Playing Music in Vista Drags Down Network Performance

[Lemur]

Not a good weekend for Windows Vista, I'm afraid. Turns out that sending audio signals to your sound card (or motherboard sound output) slows down Vista's network performance.

Here is the thread where the discovery was thrashed out. Naturally, Slashdot jumped into the fray. Microsoft's not-entirely-official response can be read here.

A lot of current speculation hinges on the idea that music playback triggers certain DRM alarms in the OS, which run online to do ... something. Others have evidence that it's strictly a call to the sound driver that slashes network throughput by up to 90%. Ouch.

From Microsoft:

• "We have been looking into this problem and are working on a doc that will go into the technical details of what we have found.”
• “Please note that some of what we are seeing is expected behavior, and some of it is not. In certain circumstances Windows Vista will trade off network performance in order to improve multimedia playback. This is by design.”
• “The connection between media playback and networking is not immediately obvious. But as you know, the drivers involved in both activities run at extremely high priority. As a result, the network driver can cause media playback to degrade. This shows up to the user as things like popping and crackling during audio playback. Users generally hate this, hence the trade off.”
• “In most cases the user does not notice the impact of this as the decrease in network performance is slight. Of course some users, especially ones on Gigabit based networks, are seeing a much greater decrease than is expected and that is clearly a problem that we need to address.”

[Xiahou]

So, who wants Linux?
My current plan is to ride XP til the wheels fall off, and then make the jump to Linux or some such. Linux already makes a great home server, and it seems to run a decent amount of games, I'm just not comfortable running it as a primary desktop OS yet.

It'll be interesting to see what the "experts" figure out as the cause of this newest Vista problem though.

[Blodrast]

[*]“Please note that some of what we are seeing is expected behavior, and some of it is not. In certain circumstances Windows Vista will trade off network performance in order to improve multimedia playback. This is by design.”[*]“The connection between media playback and networking is not immediately obvious. But as you know, the drivers involved in both activities run at extremely high priority. As a result, the network driver can cause media playback to degrade. This shows up to the user as things like popping and crackling during audio playback. Users generally hate this, hence the trade off.”[*]“In most cases the user does not notice the impact of this as the decrease in network performance is slight. Of course some users, especially ones on Gigabit based networks, are seeing a much greater decrease than is expected and that is clearly a problem that we need to address.”[/LIST]

ROFL at first two bullets ! Damn, I really need to take OS courses again - and I'll prolly fail them, too, if they are anything like the theory that M$ seems to implement... "trade off network performance to improve multimedia playback is by design" ? sheesh...
I wonder how all the other OSes handle such incomprehensible tasks as running the network and some multimedia application at the same time...


As for the last bullet, I guess they designed that "feature" with only dial-up in mind ? Gigabit ethernet was unheard of a few years ago ?

And one more thing: none of this, not one little bit, has anything to do with DRM and/or phoning home ? Well, call me paranoid... (although it's not like they are famous for being honest and transparent...)

[Husar]

And here comes Husar to the defense of the big evil company again.

Well, my network performance in XP drops as well when I listen to music since I listen to webradio....

Ok, not much to say except that my downloads are still going fast, but then I just have a 6MBit connection on a 100MBit network.
Did you also know that vista is only optimised for up to 4 CPU cores and will not be able to use 8 cores to full extent? and did you also know that most software is still 32bit and will thus often run slower than necessary on 64 bit systems. The world is really unfair, also note that 32bit systems seem to cut off at 3GB RAM, Vista 32bit apparently recognises 4GB as 3, or so some guy in the WiC forums said.

But then it's still a lot faster to install Windows than to install Linux so I'm rather unlikely to switch, don't want to learn thousands of written commands to use my OS.

And don't forget about DX10.1, they really want to make us new graphicscards again, not to mention PCIe 2.0 wgich is designed to give more energy to graphics cards and thus has a direct impact on my electricity bill which I'm not very fond of as a student. But what can you do, stop gaming and learn all day long? Good idea, but I wouldn't learn anyway.


Now seriously, who is really affected by that? Maybe LAN parties? I see this as a bug, yes, but a rather minor one, especially compared to the registration bug that probably makes your network completely useless or something.

[Blodrast]

Well, my network performance in XP drops as well when I listen to music since I listen to webradio....

Well, duh - it's webradio, so of course that one thing using the network will use up bandwidth from the other application's bandwidth... but the complaint was about multimedia and network, not about network and network.

Did you also know that vista is only optimised for up to 4 CPU cores and will not be able to use 8 cores to full extent?

