So what does the DPC latency tool from Thesycon say?

3 things first, for the rest maybe the forum can help:

- on the RayDAT Settings dialog main tab set the number of WDM devices to 0.

- on the Tab About make sure 'Optimize Multi-Client Mixing' is switched off.

- on the same tab try 'Enable MMCSS' on and off - one of both might be better.

64 is about the limit for my needs. but the point i spent all the money on the RME and rack in the first place was the 32 sample buffer size. that was the main selling point to me.

yep, tried all this already. Multi-client mixing was never on. im not sure if MMCSS helps, also Reaper has an MMCSS option in the audio setup so im not sure how they interact with each other

my main issue is this - in a well oiled, well configured PC with an RME RayDAT, what real-time CPU usage percentage should you expect it to be stable at?

I would have thought maybe 80-90% before it starts stressing out?

As it stands, my setup will start showing pops/clicks around 30-40% real-time CPU usage, and by about 50% real-time it will be too annoying to deal with.

My non-realtime "Total CPU" is often around 10-15%. When I render a track it usually renders at 2x realtime speed. My CPU is MORE than powerful enough for my projects, and I know in my gut Im not getting the best performance out of the $4000 i spent on all this stuff.

ok ramses, thanks for the instructions, i will do this ASAP and post the results.

I have a similar, rather older HDSP fanless setup  with B75 GA board and 3rd gen  i5 T CPU. I use 11rack module and GR5 via spdif or AES it depends on config,  I  never get higher than 5ms RTL  even with stacked up VSTs  regardless of boot option:  win7/HOSx 10.8

Break down your latency to see what ads to it when it spikes. I bet it has nothing to do with the RME card. 

20 MS RTL is brutal. It is likely software related. If board posts OK and card looks solid in hardware panel id try to change daw and vsts and see if that changes anything.  Something is not right thats for sure.

thanks Ramses now it makes sense. Id keep playing with memory sticks m.2 drives take the cards out, replace or disable graphics. change cpu, cpu speed steps monitor monitor.. if nothing changes id replace the mobo and do it all over again. stay away from amd and radeon. stick to nvidia and gigabyte.

It is always wise to follow proven setup and use older cards. This is exactly why i hate to be a vanguard. i always copy working rigs for my essential stuff and experiment with nonessential stuff.

Ok @ramses, I ran the LatencyMon test, heres my results:

Min DPC latency: 15µs
Usual DPC latency: 25-40µs
Peaks: 50-70µs
Peak Frequency: around 1-2 peaks every couple of minutes or so

Pastebin of the Stats tab:
https://pastebin.com/PXPiwPmw

Main tab:

Drivers tab:

Processes tab:

CPU tab:

Hi

thanks that you took the time and effort to post these values.
I noticed some things that look strange and should be fixed.
I have a certain feeling that this correlates with the issues that you experience with the RayDAT card.
In a good case of luck you get it fixed and then the RayDAT should work without issues.

1. Your CPU has 4 cores and 8 threads, LatencyMon on the one hand tells you have 8 locical processors
    but at the end of the report you see, that all DPCs and ISRs are being executed only on Core 0.
    Please check in the BIOS, whether you disabled Hyperthreading and re-enable.
    I know that some people claim, that for some systems it shall be better to disable Hyperthreading.
    BUT .. all drivers, interrupt processing, usually benefit from having more cores available to be distributed across them.

2. To be on the safe side, reinstall the latest version of your chipset drivers !

3. In your LatencyMon report the measured CPU speed is 1 Mhz and there is a big Note and a WARNING:
    "Note: reported execution times may be calculated based on a fixed reported CPU speed.
      Disable variable speed settings like Intel Speed Step and AMD Cool N Quiet in the BIOS setup for more accurate results."
    "WARNING: the CPU speed that was measured is only a fraction of the CPU speed reported.
     Your CPUs may be throttled back due to variable speed settings and thermal issues.
     It is suggested that you run a utility which reports your actual CPU frequency and temperature."
     Please check / adjust your BIOS settings so that the CPU stays on constant speed.
     While you are at it disable also Spread Spectrum, you can google what this is.

