![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
Alternatives to Creative?
- Thread starter Xenarchy
- Start date
alg7_munif
Supreme [H]ardness
- Joined
- Oct 9, 2006
- Messages
- 5,862
Check out the first post of my X-Meridian thread, I've listed some cards there.
I picked up the new Razer card today, its based on the same chipseti believe as the x-meridian this thing is incredible... I have it running to my receiver because one of my channels on my logitech 5.1 amp went and i hate using it with a dead channel :-( DDL sounds awesome though, even music seperation and upmixing sounds great, i havent tried any games yet but ill post some results after I have a chance, the benchmarks were good from what ive seen and the DSP on these cards are the only thing even close to compete w/ the XFI. The Razer driver interface is alot nicer than the OEM control panel I remember with my X-Mystique a couple years back, impressed so far i have to say, but ouch at 199.99 was more than my fatal1ty.
alg7_munif
Supreme [H]ardness
- Joined
- Oct 9, 2006
- Messages
- 5,862
I don't understand why people said that X-Fi has a better hardware acceleration in games. I'm not an electronical engineer but after a quick reading in the internet I think that I've got a better picture about the sound chip thing but I might be wrong.
From what I've read:
The CPU is not doing the sound processing because you just need a DAC if the sound is already digitally processed by the CPU according to wikipedia. All sound cards need to have a DSP to do the sound processing according to this
An interesting article about DSP.
It says that it is cheaper to design and build a single chip microprocessor that will only do specific tasks than building a multipurpose processor(the CPU). A specific processor could even do the task better than a multipurpose processor. The specific processor will take the burden off the CPU, that's why a sound chip is still needed to do the sound processing. Even an onboard sound chip will do the digital signal processing in its chip so there is no doubt that the processing is done in hardware. The real issue now is how good the sound chip will do its job to take the load of the CPU. If the processing is only done by the CPU, only a DAC is needed to convert the digital signal to analogue sound.
The X-Fi has a powerful EMU20K1 sound chip only because it needs it, the Digital Signal Processing only needs about 11% of the chip capabilities but 70% of the chip capabilities is used for Sample Rate Conversion because its DSP can only work with a 48KHz sample rate according to this. The X-Fi's DSP can also work with 44.1KHz but only in creation mode.
The old DSP from Audigy 2 also had this problem but it uses an external Sample Rate Converter according to this source
Creative's EMU20K1 can only do up to 96K/24bit if more than 2 channels are used and the X-Fi came with an onboard RAM because it needs it
I think that the C-Media CMI-8788 Oxygen HD(on my X-Meridian) is a more powerful chip than the EMU20K1. I don't know how the CMI-8788 work, either it can work with variable sample rate or it also has an internal SRC because no external SRC is required on the X-Meridian. The CMI-8788 actually support 10-channel 192KHz/32bit(Auzentech only states 24bit) output, 8-channels are used with analogue and 2-channels with SPDIF. The recording input is also at 192KHz/32bit. For more info see this datasheet, go to page 9 for a simple diagramm.
According to the datasheet all the Dolby features are software features, that's why people are getting lower fps only when DD/DTS encoding is turned on.
From what I've read:
The CPU is not doing the sound processing because you just need a DAC if the sound is already digitally processed by the CPU according to wikipedia. All sound cards need to have a DSP to do the sound processing according to this
An interesting article about DSP.
It says that it is cheaper to design and build a single chip microprocessor that will only do specific tasks than building a multipurpose processor(the CPU). A specific processor could even do the task better than a multipurpose processor. The specific processor will take the burden off the CPU, that's why a sound chip is still needed to do the sound processing. Even an onboard sound chip will do the digital signal processing in its chip so there is no doubt that the processing is done in hardware. The real issue now is how good the sound chip will do its job to take the load of the CPU. If the processing is only done by the CPU, only a DAC is needed to convert the digital signal to analogue sound.
