geekhack
geekhack Community => Other Geeky Stuff => Topic started by: tp4tissue on Wed, 25 May 2016, 21:54:30
-
OK...
So...
You haz the Intel 82579 Somewhere on one of your computers..
That computer has Dpc-Latency spikes.. There are no exceptions, if your pc has this lan adapter, you have this problem.
The latency spikes, manifest as audio/video stutters during playback.. sometimes, also causing in game input lag, or 3D render artifact/lag.
If your PC is from the SANDY-Bridge Era, 2011-2013, there is a large likelihood you have this adapter.. check device manager.
For Desktop models.. You must use Intel driver version 19.1 and BELOW, which will resolve intermittent DPC latency spikes..
https://downloadcenter.intel.com/download/24175/Network-Adapter-Driver-for-Windows-7-
For Laptop models.. such as lenovo x220, and many business class sandy-bridge builds with 82579lm, including different vendors HP, Toshiba, etc
Use the OLD intel lan driver version 11.10.84.0 and below
For lenovo users, you can directly download from lenovo @
https://download.lenovo.com/ibmdl/pub/pc/pccbbs/mobiles/83rw15ww_64.exe
^^^ This lan driver should also work for OTHER laptop machines with 82579lm Lan chipsets..
In the below image, The machine is torrenting at max bandwidth, it's been calibrated for low input lag (to play slither.io), which causes the maximum spike to be larger, because all QOS // flow control // Modulation features are disabled.
If you turn those things on, the maximum spike will be lower, but the PC will have higher input lag..
[attachimg=1]
-
[attachimg=1]
-
I've never seen a problem with it on my Asus Maximus board regardless of OS (it has the 82579V), and not all S.B. has it, my Gigabyte Sniper has a Realtek nic. Any lag I did find, I could always trace to other sources. I used both boards for my desktop (fps gaming) and as a file server, streaming video all the time without problems. If wireless G can stream Blueray HD under good conditions, certainly your gigabit can, or at least should.
Couple things...
QOS is best handled at the router level so it can shape traffic for all your systems, but keep in mind that too many rules can slow your router, so you need to be careful.
Second, why is your cpu usage and temp so high(85c)? Makes me wonder how close to over temp it gets at 100% load.
Third, torrents? There is much safer means to get movies these days.
Which brings me to something else...
Turning your torrent client can very often lead to spikes like you are seeing. That isn't to say that nic doesn't have issues, but I've seen a single torrent client as well as other software bring large, well tuned networks to their knees. They send and receive FAR more data than people realize. I mean a LOT, and cheaper home routers (even better ones) can simply get flooded out. I've even upgraded routers only to find the newer, more powerful one handled even less traffic. Do NOT go by tuning guides, the ones I have seen were way, way off. I found it best to watch transfer rates and router loads. Changing out your nic driver could hide this problem, or make it worse since the driver can alter how much I/O it can handle.
-
Well to test the X220, I'm running 2 full bitrate blurays over the network from another pc at the same time as my torrent, hence the high heat.. even at maximum 5300rpm fan speed.
It can run 100% cpu without getting to this level of heat, but not with GPU as well like with the 2 movies. running 2 videos also uses roughly 100% gpu core through madvr.
Torrents are still the best for high bitrate material. the popcorn times stuff and netflix is only for n00bie HD..
I'm not streaming from torrent.. I'm just using torrent to load the 82579 chip. To demonstrate that the fix indeed works..
I realize not all SB chips used the intel 82579 hahaha.. I said most of them did at one point.. because this was a hot seller..
If you never experienced it, maybe you lucked out, or you had the right drivers as released by Asus, as you know, they don't update their drivers as often as intel..
It's the Later intel drivers which cause the latency spikes..
For example, if you download the newest 20.x.x drivers for the 82579 series, latency spikes will for sure show up..
-
My ThinkPad W520 (Sandy Bridge i7) has this NIC.
It has severe lag issues with certain revisions of the driver. I eventually found one that seemed to calm them down to the point where they weren't apparent anymore.
-
My ThinkPad W520 (Sandy Bridge i7) has this NIC.
It has severe lag issues with certain revisions of the driver. I eventually found one that seemed to calm them down to the point where they weren't apparent anymore.
try the driver i linked, it should fix it..
the 19.1 driver from intel should calms it down a little, but you still get intermittent spikes if it's the 82579-lm
-
I am rebuilding a PC that has this NIC. It's going to be used for gaming purposes, so I'm not certain which driver version I should use. It's currently on 12.6.45.0.
HALP OH WISE TP4
-
I am rebuilding a PC that has this NIC. It's going to be used for gaming purposes, so I'm not certain which driver version I should use. It's currently on 12.6.45.0.
HALP OH WISE TP4
use this one..
https://downloadcenter.intel.com/download/24175/Intel-Network-Adapter-Driver-for-Windows-7-
19.1 is the last one which doesn't have the spikes..
-
The question I always wondered: is the latency associated with this NIC (yeah, maybe NIC isn't the best term for something onboard, but close enough) a hardware design/implementation problem?
- Is the NIC inherently faulty?
- Maybe some feature of the hardware doesn't work properly and only some drivers enable this feature?
- Is it something about the way the OE integrated it into its system design?
- Is the hardware fine but the drivers written in a way that asks the hardware to do something wrong causing the problem?
I was very 'into' trying to figure this out a while back, but when I found the drivers that worked for me I kinda left it alone after that.
-
Well to test the X220, I'm running 2 full bitrate blurays over the network from another pc at the same time as my torrent, hence the high heat.. even at maximum 5300rpm fan speed.
