Intel WLAN 5300 DPC spikes every 10 seconds

Ever since I've been aware that my GPU generated DPC latency spikes when throttling I've become DPC-paranoid

I noticed that my Intel 5300 wireless adapter generates yellow spikes ~ 1000 us every 10 seconds. It's much cleaner after a fresh reboot but after a while surfing/general use the spikes go up.

After fresh reboot:



After several hours:



I tried to update the wlan drivers to no avail. I know there's nothing alarming about yellow spikes but I was curious to know if someone had a solution in mind. Thanks.

EDIT: oh yeah i should mention that I know the wlan is responsible for these spikes because when i disable it they go away

 

so no one has a clue?

 

yeah I see the same thing on my asus G1 with the latest Intel driver, It was worse when I tried the default Microsoft driver. Not sure what you can do at this point, other than disable the wireless controller when doing audio stuff that requires low DPC latency.

 

well at least i'm not the only one...

 

Laptops with cheapo Firewire chipsets have this kind of problems.
Unfortunately, even if NP8662 is a high end laptop it still has a low-end JMicron chipset, hence the problem.
Read more here: http://forum.notebookreview.com/showthread.php?t=400025

 

are you suggesting the spikes are caused by the JMicron 5 in 1 (or 7 in 1 i cant remember) card reader? To my knowledge it is the only JMicron device present on my system. I tried upgrading the JMicron drivers to no avail. But when I disable the card reader the spikes continue, the only thing that will stop them is disabling the WLAN adapter. And correct me if I'm wrong but my WLAN is from Intel, not JMicron.

 

I would of thought this would be a byproduct of the way windows uses the Wlanautoconfig utility to search for networks.

http://www.martin-majowski.de/wlanoptimizer/ is a tool that disables the service once its running.

Its primary target is towards gamers as it normally causes one large ping spike every 60seconds.

 

interesting... i tried that tool but it didnt work though

 

You can disable it through the services panel, the problem is its the tool that finds your wireless network so nothing shows up rofl.

 

well i mean the tool seems to work but it does not get rid of the yellow spikes

 

Card reader, FireWire chipset and PCI bridge are all the same thingy which is JMicron. The problem is not that you get JMicron based DPC spikes only when using the card reader or the Firewire port. They are all part of the monolitic legacy PCI bridge bottleneck that interferes with the already measly pcie x1 link coming from the Southbridge that feeds the audio controller and all the external ports. Therefore, any non-Texas Instruments Firewire/PCI bridge will make all the DPC calls take longer time to complete which will slow down any device downstream from the Southbridge, such as the audio controller, but not only.

 

WLAN adapter is also wired downstream from the Southbridge and the lousy PCI bridge chocking the pcie x1 link.
WLAN may be very demanding on the traffic on that pseudo-pcie x1 link and I am not talking only bandwidth but also the bunch of WLAN generated dpc calls which compete for completion with the dpc calls from any other device downstream of that PCI bridge.
The problem is not only that the PCI interface imposes a bandwidth limitations on the pcie x1 link but also the fact that the PCI interface is the one that requires DPC calls in the first plase as opposed to the non-arbitrated communication along a pure pcie link.

 

wow i thought i knew quite a bit about computers but I'm completely lost! I've got some serious reading to do to understand what you just wrote... but it does seem to make sense, thanks! Bottom line is there's not much that can be done about those spikes right?

 

When you need better latency, e.g. for audio pro recording etc, you may try to disable the WLAN adapter in order to relieve the DPC queue size.

For more information, follow the discussions in the M17x threads:
http://forum.notebookreview.com/showthread.php?t=408300
http://forum.notebookreview.com/showthread.php?t=434691
http://en.community.dell.com/forums/t/19304621.aspx
http://notebookcritic.com/2010/02/23/m17x-dpc-latency-stuttering-and-lock-ups/

 

LOL! This is classic, (for me at least) after all the post's describing the problem we come to the same conclusion I did on post 3, disable the WLAN to reduce the DPC latency

Sure wish they would come out with a more efficient "south Bridge" type of port to deal with all these types of devices that are starting to take up more and more bandwidth.

 

No they won't. That's the whole point.

 

my bad I thought the point of the post was to actually help OP to fix the problem

 

I was saying that the lousy Southbridge design is the whole point with which ODM/OEM's are milking the money cow from the laptop owners. They certainly do not care much since a proper Firewire chip that would solve this particular problem costs only 5-8$. Saving on those max 8$ can screw high-end setups such as M17x beyond repair. And this is just a part of the whole problem.
I don't know if that is what you have understood from my post.

 

I guess you missed my sarcasm smiley, but hey no worries okay.

I do understand what you’re saying from both ends of the deal however. As an engineer and as a consumer.

I work at a small company that designs, manufactures and sells an electronic product. I also purchase the electronics components that go in our products.
With that I am given a task of designing a product that has a set COGS (cost of goods sold) so I have to make these same choices as your describing. And yes there are times when I want to use the more expensive part, but in the end can’t if I am to make the COGS. You have to remember there are literally thousands of parts in a notebook and if you can save x amount of dollars in the manufacturing, this adds up when you make tens if not hundreds of thousands of your product.

After all companies are in business to make money.

As a intelligent informed consumer who knows the cost and technology behind laptops it’s very frustrating for me. I find there is no perfect laptop that meets exactly what I want. So in the end I find what’s as close to that and learn to live with it.

What I was saying at the end of my post, was maybe Intel should look at creating a new interface that is more easily able to provide a bigger “pipe” to these ever increasingly bandwidth hungry “south bridge” type of peripherals like sound cards, Fire wire, audio interfaces, etc. or at least modify the way they are given these shared resources so that this latency issue becomes a non issue.

Sorry for the long rant….

 

If I may ask, does this spike cause any problems?

 

not that i can tell, except it makes my dpc latency graphs look bad

 

Could lead to audio stuttering, cracks and pops especially when you want audio pro recording. You should not have any problems though since you have a TI Firewire chipset.

 

Visible and/or audible problems appear when the spikes are 40-60k and above.
Everything under 16k can be ignored if you're not into pro. video/audio/graphics editing.