     If you want to compare with my BIOS settings in regards to processor speed, then look at this:
     http://www.tonstudio-forum.de/blog/inde … -X10SRi-F/
     It won't be compareable 1:1 .. but maybe this leads you to a good idea.
     Cross check with your mainboard handbook and internet what the different settings in your BIOS mean.
     Its not that much, in regards to speed / clocking / energy saving there are only a few.

4. Your nvidia graphics driver uses a lot of DPC routing execution time. Its shows the highest result with 514us.
    My nvidia driver for the GTX 980 only uses a fraction of it with 109us highest execution time.
    Sadly you didn't adjust the columns so that some nice information about what driver you use is not visible.
    I can tell you from own experience, that there is always a chance to catch a bad driver which led for me
    to a very bad performance. I can confirm to you, that at least this version works for me very well: 375.70
    As everything in terms of DPCs and ISRs is being executed through CPU0 (see my point #1), I could imagine
    that this might have a big impact on processing audio in time for your RayDAT.

5. Your ATAPORT.SYS driver has a ver high ISR/DPC count and if you look at the Total Execution time,
    its much higher than mine. You: 1631ms, me: 237ms.
    a) Please check also in the BIOS, whether you use AHCI for your drives.
    I would also check two other things
    b) whether you are using the proper SATA drivers
    c) whether your disks are connected to the SATA ports which connect directly to the Chipset of your mainboard.
    The performance with other add-on controllers (which use different drivers) could be worse.
    Use AS SSD tool to check:
    - which driver you use (msahci is the microsoft driver, iaStor is the Intel Raid driver, ...)
    - whether the alignment of your disks, especially on SSDs, are well (while you are in this tool)
    http://alex-is.de/PHP/fusion/downloads. … nload_id=9
   
My points #1 - #5 go somehow hand in hand with respect to audio processing performance.

Solve them one by one and remeasure after every change, so that you can tell later,  what the root cause was.
I would process the points in this order: #1, #2, #3, #4, #5

Careful with #3. Changes in BIOS ...
You surely have a digital camera. Make a photo of every BIOS setting page.
You will be thankful, if you can look afterwards, what the initial settings were from where you started doing changes.

It could also be possible that the system needs a BIOS upgrade.
I would do this as last thing, if steps #1 - #4 didn't bring any solution.

Your system seems to be highly optimized as not so much is running on your system.
In my screenshots I marked some values which are remarkable / different to your values.
I ran my tool nearly the same time like you so that we should have compareable counters, but they are sometimes very different.

I can't predict whats happening on your system if you do a change.
Therefore I said, one change at a time, and then observe whats changing on your system.

I gave you an URL to an article how I fined tuned my BIOS.
At the end I left EIST in .. and then fine tuned my Energy profiles in Windows, its all described in the article.

You can use CPU-Z to see whether your CPU clock stays stable or not.

And you can use parkcontrol to disable CPU core parking, which is also worth doing and there you can set whether you allow for throttling the frequency or not.

EDIT: So in high energy profile now, my CPU stays at 3.6 Ghz.
My full turbo speed I cannot reach anyway, as with deactivation of parking CPUs, Turbo speed cannot be reached.
You need to turn down a certain percentage (around 25% of cores or even more) before its allow to operate cores at turbo speed not to exceed the max temp target ...

And here how to check the Disk Driver and SSD alignment

Some other suggestions around BIOS / Driver / HW usage:

OK this is weird. I rolled back to LatencyMon 4.02, and downgraded to the 375.70 Nvidia drivers like you suggested

Now my DPC latency is averaging around 2-10, with occasional bumps to around 20

BUT in 15 minutes testing time, I got 3 BIIIIG DPC spikes which ive never seen before, one around 100, one around 140, and the biggest one was 434

Also the "Highest reported DPC routine execution time" was still 547 for nvlddmkm.sys NVIDIA Windows Kernel Mode Driver

Heres the new test results with LatencyMon 4.02

Stats:
https://pastebin.com/fS5Nc6jG

Imgur album:
http://imgur.com/a/NcGGK

Please don't do multiple changes at once, as I mentioned Now you don't know whether you get different numbers or new spikes because of LatencyMon 4.02 or because of your nVidia driver change...