It's the Later intel drivers which cause the latency spikes..
Ahh, laptop, makes much more sense now, I was thinking this was your desktop you pictured
I'm not sure what drivers I was using, the last two years or so, the Maximus spent it's life in my server until I updated the graphics card and had a conflict (my graphics card refused to boot in the Sniper board), and I really never update the drivers on the server. It works, leave it.
When I did swap back, I didn't play much games (but never saw spikes on music or movies) and switched to Linux shortly after. It does get spikes, but that's due to Samba, not a specific driver.
Does this issue effect win10?
-
- Is the NIC inherently faulty?
- Maybe some feature of the hardware doesn't work properly and only some drivers enable this feature?
- Is it something about the way the OE integrated it into its system design?
- Is the hardware fine but the drivers written in a way that asks the hardware to do something wrong causing the problem?
I was very 'into' trying to figure this out a while back, but when I found the drivers that worked for me I kinda left it alone after that.
Your last line shows it's a driver issue and not necessarily a faulty product.
I suspect though, like you, that it may be a matter of how it's implemented, and conflicts with something else. While we rarely deal with them these days, it could be sharing an IRQ port on some boards with something making it more pronounced on some than others (wouldn't be the first time this has happened). Or a simple driver conflict. It could also effect some models more than others, I know there is at least 3 variants of this chip, some may be more prone to problems, or the driver only effects one model more than others.
Lots of variables.
While there are complaints about the problem, it doesn't seem to be big enough for massive outcry, now or at the time it was released. Though, some of it could have been lost in the whole Sandy Bridge chipset debacle (http://www.pcadvisor.co.uk/news/pc-components/intel-sandy-bridge-recall-what-you-need-to-know-updated-3259061/) that happened around that same time.
-
- Is the NIC inherently faulty?
- Maybe some feature of the hardware doesn't work properly and only some drivers enable this feature?
- Is it something about the way the OE integrated it into its system design?
- Is the hardware fine but the drivers written in a way that asks the hardware to do something wrong causing the problem?
I was very 'into' trying to figure this out a while back, but when I found the drivers that worked for me I kinda left it alone after that.
Your last line shows it's a driver issue and not necessarily a faulty product.
I suspect though, like you, that it may be a matter of how it's implemented, and conflicts with something else. While we rarely deal with them these days, it could be sharing an IRQ port on some boards with something making it more pronounced on some than others (wouldn't be the first time this has happened). Or a simple driver conflict. It could also effect some models more than others, I know there is at least 3 variants of this chip, some may be more prone to problems, or the driver only effects one model more than others.
Lots of variables.
While there are complaints about the problem, it doesn't seem to be big enough for massive outcry, now or at the time it was released. Though, some of it could have been lost in the whole Sandy Bridge chipset debacle (http://www.pcadvisor.co.uk/news/pc-components/intel-sandy-bridge-recall-what-you-need-to-know-updated-3259061/) that happened around that same time.
This is oldddd hahaha.. p67 and h67 chipsets..
Thank god I switched to z68
althoughhhhhhh..... i suppose my laptop could be affected..
-
The question I always wondered: is the latency associated with this NIC (yeah, maybe NIC isn't the best term for something onboard, but close enough) a hardware design/implementation problem?
- Is the NIC inherently faulty?
- Maybe some feature of the hardware doesn't work properly and only some drivers enable this feature?
- Is it something about the way the OE integrated it into its system design?
- Is the hardware fine but the drivers written in a way that asks the hardware to do something wrong causing the problem?
I was very 'into' trying to figure this out a while back, but when I found the drivers that worked for me I kinda left it alone after that.
It's driver problem.. because the earlier drivers including the release drivers worked fine.
My guess is it most likely has to do with power saving features of these newer lan chips.
since they're making one design across mobile and desktop platforms...
The lan chips have very complicated power states now, and jumping up and down is the issue.
-
Just in case people didn't catch the sata voltage leak problem.. If you have a p67 or h67 chipset.. double check the revision number.
(http://event.asus.com/2011/mb/Identify_B3_Motherboards/images/part2-CS_05.jpg)
-
Necrobump. Win10 updates on this topic.
I finally installed Win10 on my laptop, a ThinkPad W520 with the affected NIC.
Windows Update force-delivered Intel driver version 12.12.140.22 (release date 2015-02-15). This driver version exhibits the DPC latency spikes.
What I did was:
- Disabled automatic download of drivers (system properties, advanced system settings, hardware tab)
- Uninstalled the NIC driver in Device Manager and checked the box to delete the driver software
- Refreshed Device Manager/scan for hardware changes, Win10 reinstalled the packaged NIC driver that ships with the OS (provider MS, version 12.15.22.6, date 2016-04-05)
And like that, latency spikes are gone (testing so far has been brief, but it's at least better than it was with no detectable spikes unlike every few seconds previously), and MS will not force driver updates. Given the age of the hardware there's really not much sense taking driver updates for anything else when it's running stable...possibly except graphics but even then, often not.
Interesting to see that the driver date and version number indicate this is a newer driver that is in fact curing the issue, but Windows Update seems to favour the manufacturer driver which is older.
Update a few days later:
It was running stable with the driver described above with low network activity, but I realized that during prolonged network activity such as copying a large file that would take a couple minutes, the DPC latency spikes picked up quite severely, worse than the Intel driver that Windows Update grabbed.
I installed the driver tp linked earlier in the thread and it solved the issue, just as it has for Win7 previously.