The X-Fi has a powerful EMU20K1 sound chip only because it needs it, the Digital Signal Processing only needs about 11% of the chip capabilities but 70% of the chip capabilities is used for Sample Rate Conversion because its DSP can only work with a 48KHz sample rate according to this. The X-Fi's DSP can also work with 44.1KHz but only in creation mode.
The old DSP from Audigy 2 also had this problem but it uses an external Sample Rate Converter according to this source
Creative's EMU20K1 can only do up to 96K/24bit if more than 2 channels are used and the X-Fi came with an onboard RAM because it needs it
I think that the C-Media CMI-8788 Oxygen HD(on my X-Meridian) is a more powerful chip than the EMU20K1. I don't know how the CMI-8788 work, either it can work with variable sample rate or it also has an internal SRC because no external SRC is required on the X-Meridian. The CMI-8788 actually support 10-channel 192KHz/32bit(Auzentech only states 24bit) output, 8-channels are used with analogue and 2-channels with SPDIF. The recording input is also at 192KHz/32bit. For more info see this datasheet, go to page 9 for a simple diagramm.
According to the datasheet all the Dolby features are software features, that's why people are getting lower fps only when DD/DTS encoding is turned on.
OK,first of all, the wikipedia article you make reference lists a few characteristics of a DSP, one of the defining is an instruction set with multiply-accumulate instructions that make digital filtering a couple of instructions instead of a loop. If a sound chip doesn't have those instrucctions it's hard to regard it as a DSP. The sounchip is still necessary to gain access to the bus, to the DMA channels, it can offload the CPU with the mixing and the voice management, even hardwired complex algorithms like the vortex2, not to mention electrical issues, you cannot just drop a DAC next to the CPU as you suggest.
Moving to your next statement, the percentage of transistors that a chip uses for certain tasks is not indicative of its computational power, it's like saying that a CPU design is not good because more than half of its transistors are spent in L2-cache and the FPU is a tiny portion. Simply there are some circuits that translate into more transistors (or gates if you design in FPGAs). As I stated before, DSP instructions vastly simplify digital filtering implementation, so all the effects just need to use different filter tap coefficients in the same functional unit.
I think it has been told repeatedly that the cards (not the chips) limit the output to 24/96 in surround because it's the specification of DVD-Audio, until some media uses 24/192 on eight channels, the utility of such feature will become apparent.
You could argue that the C-Medida chip is a more capable chip but certainly not a more powerful since it doesn't even have an instruction set nor hardwired algorithms.
Moving to your next statement, the percentage of transistors that a chip uses for certain tasks is not indicative of its computational power, it's like saying that a CPU design is not good because more than half of its transistors are spent in L2-cache and the FPU is a tiny portion. Simply there are some circuits that translate into more transistors (or gates if you design in FPGAs). As I stated before, DSP instructions vastly simplify digital filtering implementation, so all the effects just need to use different filter tap coefficients in the same functional unit.
I think it has been told repeatedly that the cards (not the chips) limit the output to 24/96 in surround because it's the specification of DVD-Audio, until some media uses 24/192 on eight channels, the utility of such feature will become apparent.
You could argue that the C-Medida chip is a more capable chip but certainly not a more powerful since it doesn't even have an instruction set nor hardwired algorithms.
alg7_munif
Supreme [H]ardness
- Joined
- Oct 9, 2006
- Messages
- 5,862
HSE said:
How do you measure the computational power? If you read the source I gave, the percentage is not the the percentage of transistors but it is the percentage of MIPS(Million instructions per second) and it looks to me that the MIPS is used to measure the computational power but I might be wrong.
I would try to find out how many actual arithmetic operations per second those MIPS translate to, and if they are fixed point or floating point before making any claims, MIPS is also interpreted by the cynical as Meaningless Indicator of Processor Speed because not all instructions are created equal, it is not the same an "increment accumulator" than a "dot product" instruction over two FP vectors.
