Intel should add a fifth core, call it "MediaCore" technology.

[XavierMace]

But you don't buy quad cores so what difference would it make if they added a fifth?

 

[Insert_Nickname]

You're seeding a bunch of torrents on a two thread system, while trying to listening to streaming audio at the same time... no shit, the scheduler is going to struggle to fit all of those tasks into time slices smoothly without something losing out. Two threads is still two threads. Start messing with thread affinities, or just buy a good CPU already.

If Larry is insisting on doing all those things at once on a single system, he should just add two more cores to his CPU... hint... hint...

Even better, get one of the 35W i7's with HT (4765T/4785T). More performance while consuming less power seems like a win-win to me.

 

[justanotherboy]

I think your problem is more SW related, flash is really really bad; try to use something else to listen to the radio. Even my Linksys E4200 can handle 5 MB/s of torrent data (the router is running transmission) while handle all the network traffic, and that thing has a ~500 MHz MIPS CPU.

 

[Arachnotronic]

You're seeding a bunch of torrents on a two thread system, while trying to listening to streaming audio at the same time... no shit, the scheduler is going to struggle to fit all of those tasks into time slices smoothly without something losing out. Two threads is still two threads. Start messing with thread affinities, or just buy a good CPU already.

:thumbsup:

 

[sm625]

It's more a driver issue than anything. During music playback on an otherwise idle machine, the cpu will sleep for long periods of time. Then the audio buffer gets low and this causes an interrupt. The cpu has to wake up and go put more data into the buffer. If the driver isnt coded right, the buffer could run out before the cpu is able to refill it. There is nothing wrong with intel's big cores being used for this type of application. They function very well for this type of application because they can wake and do their work very fast and then go right back to sleep. But if the driver isnt coded well then it all goes to crap.

Rather than having a "media core" I would rather see a simple atom core which can be used during low-to-idle periods. That would be more general purpose than something media specific. But it really doesnt matter too much since the big cores can sleep and wake very fast.

 

[Rakehellion]

I suggest a dedicated media-playback core.

This already exists. It's called the GPU.

 

[Flapdrol1337]

whoop whoop whoop whoop

it's the sound of the quad core police

it's a network issue.

 

[BSim500]

it's a network issue.

Indeed. I have an i3 in my HTPC. Audio playback takes up all of 1% CPU usage. You could stick 32x cores in there and still suffer from cr*ppy web browser flash plugin's many issues or transport glitches in general. Pumping the Internet connection full of torrents is probably causing a latency bottleneck in the net connection itself. Same reason people don't download 12x large files whilst multiplayer gaming then complain of elevated ping times.

 

[biostud]

Indeed. I have an i3 in my HTPC. Audio playback takes up all of 1% CPU usage. You could stick 32x cores in there and still suffer from cr*ppy web browser flash plugin's many issues or transport glitches in general. Pumping the Internet connection full of torrents is probably causing a latency bottleneck in the net connection itself. Same reason people don't download 12x large files whilst multiplayer gaming then complain of elevated ping times.

You just need a Killer Network and all will be fine

 

[ShintaiDK]

You just need a Killer Network and all will be fine

LOL. And you can even get Killer NIC WiFi now. And bundled with the Killer DoubleShot Pro software that will automatically pick the fastest ethernet or wifi Killer NIC connection. Or in other words just an ethernet only switch

 

[386DX]

Let me guess someone has a Realtek NIC.

 

[2is]

This is silly. You want a "5th core" for something that uses almost zero resources already even though you admittedly have no idea how well it would work.

Then we have the ultimate ignorance factor... You've been around since 2001 and you apparently don't know that there are several processors that fall between a G3258 and a 5820k. This isn't the first time you've made statements like that either.

So yeah, Intel isn't going to add a 5th media core. You're asking too much from your budget builds. So apparently yes, your options are to stick with your struggling G3258 or move up to a 5820k

You know, instead of asking for requests that will never happen, maybe you can ask Intel to make a processor that's more powerful than a G3258 but cheaper than a 5820k

 

[VirtualLarry]

This is silly. You want a "5th core" for something that uses almost zero resources already even though you admittedly have no idea how well it would work.