And I'd suggest to better stick to your previous version of LatencyMon, as otherwise we do not know exactly, whether its simply a tool issue, when number should change ... Nobody of us knows what has all changed in different releases of LatencyMon .... Then we would have to do all from the beginning with the new tool. I am not keen to spend again 2h on this tbh ...

But what you SHOULD compare between the previous version and v4.02 is, whether some outcomes are equal no matter which version you use
- is the shitload of interrupt processing still on CPU0 only ?
- are still only 4 out of 7 CPU threads visible, in both versions of the tool (multithreading)
- is the CPU frequency also not recognized by the tool ...

This would be an interesting comparison to get a feeling for whether your higher LatencyMon
version has maybe issues in measuring

But for further analysis of changes we should stick using your higher version of LatencyMon as I said.

BTW which version do you use ? And what are the settings in there ?
Higher versions are configurable how measuring shall be done, in the newest version there are 3 different options.
I would like to know to what you have set it.

its a GTX 710 passive card

as for the differences in LatencyMon 4.02

- it still seems like CPU0 is getting all the interrupts/DPC
- it still looks like only 4 CPUS are being tested/visible (but in my DAW it definitely spreads the CPU load across all 8 hyperthreaded cores, so Im not sure whats going on there)
- the CPU frequency IS recogniced by LatencyMon 4.02

originally i was using 6.5.1 (the latest i believe)

Lets stay in this thread please.
I moved my answer now to this thread to where my answer belongs to in terms of the order of actions, questions and answers.

This is a big driver package containing drivers for many different products of their product line.

In a driver you also need sometimes workarounds for specific HW .. as HW and their implementation
simply differs or can have bugs. SO .. for one HW a certain driver package can work, for another fail.

Same with mainboards, in a bad case you can also have, besides a shitty HW component or driver
a shitty mainboard with errors in the Design or in the BIOS.

Everything is possible and YOU need to look now carefully change by change like an detective
what does a change.

And the major issues I gave to you, I fear the rest is more or less on your own by trial and error.
Certain things you should nevertheless clarify with Gigabyte.

If you have multithreading enabled in BIOS and 2 different versions of LatencyMon do not see all of your CPUs ...
Well I think this is worth asking Gigabyte how this can be ...
Maybe you need to plan for a BIOS update.

Another thing is, that you seem to have many many changes to your machine if I look to all the stats.
Its much likely that you require a separate reference installation to ensure, that you didnt optimize your system to death by deactivating something or a service here and there.
Usually companies give only support on smth like a standard installation.
Can you make sure, that your optimizations are not the culprit in this case ?
I don't want to make you frightened, but at the end it comes down to this.
If you do many changes to a system, then you take over (or need to take over) responsibilites for your changes.

So .. I would recommend to you to make something like a reference installation, so that you know, that your optimizations are not the culprit here.

In terms of BIOS I would always start with the Default settings, Photos what the settings are.
Then load the optimized defaults. Photos what these settings are.
Compare, whats the difference.
And then make your "well thought" settings and test it with a Windows standard installation,
where only LatencyMon, CPU-Z and perhaps AS SSD are installed.

And then I would install only the real required drivers.

And before any change or installation of software or driver I would set a Windows Recovery Point so that its easier to backout a change completely including Registry settings etc ...

When you install your system I would also make sure not to use / disable
- secure boot and
- UEFI / DUAL Boot, choose Legacy Boot

this removes much of the error prone-ness of Microsoft Secure Boot allures to control everything ...
and also in case of emergency it makes replacement of HW easier.
If you simply change this now it could happen that you need to reinstall ...

I would buy another SSD now and do all the changes on a test system.
And I would buy another Graphic card to be able to make a quick swap to exclude issues with the one you have.

And careful .. before making changes to your system, make a system backup (disk image)
by taking a reliable software like Macrium Reflect.

Good luck !

