Windows7 ? hmm ... maybe not !

[PeterSt]

Standing by to trouble shoot here 

I guess I've already done this for you. So, your only hope is when the "Speakers" (= analogue) selection can be routed to the SPDIF output, because the "Speakers" work with KS. Maybe this can be done (with that Wire application from the Juli@) and *that* I didn't try.

The Juli@ is strange anyway;
I don't know about your version, but mine (under Vista as well as under W7) only shows the Speakers and SPDIF devices, while the manual talks about Ch34 and Ch1234 as well. No way they show up. But :
Internally they sure are there and I can let them work by a VERY strange means. I have a hunch that in XP this all shows normal (and a total of 4 devices I think). This still does not imply KS will work on XP with this card. But if you are anxious to know : plug it in one of your XP machines, drop the Foobar KS dll in the components folder, and see whether it shows up under a "KS:" device. If it does, it works.

Peter

[PeterSt]

I can't find back where I wrote about it, but one of the first things I noticed under W7 was that I was "missing seconds". Looks stupid of course and I couldn't even be 100% sure, but this was about Engine#4 which worked to some extend before I left for a vacation, and after coming back installing W7, and then the OSD Time wouldn't match reality anymore.
And now think of it, each and every day I tried to tweak it for the better (which would be some 26 days now) and only yesterday I succeeded. Btw, this is always a matter of "play time !" when this can be judged, so the time can be spent somewhat more efficiently (listen in the mean time). Also, it always takes several tracks to find out it's wrong again.

26 days.

I think I just found out what's happening here.

I received a log file from Per, and it occurred to me that in his situation per one second the exact number of bytes are being stuffed to wherever they need to go. And, since this is a recognizeable number, it just could occur to me that this "can" happen. So, it also only now occurred to me what is going wrong in my own system. Here it is :

Each one second (as per the internal clock) my system stuffs 1.4 % too many samples to their destination. That is 1.4 seconds per 100 seconds or 1 second per 71 seconds. This must be read the other way around : my time is running too slow, or IOW when it would run ~1.4% slower it would show the proper amount of stuffed samples.
This is how it functionally works out. But it is not how it technically is ...

Because the time also shows in the log file, I can see that my clock is allright, because after each one second of "a pass" just a little over one second is added. Say, 1.4% more. Per's system. shows exactly one second per a one second pass forever. To get the idea :

02:40:24.6081590
02:40:25.6081590
02:40:26.6081590
02:40:27.6081590
02:40:28.6081590
02:40:29.6081590
02:40:30.6081590
02:40:31.6081590
02:40:32.6081590
02:40:33.6081590
02:40:34.6081590

Forver and ever and ever. But this is my system :

09:51:08.3723166
09:51:09.3863183
09:51:10.4003201
09:51:11.4143219
09:51:12.4283237
09:51:13.4423255
09:51:14.4563272
09:51:15.4703290

also forever and ever.


Now, I just prove that this is not about a "sleep timer" being wrong or off. Thus, when I "sleep" 1 second, this shows a nice equal 1 second offset each time, up to the resolution (decimals) you see above. This means that the small code which is executed in between "sleeps" takes 1.4% time of a second or 0.014 seconds per second. This, while in Per's system this takes virtually zero time. Oh, it will take time, but it needs way more decimals to show it. Anyway, in Per's system over 2 minutes time this doesn't change up to the 7th decimal while in my system almost 2 seconds have "disappeared" over that time period.

That I can't depend on the "timing" is a nasty thing, but actually something to take into account always. However, once things have been setup consistently, it shouldn't change because of a new OS. Here it does. But, this is still not the most worst thing. This is though :

While this whole (SQ) audio thing depends on the most fragile stupid things in code (beyond imagination and beyond known science), here W7 is doing something which Vista does not. Actually it consumes cpu or waits for something, and I don't know what it is.

Ok, hold tight :

I just tested, and when I (just an example) had to determine the current time and all the formatting going along with that to show it properly, I can do that 6000 times to reach the equivalence of the time I loose in the process.
Ah, that didn't tell you much. Ok;
When I do nothing but performing a line of code, very similar to process an audio sample, I can do this ... 35,000,000 (that reads 35 million) times before I reach the equivalence of the time I loose in the process. That's near to processing a complete track !!  and can be performed in the 0.02 seconds I have for it.

Well, I hope you can imagine now that I feel this is totally out of control, and (something very) wrong in the mean time.
I covered for this "loosing seconds" now, but it took me 26 days throughput time, but only for Unttended. Attended will have similar issues which is worse and more dependend on the proper timing (because XXHighEnd runs totally independend of XXEngine3).

So, now you know. If you don't "hear" this, good. But, Bweh.
Peter

[manisandher]

Peter, what are, or will be, the implications of this (if any) for the rest of us?

Mani.

[PeterSt]

Good question Mani. But I can imagine not even everybody suffers from this issue. I could try to find everybody's (coincidentally posted) log files though, and look. Or, by the time you are all more up and running I can ask you to just post them (X3PB). Btw, I can only "see" this with Engine#4, because of the way it is setup.

I can also imagine that one day I'll find out what actually causes this. But as I feel it from the start this is about task switching and then possibly nothing can be done about it (the before figures from my PC was without any audio running/playing). It will need time to spend on it, and right now I'm doing other things first.

Fact is that it is quite unworkable (for me, this project), although most of it has been dealt with (ok, for my device's situation, the others (like yours) yet to come). But as we know (or as I assume it) it also has implications for Engine#3, because remember, all those suddenly having (audio stalling) problems.

Lastly, but that may not interest you much, I have KS running now without issues (Unattended only) and most of the things, whether it were gaps, preeps or OSD issues, have been solved in a fashion "huh ? allright good, but why ?". It is nothing to like, and nothing with a spur of decency.

But I don't give up fast.
Peter

[PeterSt]

Hello there,

Because this topic exists as it is, it seemed a good idea to tell here, in this place, that currently I don't feel that I have any SQ problems with W7 and Engine#3 anymore. This was discussed already here : Re: Engine#3 Vs Engine#4 comparison up to a few posts under that.

Key seems to be to use a high latency (large buffer) setting in the soundcard/driver. Or at least this helps me / my system.
I now use 2048 (samples) instead of the opposite - 48 which worked fine for Vista (at Quad A.P. upsampling = 176.4 KHz and 32 bit output).
I can't tell (yet) whether this helps everyone, but it could be an explanation to people not reporting SQ problems under W7 at all (those people possibly using a high soundcard/driver latency or not being able to set it, which will imply a high latency to begin with).

I was the one who started the stir about this; hopefully this stirs it down a bit.

If you can't agree with this (W7 still sounds like sh*t while Vista is very ok), please let it know !
Thank you,
Peter

[Robin Hood]

In the meantime, those of us who have been around for longer can perhaps more readily distinguish the differences between Vista and Win7, which should make it easier for novices to chose which O/S to ultimately end up with.

Okay, are there any final conclusions and recommendations as to the choice of O/S that provides better SQ?

[Marcin_gps]

For me and my soundcard, Win7 wins, but I blame m-audio drivers. I couldn't get decent latencies on Vista and XP, even highly tuned (nLite, vLite). On the other hand, I find the sound of "Seven" less digital/harsh, but as I said - could be specific on my system.