Steinberg ur12
128 / input/output : 6,7ms / 7,7ms

Maudio
128 / input/output : 11.9ms / 5.67ms 

ufx+
128 / input/output : 3.42ms / 3.49ms 

All values given by ableton.

With these settings, the only issue i have is with the ufx, the audio is totally distorted with noise.
Both steinberg and maudio work fine.

Is it normal ?
Is it my system ?
I do not understand at all.

Also, prior to purchasing the rme, i was making a song, 6min about,
120 tracks, perhaps more than half of them are u-he zebra,
used 1024 buffer size with MAUDIO

I plug in my brand new ufx+, it cannot run the project with the same settings

Since then ive improved the bios settings, tried all possible tweaks i could find scattered across the internet.
And here I am.

My system is at least twice as fast as before the bios tweak.
I can run the project file at 512 with the MAUDIO.

HOWEVER, the ufx+ is STILL slower than both low end cards.

It is very disconcerting.
To what point do I need to improve the system so that the UFX+ outperforms the MAUDIO ?
Will it happen or will it not ?.......

I 've checked the results, and the RME is "quicker" in terms of latency(and that is to be expected).
So what I can figure out of these, is that the processor cannot handle that fast of a response, because it has to process too many things at any given moment.

Since you 've tuned DPC Latencies, perhaps the RAM is not so fast, or its latencies are not in tune with the processor.
With the Steinberg and the M-Audio the processor has more time to calculate and deliver, because of their higher RTLs.

Yes this obviously says it all. UFX is performing worse in high load scenarios. What is the cpu load with this project?
And on very light loads does the ufx then work at very low buffers or never?

How you know that there is lag ...

By observing kernel latency timer values.
Over 1000 = 1ms this is too high and a system is regarded as not suitable for audio.

The lesser this value is with best no high spikes / peak the better your CPU cores are able to quickly react on a workload. I mentioned already my usual and best values on my desktop system.
On a company laptop Lenovo T540 P around 6y ago I achieved kernel latency values around 50-60 us. For Internet browser I had to enable the separate nvidia GPU for this application otherwise I got audio dropouts at 32 samples ASIO buffersize because graphic use of a GPU which is integrated into CPU has higher prio for things like accessing the shared DRAM and this can cause issues.

idle system (freshly booted, just logged in, wait until all startup processes and services have loaded (no disk activity)).
No applications running or opened.
During test no movement of mouse pointer as this also generates interrupt's of mouse driver.

General tuning starts at BIOS level as here you define for your cpu certain parameters in terms of energy saving. The higher the sleep state that your cpu cores are allowed to enter the more time is required to wake them up until they are in useable state. This alone can take up to over 250 microseconds.

See this blog how I tuned my bios settings:
http://www.tonstudio-forum.de/blog/inde … -X10SRi-F/

Next steps are quite basic:
- disable not required devices in Bios
- install only required drivers, no Raid stuff etc
- avoid installation of  add-on tools from laptop vendor. Be it for upgrades or blower control or wifi add on GUI. Most of these tools are badly coded and often cause much lag by too high check frequencies and who knows what they are all doing
- then it can happen that certain versions of drivers can be bad. So learn to know your hardware and what LatencyMon values you usually get
- you need to take the high performance energy profile otherwise lag/dpc will be higher
- try to disable cpu core parking
- set windows to priorize background tasks

Disable / stop certain Windows background processes. Be it animated icons in the startmenue, file indexing, disk defragmentatin (for ssd a nogo anyway).

You can also stop certain not needed Windows services. Especially those related to collection of user data.

Use O&Owin10shutup to harden Windows which also will stop some of these data collectors.

Avoid in windows task planning that certain maintenance programs will be started automatically.

When installing open source software some installers install not wanted 3rd party tools, sometimes you can only see it when choosing advanced installation options.

If you got your PC preinstalled then perform a new installation on your own as all preinstallation that I saw so far are more or less crap.

If you have ssd then look too leave around 10% unpartitioned space at the end which is needed by most ssd controllers to have enough free cells for heavy io operation .

Basically avoid everything which generates unwanted background cpu load.

Norton and Kaspersky Internet security solutions are fine, careful with others they might suck to much performance.

I personally did remeasuring after every single change driver and software installation. Then you learn most about your system and can identify what causes isues and what not.

Yep all done !
Maudio and steinberg still faster under load than rme.

On your given scenario, yes. Because it depends on how the plugins are coded, how ableton utilises each core, usb speed, and what other tasks the i9 has to process.