Then we have the ultimate ignorance factor... You've been around since 2001 and you apparently don't know that there are several processors that fall between a G3258 and a 5820k. This isn't the first time you've made statements like that either.

I certainly know nearly all of Intel's and AMD's CPU stack. The point being, if a G3258 isn't enough grunt, then what is? Where does it end? If you keep going up the stack, you end up at something like the 5820K. That's not ignorance; that's hyperbole.

So yeah, Intel isn't going to add a 5th media core. You're asking too much from your budget builds. So apparently yes, your options are to stick with your struggling G3258 or move up to a 5820k

You know, instead of asking for requests that will never happen, maybe you can ask Intel to make a processor that's more powerful than a G3258 but cheaper than a 5820k

Oh, so even though my processor is only 20% utilized, you're going to say I need a quad-core? You call me ignorant, but you consider a CPU that is 20% utilized, to be "struggling"?

My point, was not that Intel needs more CPU cores for more software grunt. Rather, my point was that having a dedicated core for media, would allow it to run a lighter, more-direct, software stack.

One of the responses suggested a Killer NIC. My suggestion is along the same vein. The idea is to short-cut the normal software stack layers, and create something that is dedicated (even if that were done "virtually") to media-playback tasks. Judging by the number of motherboards for gamers that have adopted the Killer NIC, and have "Killer" in their names, either that's some good technology, or some "Killer" marketing. Intel could learn a thing or two from them either way.

 

[VirtualLarry]

it's a network issue.

That may well be. I'm using a WD switch (they were so successful in the networking space that they aren't, now), and it supports some purported QoS. There are 8 ports. 4 are "medium", two are "low", and two are "high". I have all three of my PCs, and the uplink (all gigabit) connected to the four "medium" ports in the middle. When I connected some of my PCs to the "high" ports, my streaming and other things were even more inconsistent. So I may replace that switch with a more "normal" unmanaged SOHO gigabit 8-port switch.

Indeed. I have an i3 in my HTPC. Audio playback takes up all of 1% CPU usage. You could stick 32x cores in there and still suffer from cr*ppy web browser flash plugin's many issues or transport glitches in general. Pumping the Internet connection full of torrents is probably causing a latency bottleneck in the net connection itself. Same reason people don't download 12x large files whilst multiplayer gaming then complain of elevated ping times.

Well, I'm only seeding, not downloading, so the downstream pipe should be pretty clear.

I did a Verizon network diagnostic, and it said something about middleboxes, and a 1448 MTU. I'm not entirely sure what that's about. I'm not on PPPoE, but I do have an IPv6 tunnel through HE.net, but I thought only the IPv6 traffic had a reduced MTU, I thought that my IPv4 traffic went out unmolested through my FIOS connection.

 

[soccerballtux]

mediocore technology?

 

[TheELF]

My point, was not that Intel needs more CPU cores for more software grunt. Rather, my point was that having a dedicated core for media, would allow it to run a lighter, more-direct, software stack.

It has one, it is call an intel hd graphics.

 

[TheELF]

Well, I'm only seeding, not downloading, so the downstream pipe should be pretty clear.

You still have to tell the stream that it should keep playing,upload hurts your communication with the servers a lot.

 

[2is]

I certainly know nearly all of Intel's and AMD's CPU stack. The point being, if a G3258 isn't enough grunt, then what is? Where does it end? If you keep going up the stack, you end up at something like the 5820K. That's not ignorance; that's hyperbole.

How about all the processors in between the two that you apparently know about but are ignoring? "where does it end" is up to you. Chalk up one more thread where you're complaining about performance. It will end when you decide to configure a system that can meet your needs. It's up to you to decide if, when and how that happens. There's no need for a "media core" that's beyond silly. Notice how nobody who's replied to this thread has this issue, and no one has a "media core" enabled CPU.

 

[JeffMD]

Interesting thread... maybe I can shed some light on where the issues in media playback are. Intels processors are super efficient in decoding and displaying video playback. In dual GPU situations where laptops use the intel iGPU and a dGPU, there is no need or want to use the dGPU like nvidia for movie playback. The cpu and iGPU do it perfectly.