Not sure how that's relevant... I don't recall any criticisms regarding this issue, so I'm not sure why you're bringing it up at all.

Oh, and I am not sure I agree with your position here - trying to "defend the big evil company". The fact that some parts of their products are ok does not mean there are no parts that are not ok. Your dismissal of (inexistent) criticisms doesn't invalidate the very real criticisms of very real problems in their stuff.

That said, I'll address your other issues, even though, again, I'm not sure how they are relevant to the topic.

and did you also know that most software is still 32bit and will thus often run slower than necessary on 64 bit systems.

Then run it on 32 bit systems!
Again, I realise you're being tongue-in-cheek (at least I hope you are), but I'm not sure what point you're trying to make (unless, like I said above, you're trying to invalidate all criticisms by showing some obvious non-issues).

The world is really unfair, also note that 32bit systems seem to cut off at 3GB RAM, Vista 32bit apparently recognises 4GB as 3, or so some guy in the WiC forums said.

I don't know what Vista does, but the first part is incorrect. First of all, OSes do recognize 4GB as 4GB. They can also support a lot more memory - have a look at any of the "-bigmem" linux kernels, guess what the "bigmem" allows them to do ? Here's the output from top on a machine I'm connected to, I bolded out the relevant part.

top - 16:10:00 up 122 days,  3:22,  1 user,  load average: 0.00, 0.00, 0.00
Tasks:  75 total,   1 running,  74 sleeping,   0 stopped,   0 zombie
Cpu(s):  0.0%us,  0.2%sy,  0.0%ni, 99.8%id,  0.0%wa,  0.0%hi,  0.0%si,  0.0%st
Mem:  16377332k total,   682624k used, 15694708k free,   278956k buffers
Swap:  2097132k total,        0k used,  2097132k free,    47920k cached

The part about 4GB vs 3GB _may be_ related to the following issue (Note: this is the case for linux, I don't know about windows, but my guess is that it is a similar issue): on 32 bit systems, there is a limit of 4GB for _one process_. One process cannot access more than 4GB of memory - theoretically. The practical limit, however, is about 3GB for linux, 1Gb out of those 4GB being reserved for the kernel. This limit can be raised by hacking the kernel, up to 3.7GB, from what I've read.

But then it's still a lot faster to install Windows than to install Linux so I'm rather unlikely to switch, don't want to learn thousands of written commands to use my OS.

I'm not at all sure about the installation part, to be honest. If you don't run in any problems with drivers (which is certainly an issue for linux, although sometimes can be resolved relatively easily), it may even be shorter, depending on what distro you use.

Hey, I'm not trying to convert anybody to use linux (although I'd rather people made an informed choice), but you do know you're exaggerating about the "thousands of commands" part, right ? You can do most of your tasks in the GUI that all linux distros offer these days - this has been the case for a few years now...

[Husar]

Well, duh - it's webradio, so of course that one thing using the network will use up bandwidth from the other application's bandwidth... but the complaint was about multimedia and network, not about network and network.

I know, it was some sort of bad joke, I'm sorry you can't expect much better from me.

That said, I'll address your other issues, even though, again, I'm not sure how they are relevant to the topic.

They're actually things I read about and which concern me to some degree. I wonder why Vista runs fine only on quad cores when it's very likely that we will see eight cores in the (near) future. also makes me wonder at how many cores XP will stop getting optimal performance out of them. It was just relevant to show that I have my criticisms as well before anyone thinks that I'm drroling everytime I see a windows logo.

Then run it on 32 bit systems!
Again, I realise you're being tongue-in-cheek (at least I hope you are), but I'm not sure what point you're trying to make (unless, like I said above, you're trying to invalidate all criticisms by showing some obvious non-issues).

I brought that one up before and I meant that if the programs were programmed in 64 bit, they might be faster. Supreme Commander for example lags a lot even on my dual core, maybe if I were lucky, it wouldn't lag as much if it were programmed in 64bit, but then I'm not an expert on these matters. If a 64bit mode could be made optional like DX10 or so, or maybe choose a version upon installation, that might help with some games.

I don't know what Vista does, but the first part is incorrect. First of all, OSes do recognize 4GB as 4GB. They can also support a lot more memory - have a look at any of the "-bigmem" linux kernels, guess what the "bigmem" allows them to do ? Here's the output from top on a machine I'm connected to, I bolded out the relevant part.