About ram, the easiest is to increase ram speed, from let's say 2933MHz to 3200MHz and check! It's a trial and error task. If the value you set is optimal, everything is stable. If not, BSOD ! Simple as that.
The more difficult tweaking involves changing Latency timers, and a lot more trial and error checks!

Also, everything that Ramses said! EDIT: Just saw you 've, allready, did those!

Also, how to increase the ram speed, using a tuning software ?
I have xtu, but it actually increases DPC... so I am not using it

Your cpu is surely powerful for benchmarks like number crunching or computation. All of thede benchmarks or also appl cations like video rendering have one in common.
It's not time critical when data is being computed or when it arrives at disk.
Audio data has realtime requirements unlike other applications.
Forget about tuning ram, this doesn't bring anything useful.

I understand all the tweaks proposed, I’ve implemented them and I do under in a much deeper way by now the time sensitive nature of the problem. However, I am not talk in ABSOLUTE terms here, I am making a comparaison on the same system.

This:

Steinberg ur12
128 / input/output : 6,7ms / 7,7ms latency
Maudio
128 / input/output : 11.9ms / 5.67ms latency
ufx+
128 / input/output : 3.42ms / 3.49ms latency
All values given by ableton.

With these settings, the only issue i have is with the ufx+ ! the audio is totally distorted with noise.
Both steinberg and maudio work fine.

C states on or off
Tweak a b c, 1, 2 or 3,  x y z applied or not !
Systematically Maudio has the advantage here !

Why ?
Normal or not ?
Will rme eventually outperform Maudio
Just haven’t found the magic tweak yet ?

Those are clear cut questions I believe and there are only 3 possible answers, yes, no, I don’t know...

Here is a short thread with some Diva settings you could try:

https://www.gearspace.com/board/music-c … iller.html

I don't know if it still applies, because it's an old one, but worths a shot!

The Maudio runs more plugins than the rme.
It’s like saying to someone who bought a Ferrari, well if you don’t need to drive it fast, I don’t see the problem, it works.

At high buffer settings ? What would that test reveal ? I run more at 128 and at 1024...

“ MAudio's RTL is significantly larger, thus the CPU had more time to deliver.”
Rme does not need more than a core i9 to run at its full potential or close does it ?...

RTL is shorter okay.... so ?
I mean practically ?

Because what I experience is the possibility to run my projects at lower latencies with basic soundcards, than with the rme.
Or am I hallucinating ?

Did you look at the numbers I posted ?

Metalhead keys, there is this one project, 6 diva plugins
Maudio can handle it
Steinberg can handle it
RME FIREFACE UFX+ CANNOT HANDLE IT AT ANY BUFFER SIZE UP TO 1024.

RTL seems pretty academic at this point, don’t you think ?
At 1024 there is still audio issues !!
How does the rtl comes into play to explain that.
I’m probably missing something then

At the end of the day I just want to know if I am using the rme well below it’s potential because of the dell.
You believe so, others here are not pronouncing themselves, not giving a clear answer. So it is not reassuring.

Using 11 threads instead of 12 in LatencyMon I get 60 microseconds for the highest reported DPC routine execution time.
That is well below 1ms.
I have removed core 0 because there is a well known issue for Dell there, a driver with high DPC uses that core.
And when I say high, in my case it’s 0.6ms.. still below 1ms
And even with that core activated LatencyMon is happy with my system. Max dpc at, well, around 0.6ms.....

Ramses !
I am adding vsti on purpose for exactly that reason : to test what is possible with the rme !
I understand very well vstis, cpu load etc.... been using them for years. Like I said in a previous post I may have more than 50 running in a 120 track project. WITH THE MAUDIO.
and obviously the first thing I did was to plug the ufx and to my surprise it couldn’t handle the load.
Why rtl would affect that ? Forgive my ignorance but if I’m so puzzled maybe it’s because I’m missing some crucial bit of information. Rtl is lower which means that, shorter rtl card can’t handle as much as workload as larger rtl cards,
Then having a shirt rtl must have some benefits ?
So what needs to be done to help the short rtl cards achieve the same workload as the long rtl cards ?

To your other question: Cpu usage is roughly the same for both.
I have loaded the cpu ON purpose !
With the 6 very heavy virtual instruments, diva, cpu goes from 70 for large buffer sizes, up to 100 for smaller buffer sizes.
But to my astonishment (i am too ignorant to understand why, some say it’s the rtl... I don’t get it)
The rme cannot handle the project at any buffer sizes, and so, this is the price to pay to have short rtl ?
What do I get with this short rtl in return ?!