So then, whats it take to decode an HD movie? On any and every modern cpu, nothing. If we go all the way down to intel's dirt cheap windows tablet cpu, the Intel Atom Z3735G, it can still do 1080p video just fine using local file playback with a good player like MPC.

So what causes stutter? Well if we look at youtube and HTML 5.0, the HTML player itself has some overhead. This is ignoring of course and issues with the bandwidth not getting data fast enough. I find that youtubes html player is pretty solid, and any phone with a Snapdragon 410 should have little trouble with it. FLASH players are a little more troublesome. Flash video has not had any updated and improvements for quite some time, is based on old tech, and has always been slow. I compress small flash videos all the time and for years now it's hardware acceleration has never worked on my PCs. Another issue is 10bit video. Only really seen on anime fansubs, 10-bit color video tends to reduce color banding at low bit rates so even if you don't have 10bit color displays, there is still reasons to use it. Unfortunately no hardware supports 10bit video decoding, so playing back 10bit videos requires %100 software decoding from the cpu. Also another note that deals a lot with subtitles, until the past year or two styled soft subs would also require enough cpu power to give low powered systems and issue. Thankfully players like MXplayer on android have a built in refresh limiter that makes them much easier to handle.

 

[Arachnotronic]

I certainly know nearly all of Intel's and AMD's CPU stack. The point being, if a G3258 isn't enough grunt, then what is? Where does it end? If you keep going up the stack, you end up at something like the 5820K. That's not ignorance; that's hyperbole.

http://www.newegg.com/Product/Product.aspx?Item=N82E16819116989&cm_re=4690-_-19-116-989-_-Product

 

[VirtualLarry]

How about all the processors in between the two that you apparently know about but are ignoring? "where does it end" is up to you. Chalk up one more thread where you're complaining about performance. It will end when you decide to configure a system that can meet your needs. It's up to you to decide if, when and how that happens. There's no need for a "media core" that's beyond silly. Notice how nobody who's replied to this thread has this issue, and no one has a "media core" enabled CPU.

I should probably mention at this point, that the problem doesn't seem to manifest in Windows 7 64-bit. That's why I was talking about the scheduler differences. MS's Multimedia scheduler seems to work, for media playback, but it hobbles network transfers.

My idea, was to have a (virtually, if needed) dedicated CPU core for media playback, that wouldn't subject the rest of the CPU's cores / OS to scheduler hacks that slow down other things on the PC just to ensure media playback is smooth. It's the (semi-)obvious solution, just like Killer NIC.

Edit: I believe, that if we were all on XP still, without MS's newer Multimedia scheduler available, that doing heavy network transfers, would indeed cause media playback to lag and skip, just like Linux does.

http://www.zdnet.com/article/follow...er-performance-in-vista/?_escaped_fragment_=#!

This short article claims that the slowdown was fixed in newer versions of Windows. I'm curious about how.
http://www.ghacks.net/2007/08/29/fix-for-slow-vista-network-transfer-while-playing-music/

A more technical article, by Mark Russinovich:
http://blogs.technet.com/b/markrussinovich/archive/2007/08/27/1833290.aspx

This paper may be relevant, it came up during my search:
http://infocom2003.ieee-infocom.org/papers/05_03.PDF

Edit: After reading Mark R's blog post on the implementation of the MMCS, it semi-implements (sort of) what I was suggesting, but it also explicitly throttles network performance, which is the problem with the MMCS. If Intel implemented a way to dedicate a (virtual) core to MM playback, much in the same fashion that the MMCS boost that task to realtime, to prevent CPU starvation for that thread, but in a way that the rest of the system didn't have to be throttled, I would be for it.

Essentially, you would have to break the SMP-ness of the OS, and have one CPU core running the MM task at realtime, and have ONLY MM-playback-related ISRs fire on that CPU, while directing the rest of the system ISRs to fire only on the other "normal task" cores of the system. I don't know if you can program the APIC to do that or not. Maybe you can, and this is something that MS could look at for Windows 10. I wonder if they have looked at improving the MMCS?

