Comparing an LSM to a Desktop Computer |
Post Reply | Page 12> |
Author | ||
Elliot Thompson
Old Croc Joined: 02 April 2004 Location: United States Status: Offline Points: 5176 |
Post Options
Thanks(0)
Posted: 20 May 2015 at 11:35am |
|
https://www.youtube.com/watch?v=TMBMZ_XX6g0&feature=youtu.be Here is a quick comparison of an LSM to a Desktop Computer used as a VST Host. Both devices offer a +12 dB boost at 7.850 kHz with a Q of 1.0 The LSM is 48 kHz/24-bit whereas the Computer is configured to 96 kHz/24-bit.
The Computer is an old single-core 3 GHz operating on Windows XP SP-2. The CPU consumption throughout the test was 2%. The soundcard streaming the signal is an old M-Audio Audiophile 2496. Latency is 10.7 ms under a 1024 buffer setting at a sample rate of 96 kHz.
The second computer conducting the recording was configured to record at 96 kHz sample rate.
The Latency differential amongst LSM and the Desktop Computer offers a Flange Effect which is too low to make a significant difference. With the CPU hovering at a mere 2%, the buffer could be reduced by half and still would not bottleneck the CPU if it really mattered.
The track used is an old Madonna vinyl track from 1983.
Best Regards, |
||
Elliot Thompson
|
||
Darkstar
Registered User Joined: 08 October 2014 Location: Italy Status: Offline Points: 326 |
Post Options
Thanks(0)
|
|
Does this mean we'd better be off using FabFilter Pro-Q over a laptop and audio card rather than Eq-ing through LMS?
|
||
Bass =/= Enough
|
||
Elliot Thompson
Old Croc Joined: 02 April 2004 Location: United States Status: Offline Points: 5176 |
Post Options
Thanks(0)
|
|
I would imagine it is a matter of flexibility, convenience in addition to preference on which direction one would choose. Best Regards, |
||
Elliot Thompson
|
||
toastyghost
The 10,000 Points Club Joined: 09 January 2007 Location: Manchester Status: Offline Points: 10919 |
Post Options
Thanks(0)
|
|
Over 10ms is a lot of latency for just a basic EQ. Why did you not make them both the same sample rate? Also what is the latency of the dedicated processor performing the same task?
|
||
Teunos
Old Croc Joined: 23 November 2008 Location: The Netherlands Status: Offline Points: 1799 |
Post Options
Thanks(0)
|
|
My ev netmax has a fixed latency of 2.29ms fram analog input to analog output regardless of what processing it does for as long as of course no fir filters are introduced. Quite a difference from 10ms. Also it is stand alone, flexible and does not require a computer.
We have had this discussion in another thread, but to me 10ms is unacceptable. |
||
Best regards,
Teun. |
||
shagnasty
Old Croc Joined: 30 July 2007 Location: Guildford, UK Status: Offline Points: 7685 |
Post Options
Thanks(0)
|
|
KT quote a max latency for end to end live audio usage, I forget what the figure is, but it isn't near 10mS, 4.5 springs to mind as the point as which IEM systems become useless, hence Hyper/SuperMAC run L2 on Ethernet to avoid the "packet wrapping" stage an IP transport would need, from memory even Dante is very close to the limit (a router hop would trash it) so I am not surprised at this result.
PS does anymore know what LSM is a TLA for (I assume not Loud Stage Manager!!!!) |
||
Elliot Thompson
Old Croc Joined: 02 April 2004 Location: United States Status: Offline Points: 5176 |
Post Options
Thanks(0)
|
|
Possibly you missed this part
That means I could have reduced the buffer to 512 which, would have raised the CPU to a mere 4% if I wanted reduce the latency. Or I could have dropped the sample rate to 48 kHz and reduce the buffer rating to 256 which would have gave me a CPU consumption of 4% if I wanted to reduce the latency. With computers the CPU consumption is the deciding factor and, when your CPU is hovering below 10% you have lots of headroom to adjust the latency on how you see fit. The majority of the workload a computer sees does not come from audio processing but visual graphics. Personally I do not use audio cards that have the capability of delivering high sample rates at low sample rates. It defeats the purpose in investing in a high sample rate sound card. The M-Audio is configured to 96 kHz at all times, so that is how the host was used. VST Plug-ins varies per developer and, adding more bands will not always bring forth a huge spike in CPU consumption. Equalisation does not require a lot of processing unless the developer is adding additional processing such as analogue emulation. For straight forward digital audio processing, the consumption will always be low. Best Regards, |
||
Elliot Thompson
|
||
toastyghost
The 10,000 Points Club Joined: 09 January 2007 Location: Manchester Status: Offline Points: 10919 |
Post Options
Thanks(0)
|
|
So why not do the test with less latency if it is so easy and possible? Your comparison as it stands is not really a fair one.