Ramses, all the data is here, but in picture format, it should be ok to open
https://drive.google.com/drive/folders/ … cbk7Y2Tekk

If the rtl is the input + output latency
Then for example, with:
MAUDIO at RTL = 17.6 ms
and UFX+ at RTL = 47.6 ms

AT THE SAME LOAD (heavy divas etc...)

Maudio plays is back perfectly
And ufx+ does not

Now I’m not sure if that the way to obtain the rtl, but if it is, can you explain ?
Ufx would have a larger rtl but still doesn’t produce clean audio ...

Yes, that is the way!
I must admit that this is the only test that has got me thinking, as to why it's happening. All the other tests add up, regarding RTL, buffer size, CPU load..

Ok but this is actually the only test that matters !
Because it’s practical , real world test...

Just now:
My heavy project/song plays back with the rme at 1024, that’s rtl=47.6ms
The Maudio plays it at 512, that’s rtl=35ms

This is the only test that matters, because it showcases the limits in possible loads.
The other tests only showcase the better hardware capability of the rme, it can simply go down to smaller rtls, yes of course.

But the real world application here, for me shows that I need to increase latency to playback my song, more than with the Maudio.

I want to stop annoying everyone here, but for that I need an explanation !

You 're not annoying anyone!!
We're all trying to figure out why this might be happening, and help you, and potentially others might be helped too!

Are these results with the "Limit ASIO to 32channels" option enabled on the RME?
If yes, I 'll need to think more on this!

I think we reached the point where further support is not possible via remote/forum.

Based on this information you should be able to understand what most likely the problem is: your DELL laptop and maybe also to a certain degree the installation of Win10.

All tuning tips had the intention to mitigate problems of your computer, but without guarantee that it can solve all problems of your system.

RME drivers are the best in the industry, but they can not fix issues deep inside of your machine.

Get a proper system for audio and your issues will go away.

BTW you seem to be back at the beginning ... seems you learned nothing from this thread. Comparing a 4 channel USB2 audio interface with slow usb2 transfer modes and only 4 channels is something different compared to a
Recording interface that has to transfer 188 channels using USB3 transfer modes.
For this your laptop is currently unable to deliver the required performance to do this sustained in time. If he would be able then you could also be happy customer with a system that has much better RTL.

Last attempt from my side ... Did you check already for a broken usb3 cable ? Do you see crc errors in the driver settings window (you need to keep this open for checks to run).

The only thing tp check is if the rtl of the m-audio and steinberg is true. There have been cases where it was double the reported. Then it would all make sense.

Contains interesting information regarding
- hard pagefault's
- that 2000us are only arbitrary value's for DPC
and ISR. the smaller the ASIO buffers are the smaller the sum of ISR and DPC needs to be.
- and some other interesting stuff

https://www.resplendence.com/latencymon_using

I assume only a careful new installation of Win 10 with remeasuring after each installation step,  with selected good drivers and leaving out bad stuff etc will bring insight up to what level the dell laptop will be useable for audio.

With 32gb of DRAM you could also try to remove the page file entirely to get rid of hard pagefault IF these should cause audio drops.

I think this is a job for an experienced person in this area who comes to right conclusions and actions. It's all not so easy and requires much skill and experience.

Its certainly easier and more straightforward (and not so frustrating / time consuming) to get a turnkey system optimized for audio.

It seems that no one can decide confidently if the shorter rtl (reported by live) for the Maudio vs rme (17.6 vs 47.6 ms) is NORMAL or NOT NORMAL.

Same for steinberg ur12 by the way ! It handles even shorter rtl than the Maudio !

Both manufactures are lying, yes maybe... ok then.
Can you be certain about this ?
That the rtl are cut in half for both ?
How to measure it ?

But even cut in half:
Again please integrate more carefully what I am showing :

17.6 ms vs 47.6 ms !
Please double 17.6ms and see if you can reach 47.6ms..........

“ the smaller the ASIO buffers are the smaller the sum of ISR and DPC needs to be.”
Why Maudio is ok then, even they cut the number in half to cheat ! It still produces clean audio and rme doesn’t !

Dell users complain when the LatencyMon shows a value way way upwards of 1ms for DPC, I’m at 0.06ms.

rme vs Maudio
Normal or not normal ??
You seem to say not normal.
Certain ?

Understand that if i purchase a pro workstation i want to be certain beyond a doubt that my stupid little test, rme (1 million ch) vs Maudio/steinberg (2 ch) does not give the same result at all in the new system !