'tfind these models on the market. Still you can purchase Aureal Vortex2 based SQ Digital PCI sound cards (sometimeslabeled as Xitel, VideoLogic or Terratec), which are equipped with an optical digital out. But those have nothing todo with SQ2500 and they're based on older Vortex2 chips, the same as MX300 is based on. SQ Digital PCI usually havedark brown or dark green textolite.
Aureal SQ2500 sound card is $25 more expensive than MX300 but for this price you get a better performance in games,an electrical digital out and wonderful 3D sound. Anyway, the choice is yours, but in our opinion ,MX300's time slippedaway and there's no point in purchasing it now.
Among SQ2500's lows can only be mentioned the unclear situation with new drivers and Aureal's fate in general.
P.S. One day Media Vision company, ProAudio Spectrum family manufacturer, announced its bankruptcy.Some time later appeared a new company called Aureal. The name's not important but ideas and technologies are. Let's beoptimistic.