257.15 Beta: DPC Latency issues (like ever)

I ran Windows Performance Analysis Tools DPC trace & I'm not having any issues with these drivers. Are you sure they are the culprit?

Keep in mind something as simple as the Windows power saving feature can cause issues, try setting it to performance & see if it helps. Also you can try disabling your power saving features in the bios, they can cause this issue as well, I do believe Windows performance power plan disables most of the of ones set in the bios.

I just took these, you can see quite a difference between plans:

1. Windows High Performance Power Plan


2. Windows Power Saving Power Plan

 

Thanks for the suggestions, I really appreciate it. But if you read my post you should notice how I repeatedly stated that "the 2D mode" of Nvidia's driver causes DPC latencies while "all 3D modes" do not (all means, Low Power 3D and both Performance 3D clock-rates that the 9600M GT uses).

I also already stated that using Rivatuner to force the Nvidia performance mode to a fixed 3d state fixes the issue (as does using Microsoft's Standard VGA driver that also uses a fixed clockrate roughly corresponding to Low Power 3D mode).

Furthermore I know really alot about the possible culprits that can cause DPC latencies/spikes in Windows and even some special Bootcamp/Mac related ones and thus can pretty much rule out everything else. Please read here again to get an idea of some of the various culprits: http://www.rme-audio.de/forum/viewtopic.php?id=4320

Just to underline it: I get pretty perfect DPC latencies of around 100 us and below as long as NVidia's driver is not in 2D mode! In 3D mode I can use professional audio hardware/software at lowest audio buffer settings without dropouts.

The issues were present from the very first Bootcamp/Nvidia driver that came out with the 2008 Unibody Macbook models and still persists with the new generation 257.15 Beta driver. The only thing that was fixed is that there are no more *enormous* spikes whenever clock-rates are switched (which is a very good thing already, but not enough).

Besides: DPCs should not be measured in Idle, but preferably in a loaded/working state (or both). That does not only mean CPU load, but also hardware load since you want to find out which hardware/driver occupies the CPU longer than it should.

 

Well I usually suggest the basics to start with, since everytime I jump to the more technical I get complaints that the person can't follow what I'm saying. Please post a DPC Interupt Performance Analyzer Trace File. That's the only way to get to the bottom of it.

 

You are using ASIO correct?

 

Yes, I am using ASIO. Unfortunately the switching of Nvidia clock-rates is not fully fixed either (see screenshot below).



Here is a trace of 2D mode while running Ableton Live (audio is *disabled*):



In one trace run I had exorbitantly high DPC readings via DPC Latency Checker. Interestingly these turned out to be a combination of the Nvidia driver and usbport.sys while using Live's GUI (which is somewhat questionably programmed though and causing its own load of problems).

 

no its not it for nv4_disp most nvidia desktop cards lol

 

Guys, could we concentrate on the issue here instead of discussing driver versions? The 257.15 driver is officially offered for download for my 9600M GT when I click my way through NVidia's driver choices. The problems are the very same as with any former driver.

Here is the official Nvidia list of supported products to end this part of the discussion:

GeForce 300M series:
GTS 360M, GTS 350M, GTS 250M, GT 335M, GT 330M, GT 325M, GT 320M, 310M, 305M

GeForce 200M series:
GT 230M, GT 240M, G210M, GTX 285M, GTX 280M, GTX 260M, GT 220M

GeForce 100M series:
G 110M, GT 130M, GTS 160M, G 105M, G 103M, G 102M, GT 120M

GeForce 9 series:
9600 GT, 9400

GeForce 9M series:
9400M G, 9650M GS, 9200M GS, 9500M GS, 9400M GS, 9800M GS, 9300M GS, 9800M GT, 9100M G, 9800M GTS, 9500M G, 9700M GT, 9600M GS, 9650M GT, 9400M, 9300M G, 9800M GTX, 9600M GT, 9700M GTS

GeForce 8M series:
8800M GTS, 8600M GT, 8400M GT, 8200M G, 8800M GTX, 8800M GS, 8600M GS, 8400M G, 8700M GT, 8400M GS

