Hard Page Faults ! (learned something new & important)

[Cableaddict]

I recently posted this info on my "Does faster ram allow for a lower HW buffer?" thread, but it deserves it's own thread, so -

We all know about low DPC latency, but many of you may not know about Hard pagefaults. They are related, and evidently extremely important.

note: I'm not a computer expert, so do your own research for details or maybe even to find out if this isn't as important as I think it is.

But I DEFINITELY think it is. Especially in the quest to optimize for the lowest HW buffer possible.
-------------------------


A pagefault, evidently, is like really bad DPC activity, but just for a few ms.
A HARD pagefault seems to be just really bad ones, or maybe pagefaults that happen repeatedly.
I'm not sure. (Correct me if I'm wrong.)

A common cause of bad pagefaults is a buggy driver. It turns out that some NVIDIA drivers of recent past have been horribly buggy in this regard. Despite having super-low DPC latency in my current rig, I have been getting obscenely large hard pagefaults. I mean right off the scale. And once they started, they happened at an alarming frequency.

I tried newer NVIDIA drivers, even older NVIDIA drivers, but no love. Still suspecting this as the cause, I decided to replace my 5 year old video card with a new, GDDR5 based one. I figured that NVIDIA might be more concerned about this problem with their newest drivers, so fingers crossed.

Sure enough, WITH THE NEW CARD, I NOW HAVE ZERO HARD PAGEFAULTS !

And not for nothing, but beforehand I couldn't run a 64 buffer unless I turned off several VSTi's, and even then I got minor crackling. Now I'm running 64 buffer with everything on, and not even heavy sustain pedal causes problems.

So... it seems to be a big deal.

Again, I don't really know. YMMV.

[Dstruct]

How does on see if there is a hard page fault?

[gibi25]

I use this:
http://www.resplendence.com/latencymon

[ivansc]

Interesting... I too get a fair sized reading in that software but since I already get pretty good performance have never bothered to explore any further.
But I have had issues with Nvidia drivers on previous occasions.
Off to google/explore.

[Stella645]

Have seen many recommendations to run a clean custom install and only install the basic driver for Nvidia cards.
The other installs can be more problematic than the driver itself.

[karbomusic]

Originally Posted by Dstruct

How does on see if there is a hard page fault?

In modern times this shouldn't really be a problem unless the machine is under memory pressure. AKA it means a page in memory has been previously paged out to disk, and now the OS has to go to disk to retrieve it. Short story if you remember the days of not having enough memory and your disk thrashing because it's constantly reading/writing from the page file, that's what this is.

So, it shouldn't even be relevant unless there is memory pressure. If you wanted to measure them however, you can look via perfmon.exe under the memory counter:

Finally, when investigating page fault issues, it is important to understand whether the page faults are hard faults or soft faults. The page fault counters in Performance Monitor do not distinguish between hard and soft faults, so you have to do a little bit of work to determine the number of hard faults. To track paging, you should use the following counters: Memory\ Page Faults /sec, Memory\ Cache Faults /sec and Memory\ Page Reads /sec. The first two counters track the working sets and the file system cache. The Page Reads counter allows you to track hard page faults. If you have a high rate of page faults combined with a high rate of page reads (which also show up in the Disk counters) then you may have an issue where you have insufficient RAM given the high rate of hard faults.

https://blogs.technet.microsoft.com/...f-page-faults/

If memory serves, if Reaper is waiting due to this it will cause transport bar flash/turn red.

[Cableaddict]

Originally Posted by gibi25

I use this:
http://www.resplendence.com/latencymon

That's what I use also.



Karbo, that info is interesting.

At the same time, if I was having "insufficient memory" problems, (which I doubt) then the sole cause of that had to be the older NVIDIA card / drivers, since as I wrote, my hard pagefaults are now virtually non-existent. - And nothing else changed in my system.

I remember way back in the Stone Age (Windows XP) there were important issues with video cards and something called "shadow ram." Something about how the cpu had to use ram to "handshake" with the card, so certain cards were too much for the XP limit of 4 GB. (Something like that.)

Maybe that's still the case.

More research needed .....

[Dstruct]

Originally Posted by gibi25

I use this:
http://www.resplendence.com/latencymon

This tool says that REAPER is having the most hard page faults here.

[Stella645]

Originally Posted by Dstruct

This tool says that REAPER is having the most hard page faults here.

And? Is there a problem with your system performance?

[Dstruct]

No problems here.

[karbomusic]

Originally Posted by Cableaddict

At the same time, if I was having "insufficient memory" problems, (which I doubt) then the sole cause of that had to be the older NVIDIA card / drivers, since as I wrote, my hard pagefaults are now virtually non-existent. - And nothing else changed in my system.

A hard page fault only occurs if a page of memory is requested and it is on disk instead of the memory stick or potentially... Memory that services hardware is usually different, it can't be paged, ever (look up non-paged pool memory) it's likely a hard page fault concerning the interrupt to service a DPC or similar but would need to look that up. The bigger point I wanted to make is there is probably no need to deal with page faults (if you have enough memory) since other methods such as DPC measurements are skinning the same cat.