Well, I just repeated what some random guy on another forum said, apparently Windows in 32bit only supports 3GB RAM. I read about something like that from a more knowledgeable source before but don't remember what was said. for all I know that guy could just have a malfunctioning RAM stick.

The part about 4GB vs 3GB _may be_ related to the following issue (Note: this is the case for linux, I don't know about windows, but my guess is that it is a similar issue): on 32 bit systems, there is a limit of 4GB for _one process_. One process cannot access more than 4GB of memory - theoretically. The practical limit, however, is about 3GB for linux, 1Gb out of those 4GB being reserved for the kernel. This limit can be raised by hacking the kernel, up to 3.7GB, from what I've read.

According to memtest, the limit in XP 32bit is somewhere around 8xxMB, the program told me to start another memtest to test the other parts of my RAM. Made me wonder how games use more RAM then, but I never bothered to find out.

I'm not a Linux expert and currently I'm too lazy to become one since my Vista is still running fine.

[Blodrast]

I know, it was some sort of bad joke, I'm sorry you can't expect much better from me.

Nah, it's okay, if you were kidding it's all right, except that as you know sometimes it's kinda hard to tell when someone is serious and when they're joking on these internets.
It's all good though.

They're actually things I read about and which concern me to some degree. I wonder why Vista runs fine only on quad cores when it's very likely that we will see eight cores in the (near) future. also makes me wonder at how many cores XP will stop getting optimal performance out of them. It was just relevant to show that I have my criticisms as well before anyone thinks that I'm drroling everytime I see a windows logo.

Don't drool! The ladies don't like it!
I think the main reason is that it is very hard to optimize something for an architecture that doesn't exist yet.
Yes, people use simulators, but simulators, as their name suggests, are just simulators - and they are not always realistic enough, or the final product may be different from whatever flavour of simulator people were using at a given point in time.
Take for example the SMT technology (Intel's Hyperthreading), which was much of a flop, imo (do you hear anything about Hyperthreading from Intel these days ?). They played with the architecture on simulators quite a bit before making the product, and it looked good on paper... Then they built it... and it was far from getting 30% more performance as marketing figures claimed.

That's what I think is the main reason...

Another reason for that is another pretty hot topic in the research community - how do we manage all these cores ? How do we write software for them ? Do we need new programming languages ? New parallel programming paradigms ? How do we easily make old software run better (i.e., take advantage of the extra resources) on them ?
Lots and lots of problems, and no definite answers yet.

Hope this sheds a bit more light into your questions.

I brought that one up before and I meant that if the programs were programmed in 64 bit, they might be faster. Supreme Commander for example lags a lot even on my dual core, maybe if I were lucky, it wouldn't lag as much if it were programmed in 64bit, but then I'm not an expert on these matters. If a 64bit mode could be made optional like DX10 or so, or maybe choose a version upon installation, that might help with some games.

I don't think 64bit has anything to do with running faster, it's just addressing some address (pardon the pun) limitations, mainly, afaik.

[Alexander the Pretty Good]

Maybe the thread title should be just "Vista Drags Down Network Performance."

What an awful OS.

[Lemur]

FWIW, there's finally a decent technical explanation of what's going on.

[Xiahou]

FWIW, there's finally a decent technical explanation of what's going on.

Wow... so basically they're saying that even with the quantum leaps forward in hardware over recent years, Microsoft couldn't manage to code an OS that can handle LAN traffic and playing a song at the same time without bogging down the CPU? That's crazy.

Really, how bloated and inefficient can you get?

[Lemur]

My favorite line from the article: At this point, it becomes clear that the process scheduler folks and the networking folks are bitter enemies and do not converse.

-edit-

Everybody knows that starting with Windows NT, Microsoft copied the networking stack from BSD Unix. Bad form, poor sportsmanship and all that, but it meant that the networking was pretty stable. I remember reading that Microsoft was throwing out the entire networking stack and re-writing it from scratch for Vista. And I remember thinking, "Hmm, that could go badly."

At least with a pirated clone of Unix, it worked as advertised. Reading about the shivs and hacks needed to square the circle in Vista makes my eyeballs ache.

[Husar]

Well, I always say, better a good copy than a bad new invention. But people always complain about good copies for some reason.
The Chinese government runs a whole country based on good copies.

[Bijo]

[offtopic]And yet again Vista shows errors in its system. Why would one utilize this operating system in the first place (unless necessitated)? DX10 (while DX9 still performs quite well)? The trouble Vista brings is sufficient reason to await a superior version, even if its arrival would take many a moon.[/offtopic]