And do the math, a 32-tap filter for a mono stream sampled at 192KHz would only require about 6 million scalar additions per second and 6 million scalar multiplications per second, suddenly the 1100 MIPS for DSP that the article mentions look like overkill, even if the instructions are are scalar operations, as it would be capable of filtering around 90 streams at that sampling rate and 180 if the instructions are "multiply and accumulate".
And do the math, a 32-tap filter for a mono stream sampled at 192KHz would only require about 6 million scalar additions per second and 6 million scalar multiplications per second, suddenly the 1100 MIPS for DSP that the article mentions look like overkill, even if the instructions are are scalar operations, as it would be capable of filtering around 90 streams at that sampling rate and 180 if the instructions are "multiply and accumulate".
alg7_munif said:
simple
X-Fi has acceleration by DSP. The CMI8788 card has not acceleration.
Jep and its more cheaper without the DSP. Because the PC allways contains the CPU.
The problem is the inaccurately use of the word DSP.
This ist double wrong. (one bye your opinion and one bye extrmetech)
The X-Fi can handle all funktions up ti 192kHz 32Bit.
Yes the EMU20k1 can do SRC by himself, (with extrem high Precision)
The CMI8788 has no possibilities of a SRC. This architecture does SRC calc on on the CPU.
Nope is wrong. The extremetech aritcle appeared very early.
See in the ixbt articles. That article is qualitatively much better.
http://www.digit-life.com/articles2/multimedia/creative-x-fi.html
No. No.
The CMI8788 is an audio PCI controller, totally without possibilities of soundprocessing.
The EMU20k1 is a PCI controller and(!) a high performance soundprocessor.
Jep. This channels are in-/output channels. (left front, rear right. .. aux in) at fix frequencies. (44.1, 48, 96, 192kHz, with SRC by CPU).
Nope. All 10 channel are digital, or I²S (or AC'97) analogue. It means in- and output.
The EMU20k1 chip himself can handle 4096 channels.
look there (the Theatron DTS is a CMI8788 card too)
http://www.technic3d.com/?site=article&action=article&a=339&p=11
read the original description of the CMI8788. Look at processing, for OPCodes, for voice mixing. You will not find evreything of these.
http://www.cmedia.com.tw/files/doc/PCI/OxygenHD CMI8788 Datasheet Rev0.6.pdf
CMedia goes another way. They do soundprocessing on the CPU by the audiodriver, this is cheaper and the controller can be more (many more) simple.
alg7_munif
Supreme [H]ardness
- Joined
- Oct 9, 2006
- Messages
- 5,862
I think that the voice mixing is done by the CMI8788 chip, it also can handle up to 128 voices. On C-Media chips EAX effects are done by the CPU and this is shown on the website that you gave:
The EAX feature and DD/DTS encoding feature is a software feature if you read the the datasheet of the CMI8788 correctly. Not all games have/need EAX effects to sound "good", in games without EAX feature like oblivion I think that you will still get a hardware acceleration and if you use DD/DTS encoding, CPU will only be used for the encoding.
The EAX feature and DD/DTS encoding feature is a software feature if you read the the datasheet of the CMI8788 correctly. Not all games have/need EAX effects to sound "good", in games without EAX feature like oblivion I think that you will still get a hardware acceleration and if you use DD/DTS encoding, CPU will only be used for the encoding.
Can you prove this? From what I've read from the datasheet under the hardware feature:MixBar said:
alg7_munif said:
Nope. The CMI8788 don't can do a hardware mixing. Look at the CMI8788 Docs DirectSound3D is a software feature.
the 128 voices rendert completly in software. More voices use, more performance drop.
You must understand for what SCR is used.
A soundadapter works always with one fix frequency. (on CMI8788 there can either 44,1, 48, 98 or 192kHz). If the sound adapter now to play a sample differently to this output frequency, you need a SCR. For example: You play a 44,1kHz CD and your CMI8788 is in the 96kHz output mode. The driver resample the CD stream to 96kHz.
This can do, by the windows kernel mixer, by driver only, or by hardware. And the CMI8788 has no sound processor, that can do it.