PS: now you know why there is a market for having optimized system for audio recording.
For office PCs this all is not critical. But if you want to have best performance for audio processing,
then it can become at times tricky, its easy to overlook something if you are not familiar with all the stuff and tuning.
I took my time, had already a good basis skill from my Unix time also with developer background.
For somebody who does not have all this Hardware / OS and Troubleshooting experience this can become tedious work.

This encouraged me to tell the community, how to build a stable and well performing system based on quality products patially taken from Enterprise grade parts (Xeon CPU, ECC RAM, server board):
http://www.tonstudio-forum.de/blog/inde … mponenten/

The system would have costed more then double the amount of money if I would have bought it from such a manufacturer, who optimizes it for customers for recording purposes. At that time sadly with older components, why I took the challenge to do it on my own ...

The nvidia downgrade in combination with your particular HW.

Try also other PCIe sockets please or other recommended versions of nVidia SW.

BTW .. you mentioned it already, some aspects of your system are even better than mine ...
But .. the essential things are working for my system, its also optimized in many ways, so that it works perfect.

http://www.tonstudio-forum.de/blog/inde … cks-de-en/

Yes and pls repeat your 15 min test and post the results to have an new baseline to compare against.
We can not exclude, that maybe measurement / results of the 2 different versions changed.

BTW .. I also did the performance test of 400 track project with ASIO buffersize 32 sample with the RayDAT.
Same excellent results like with the UFX+.

http://www.tonstudio-forum.de/blog/inde … cks-de-en/

Ramses, off the topic, but not necessarily. We talked about data sampling rates few months ago. Have you ever thought about bus data being affected the same way. I usually use lower frequency processors t,p, and s, and rather slower memory for my rigs. definitely not overclocked. Is usually less expensive too. 

I have a rig with the same board as discussed here but using 4770T  and vengeance memory.  I use ZOTAC GT 710 pci x1 to save my x16 slots. It keeps some heat off my main cpu. everything is bellow any suspicion threshold. I wipe off all pci fingers to minimize corrosion and dust the boards with the compressor.

I avoid fans if possible and and use PSU about 30% above any tested sustained peak.  everything is always fan-less and clocks set to the lowest end if dedicated function allows so. I use FLIR to look for suspicious cards and joints and get clues on heat dissipation. As my wife said before i have a dude that dremels extra holes and fill with a fine mesh i find honey combs too open to metal particles.

jackyloup, it would also be great to know which PCI port you plugged your RayDAT into.

New reference installation is mandatory.
New disk, start with BIOS defaults, photo series BIOS, then choose optimized defaults, photo series.
Find out differencies between BIOS default and optimized settings (the optimization is usually NOTHING, but who knows).
Ensure at this point to disable UEFI and secure boot (its crap).
Ensure that disks run in AHCI mode.
Disable all the HW you do not need (serial, lpt, sound, wlan, bluetooth, ...)

Then install windows. *** Make a Disk Image here that you can easily come back to here ****
Then install your test tools *** 2nd Disk Image ***

Do here your first baseline of system performance behaviour.
Only one change at a time (be it driver, bios, whatever change) and set before each change a system restore point !!!!
Remeasure boot, shutdown time, LatencyMon (3-5 min is enough for that), AS SSD.

Install for your mainboard: chipset driver, sata driver if needed, usb, network driver. Leave out all the other stuff like updaters and such stuff, fulfill purely the basic demands of what hardware you want to use on your board.

If you get issues already here then this would be strange ... concentrate troubleshooting at this point.
Because it won't bring any benefit when you install more and more.

If you install Win7 SP1 then this should be enough ... security patches are at this point IMHO not required.
If you want to install Windows upgrades, make sure you parametrize it to install only the securite upgrades,
not additionall add-on stuff .. you find references for this in google, its only one tick box ...

Current BIOS state:

C states - all disabled
SpeedStep - disabled
All network controllers - disabled
SATA express - disabled
Marvell SATA controller - disabled
Hyperthreading - enabled
Internal Sound - disabled

Im not sure where to disable M2 in the BIOS. Im running the latest BIOS (F8) from the Gigabyte site