Quadro FX Notebook series:
FX 3800M, FX 3700M, FX 360M, FX 1600M, FX 380M, FX 370M, FX 3600M, FX 2800M, FX 2700M, FX 880M, FX 770M, FX 570M, FX 1800M, FX 1700M

Quadro NVS Notebook series:
NVS 5100M, NVS 3100M, NVS 150M, NVS 160M, NVS 140M, NVS 320M, NVS 130M, NVS 135M

ION series:
ION LE, ION

 

Is there a previous driver in existence that does not exhibit this issue? If so I would suggest just using it. I realize you would like the issue resolved in current and especially future drivers but I don't think you'll get it working with the latest beta by any means. If you have not already posted on Nvidia's forums I would suggest doing so even though this has been a long standing issue and I'm quite certain they are aware of it they just have more important fixes to implement especially in the wake of a new product launch.

 

http://forums.guru3d.com/showthread.php?t=264258&page=3

old post to dpc

 

As both a gamer and audio user who uses both Macs & PC's (for audio & graphics) I'd just like to chime in a bit of support for the OP's efforts here. I pretty much gave up having y nvidia drivers clock down into low power 2d mode during dual/triple monitor usage. This is the case with my GTX285's under Win7 though I no longer run bootcamp on the laptop as I got tired of maintaining both OSes and went for a clean snow leopard install there. (I also don't run boot camp on my Mac Pro anymore, as I have a Win7 Xeon class machine that's separate.)

Also I have noticed that when my machine shuts off the monitors it does go into low power 2d mode, so at least I'm getting the power savings from that when the machine isn't in use.

 

Something I just noticed though I find it highly unlikely to be related to the original poster's topic.... my DPC was severely spiked (hadn't checked in quite some time), and by severely spiked I mean the bars were at the top of the latency checker window so I was racking my brain trying to figure out wtf was going on and finally after about an hour and a half I closed my Daemon Tools sidebar gadget and everything is back to normal. Very strange because I never noticed it before however I'm not sure if I ran the checker after installing the latest Daemon Tools update. I use pro standard so I'm not sure if the gadget is the same as the free edition and if the problem would exhibit itself there. Just goes to show you how little it takes to impact DPC latencies with a little bit of crappy coding.

 

all fine here.


looks more like something else is wrong.

 

None that I know about. I started with the very first 9600M GT driver coming with the first 2008 Unibody Macbook Pro and had the problems from the start. No driver after that solved the issue eventhough I used NVidia then newly setup Feedback webform to report the issues (just did it again).

I will give the forum a try. But thinking about how this issues persists for over 1.5 years now I have little understanding for any "more important fixes in the wake of new product launches".

BTW, the current 21" Imac that come with 9400M do *not* exhibit the DPC issues. And according to this old post there were such problems with the 9400M in the past which where then fixed by a driver update:

 

According to your profile you are using a different graphic-card. I am using a 9600M GT.

 

http://forums.nvidia.com/index.php?showtopic=157614&hl=dpc

 

Enabling 2 monitors does not invoke 3D Mode (easy to check in EVGA Precision).

You will find no Active 3D Process !

It does however run the same clocks as 3D mode as the newer Nvidia GPU's do not actually have separate 2D and 3D clocks due to the newer power saving mode (3 stages of lower mhz).

You can however simply set the clocks as low as EVGA Precision will let you and save the profile then set up a 2nd profile for 3D use and tell Precision to use them for 2D use and 3D use.

That is the way I do it and this includes fan speed, when I disable my TV I get power savings like 301/602/100 right now

 

ah hell, just get this (since your running W7)
http://us.store.creative.com/Sound-Blaster-XFi-Surround-5.1/M/B0017QQQAE.htm
But if im being too ignorant, ill get a little technical because your problem is that your using a windows OS on a macbook that probably interferes with what windows was intended to initially run on (a PC) and the ASIO drivers on it. Now the way ive only experienced high Us was due to the Nvidia Nforce/ Creative "snap crackle pop" issue. Ive also heard that W7 has alternatives to ASIO and WASAPI interfaces. but i havent read much on that.

 

So ?
my point was their nothing wrong with 257.15 regarding dpc latency.

 

Reproducing issues is best done using the same hardware, you know.

 

OP might wanna change topic tittle then.