Optimising PC for KS Special Mode

[manisandher]

Hi y'all,

I'm having the day off, after a few hard days working in Moscow last week. I've just finished some gardening in the glorious UK sun (no, really!), and have spent the last 30 minutes relaxing to some tunes through XXHE.

It occurred to me that I've never actually tried Special mode with my latest PC build, so I decided to give it a go. The first impression is that it's 'sweeter' than Adaptive mode. Adaptive sounds organic and full... if a little 'thick' in comparison.

In any event, I'd like to optimise my system to take full advantage of Special mode. I have been through all the threads on this forum with regards to optimisation ('Tweaking Dead', 'Power Setting', etc, etc) but still have an issue. Here are the lowest latency settings I can achieve:

Adaptive: buffer=32, Q1=1
Special: Q1=92

Any ideas as to how I could decrease the Q1 value in Special even further would be most welcome.

Mani.

[Flecko]

For adaptive mode not the smalest buffer size is important. You have to set the right size of your device. How you can determin it you can read in the box popping up if you hold the mousepointer over the ks-mode field.  As far as I understood you use the same settings for special mode. Sorry if you already know that...because you use 32 I just assumed that, it is unusual small.

[manisandher]

Hi Fleko,

Sorry, I forgot to mention that in Adaptive, I have Q1=1, buffer=32 (now corrected in initial post). Yes, I know this is low - I built the PC especially to get theses low latencies to work without any glitches. My RME card seems happiest when set to 32 samples.

But I'd like to get the Q1 buffer size down further in Special.

Mani.

[Marcin_gps]

Hi guys,

I just got Cantatis Overture sound card and I'm pretty shocked that I can't achieve decent buffer sizes with it. Actually, M-Audio  Transit (USB) that I've used so far could play at 64 samples without any glitches (24/44, Q1=1, Adaptive Mode) while Cantatis' lowest buffer size that I managed to squeeze is 512 samples (24/44, Q1=1, Adaptive Mode). I thought that it's a matter of performance, so I overclocked my CPU to 3.8GHz (4 cores, 4GB of DDRIII 2000MHz modules) and guess what - nothing has changed! That leads me to conclusion: drivers' "performance" is more important than overall performance of a PC But there is one thing that astonishes me - Leo reported that he could play at 4 samples (special mode) with the same sound card and slower PC - what the hell?

[Marcin_gps]

Hi Fleko,

Sorry, I forgot to mention that in Adaptive, I have Q1=1, buffer=32 (now corrected in initial post). Yes, I know this is low - I built the PC especially to get theses low latencies to work without any glitches. My RME card seems happiest when set to 32 samples.

But I'd like to get the Q1 buffer size down further in Special.

Mani.

Mani, which PCI slot do you use for RME?

[manisandher]

Marcin,

The Gigabyte P55M-UD4 mobo only has one PCI slot (it has 3 PCIe slots, one of which I'm using for the video card).

I was tempted to buy the PCIe version of my card, but have decided that the money would be better spent contributing towards the two NOS1 DACs that I want.

I if had a PCIe card, would choosing one slot over another make a difference?

Mani.

[AUDIODIDAKT]

Hi,

There can be audible differences in the PCI slot you use.
The rule is: use PCI(e) slot nearest to processor for audio. (shorter lines)
But for me this is not thru (because its a "bad" slot, dont know why but it just sounds worse)
Do experiment with this, you can also do some shielding of your soundcard (or maybe it is already within a metal casing) especially when sitting next to videocard.


Cheers,

[AUDIODIDAKT]

Mani,

Are you planning to use 2x NOS1 in one system ?
As in bi- or quad amp control ?

Would be nice ? (all on 1 master-clock)

[manisandher]

Hi Roy,

No, actually I want to use them in two separate systems - office and listening/living room. In my 'old' days, I might well have been tempted to go for multiple DACs for my multiamp system, but those days are gone (for now at least). I'm currently using a pair of crossoverless Quad 2805 speakers - just a single NOS1 and stereo amp required.

FWIW, I really like the Quads - it's so good to have totally coherent sound. My one criticism had been that they sound a bit 'dead' - what I call 'cardboard' sound. But with KS Special mode, they have come alive in an amazing way - I'm getting seriously good sound out of them now (powered by an old Pass Labs Aleph 4 single-ended amp that I've had knocking around for years).

Mani.

[PeterSt]

Wait for 0.9z. You will be amazed.
I hope.