Also your load is currently an EQ, which is hardly representative of the entire processing a dedicated unit will do in a gig situation. |
||
shagnasty
Old Croc Joined: 30 July 2007 Location: Guildford, UK Status: Offline Points: 7685 |
Post Options
Thanks(0)
|
|
Afaia most DSP solutions don't increase latency with workload as they are FPGA based and that ultimately makes them inherently timely, CPU based solutions will increase in latency, but not that linearly, with only a LMS style system running you could possibly not load the machine to the point it increases at all, that said, add limiters which people on here seem to dribble a lot about, some dymanic EQ and you could be in trouble, I note hardware accelerated cards feature hard in Pro DAW systems and some digital desks even offer VST off-load racks, which I will wager do not contain i7/Xeons/Piledrivers but something a bit more defined in use...
|
||
Elliot Thompson
Old Croc Joined: 02 April 2004 Location: United States Status: Offline Points: 5176 |
Post Options
Thanks(0)
|
|
I don't believe you are familiar with how low of a percent the computer is operating. Considering that it is a single-core can you imagine where you would stand with an 8 Core CPU with someone who knows how to configure it solely for audio purposes? They have recording guys that are tracking with multiple tracks ranging from 16 - 32 channels with effects on each channel without a single glitch on their computer. We use eight channels at best. So I dropped the buffer rate to 124 in addition to reducing the Sample rate to 48 kHz. This gives me a latency of 2.7ms
This poses no problem for the Desktop
Computer. As you can see in the Task Manager, the average is 2% with peaks at
4% Since Youtube Blocked it, you can watch it on Google. https://drive.google.com/file/d/0BysEZgz2cg_yNHBGS3hyMmtzSGc/view
The Desktop is playing on the Left Channel
while the LSM is playing on the Right Channel. I merged both channels together
which offers a flange affect. If I wanted to drop the Buffer rate to 64, I am
more than certain the computer would go into cardiac arrest but, as far as I am
concerned it is not necessary. It appears many are focusing solely on latency and missed the reason I made the Topic. The topic was for you to listen to the difference in sound both components offer. For neither of them sound a like. So lets focus on the sound difference amongst both devices. Best Regards, Edited by Elliot Thompson - 20 May 2015 at 4:42pm |
||
Elliot Thompson
|
||
azlan
Registered User Joined: 09 January 2012 Location: W12 Status: Offline Points: 364 |
Post Options
Thanks(0)
|
|
I hate to say it, but I'm not sure what the point of this is? If you are advocating using a PC insead of an LMS, then latency is absolutely critical, especially as a typical modern digital mixing signal chain will already have a not insignificant amount of latency, meaning that the pc really needs to keep the latency as low as possible.
Products like waves multirack already prove that low latency io from a computer can be done reasonably reliably, and with good results. However to really push the usefulness of such an idea, the real genius would be to reduce the signal chain to a single a/d (or none in the case of a dj) at the preamp, and keep everything digital all the way to the amps, ideally using the smallest number of format/transport conversions as possible. If I remember right, using DVS (dante virtual soundcard), you can get achive a looped through latency of <5ms (the same as being about 5ft from the sound source), using a combination of this and some decent spec routers, and something like the midas network bridge to get aes to feed the amps, you could achieve a decently stable,low latency system, whilst avoiding multiple conversions (which would ultimately add further latency and lose quality), this would allow you to use the host PC as an LMS, and if you had enough processing power left over,use it as a playback system, digital inserts, or just about anything else you would want. As an aside, if I recall correctly, it should be possible to lower the analogue latency by actually increasing rather than decreasing the sample rate. |
||
Darkstar
Registered User Joined: 08 October 2014 Location: Italy Status: Offline Points: 326 |
Post Options
Thanks(0)
|
|
Ok... Another question then: does the fact that FabFilter provides "zero latency", natural phase and linear phase make it a better option?
What they claim is that in natural phase mode the eq acts about as precise as analog eqs, whilst classic digital eq-ing gives bad phase alteration. Edited by Darkstar - 20 May 2015 at 8:52pm |
||
Bass =/= Enough
|
||
Post Reply | Page 12> |
Tweet |
Forum Jump | Forum Permissions You cannot post new topics in this forum You cannot reply to topics in this forum You cannot delete your posts in this forum You cannot edit your posts in this forum You cannot create polls in this forum You cannot vote in polls in this forum |