 

[XavierMace]

Edit: I believe, that if we were all on XP still, without MS's newer Multimedia scheduler available, that doing heavy network transfers, would indeed cause media playback to lag and skip, just like Linux does.

I'm not sure why you believe this when one of your own links mentions the problem doesn't exist under XP. You've apparently just now found out that Vista had a lot of bugs at launch (poor SMB performance anyone?) and somehow decided that this is some widespread and still relevant issue that should cause Intel to overhaul their processor design to fix. There's a reason all the articles you linked to about this are roughly the same age.

That may well be. I'm using a WD switch (they were so successful in the networking space that they aren't, now), and it supports some purported QoS. There are 8 ports. 4 are "medium", two are "low", and two are "high". I have all three of my PCs, and the uplink (all gigabit) connected to the four "medium" ports in the middle. When I connected some of my PCs to the "high" ports, my streaming and other things were even more inconsistent. So I may replace that switch with a more "normal" unmanaged SOHO gigabit 8-port switch.

That's what you call a crappy switch and probably part of your problem.

Also, task manager is not a good tool for troubleshooting network performance "issues".

 

[VirtualLarry]

I'm not sure why you believe this when one of your own links mentions the problem doesn't exist under XP. You've apparently just now found out that Vista had a lot of bugs at launch (poor SMB performance anyone?) and somehow decided that this is some widespread and still relevant issue that should cause Intel to overhaul their processor design to fix. There's a reason all the articles you linked to about this are roughly the same age.

You didn't read them very carefully. The problem that XP "didn't have", was network slowdown, during media playback. What I stated was, that heavy network traffic could cause XP media playback to skip.
Those statements are not in conflict, and the reason is explained in Mark R's blog post. If you read that, you can see this this is not a bug in Vista, rather, it is intentional design. (The MMCS rate-limiting network traffic PPS, so as to reduce interrupt load on the CPU cores.)

 

[VirtualLarry]

That's what you call a crappy switch and probably part of your problem.
Also, task manager is not a good tool for troubleshooting network performance "issues".

Oh, that's the thing, though, I don't see the issue in Windows 7, really, so if the switch was truely the cause, I would expect the problem to manifest in Windows as well. (I should clarify, I did see some issues when I was using the "low" and "high" ports, it seemed like it was using XON/XOFF-style rate-limiting QoS - I know that ethernet doesn't use XON/XOFF, that's an analog modem term, but I know that there is some 802.11 flow-control protocol, but I don't recall the proper name for it.)

I should also note, that there are some threads here, I don't remember if they are in the *nix forum or where, that I read some people complaining about Firefox on Linux, specifically, lagging, especially on JS-heavy sites like Amazon.com.

This radio player site is very JS-heavy too.

The problem may simply be Firefox on Linux, seeing as how FF is basically single-threaded, and Flash Player depends on the browser to feed it data to stream, so if the browser is running some other task at the time, and there's a break in the buffer, well, there you go.

 

[escrow4]

]Oh, that's the thing, though, I don't see the issue in Windows 7, really,[/B] so if the switch was truely the cause, I would expect the problem to manifest in Windows as well. (I should clarify, I did see some issues when I was using the "low" and "high" ports, it seemed like it was using XON/XOFF-style rate-limiting QoS - I know that ethernet doesn't use XON/XOFF, that's an analog modem term, but I know that there is some 802.11 flow-control protocol, but I don't recall the proper name for it.)

I should also note, that there are some threads here, I don't remember if they are in the *nix forum or where, that I read some people complaining about Firefox on Linux, specifically, lagging, especially on JS-heavy sites like Amazon.com.

This radio player site is very JS-heavy too.

The problem may simply be Firefox on Linux, seeing as how FF is basically single-threaded, and Flash Player depends on the browser to feed it data to stream, so if the browser is running some other task at the time, and there's a break in the buffer, well, there you go.

So why are you not using 7 or better yet 8.1? Linux drivers are always second rate hacks anyway. When was the last time a company released Day 1 perfectly working Linux drivers for a slab of hardware?