Marcin, yes, it is the driver which is far more important. Don't say I didn't tell that yet 


I myself keep on falling back to Adaptive Mode. It is very strange; the other day I talked about some bootleg recording and thought it was the best recording ever. Two days ago the same happened, this time some Made in Japan bootleg. Totally wrong for mike placements, but in the mean time the most fat (hard rock) sound you can imagine, and like you're in the middle of it. Played the loudest ever for 1 hour in a row, and my wife didn't make a single complaint (always the best measure). But then we went to a Deep Purple concert some years back, and that helps of course.

Anyway, after a while (I mean weeks) listening to Special Mode with me it is always so that going back to Adaptive Mode brings the real sensation. Indeed it is warmer, fuller, and yes, organic would be a correct expression. But (or and) ...
I applied a special "trick" to the DAC (hardly feaseable for real production) and the bass reproduction comes with one word only : wow. This excels in Special Mode for its preciseness, but the fun is that with Adaptive Mode it just can have it. So, because it is more precise, the organic glow Adaptive throws over it just doesn't make it smeared. This is where "fat" indeed becomes fat (I must be careful because it may be a dutch expression "fat"), and doesn't turn up into a mush.

A last thing which is also funny (or interesting) I think : with the ultra low latency KS mode (remember, I can use 4 samples @ 24/358.2 now) the drums start sounding towards ... paper ! Cardboard if you want. But my speakers aren't cardboard like;
You may recall the (or mine at least) expressions about Arc Prediction and how it makes everything more "dry". So, shorter, giving way to the next sound coming. Well, the first thing I noticed from the 4 samples latency is a very clear "more short" again. But also too short ! So I went to 16 samples and later to 64, and although it improved, I was struggeling with a "what happened to my best sound ever ?". ... Adaptive Mode brought that back.

Allright, as said, wait for 0.9z. But be ready for a totally different sound from Special Mode (which currently is no option -> back to the old sound). Many protests may hunt me shortly ... haha

Peter

[Marcin_gps]

For me the lower the buffer is set, the sharper the sound becomes and this is the same regardless of sound card/interface used.

I am trying to find the best playback options for Cantatis Overture (great piece of hardware, I could throw my Beresford through the window right away... :D). QAP has more air, better holography, overall it's lighter while 24/44 has more punch. That would confirm Peter's conclusion about applying upsamling on a oversampling DAC. It's worth of trying though, I bet few folks would prefer QAP over redbook, even without NOS DAC.

For those who're trying to pursuit the lowest latency "train", you might never catch it :D yesterday I performed a series of actions to achieve 256 samples, well, I failed.
Here's what I've done so far:
1. Assigned IRQ priority to sound card
2. Switched power plan to performance
3. Forced installation of VIA drivers for Cantatis
4. Removed GPU drivers and set hardware acceleration to none
5. Changed PCI slot
6. Overclocked from 1GHz to almost 4GHz (quad core)
7. Installed fresh Windows on a formatted disk

Nothing improved - null! My theory is that it also depends on chipset and southbridge implementation. I can't adjust PCI Latency Timer as well. If my theory is right I will be able to achieve lower latencies with other PC and other chipset. I'll try soon ad let you know how it worked.

[PeterSt]

Haha, once again : wait for 0.9z. Hopefully today. Or otherwise tomorrow.

What you overlook is what I by now can do with the driver. 0.9z will hopefully proove that. To you too. And then I am still not finished with it.

With my 24/352.8 and 0.9y-8c there was no way I could use Special Mode with any setting I could find. It played allright, but with a lot of glitches and ticks. Now I can use 4 samples and no glitch to be heard. I didn't touch the PC whatsoever, or IOW, it is not about that. It is about the driver allright, and it is there where I changed things. You will see ...



Whether you will like the sound is another matter. It is special, and again new. With again more mid detail. It might work for you all, but I don't think it does with me. So, to have better conclusions I really wonder what you (all) will say about it.

Peter

[Marcin_gps]

Great, so we're all looking forward to that new great special mode

[Marcin_gps]

Haha, I switched from 24/44 to 16/44 and now I am able to set 128 samples That should be obvious - less data, lower latency, right?
But it sounds so much worse on a 24bit DAC. Also there is a noticeable difference between DAC needs 24bits and DAC needs 32bit settings. How is that possible? oO

[PeterSt]

Well, assuming that your soundcard can take both 24 and 32 bits (which is not common at all), when set to 24 bits you're using less bandwidth (because I'm sending 24 bits instead of 32 bit in that case).

But it sounds so much worse on a 24bit DAC.

I don't understand ?