HQPlayer4 EC modulator tips and techniques

[ted_b]

Jussi has produced new SDM modulators (known as EC modulators) in his newest HQPlayer version 4 releases.  The feedback on these VERY cpu-heavy modulators is nothing short of game-changing....for example folks are realizing they can only run much lower DSD bit rates (say DSD128 vs DSD512) and don't care...they sound better at 128 than the earlier ones at DSD512!

 

Anyway, the huge HQPlayer threads are getting bogged down in hardware discussions, being that folks are having a hard time finding the right combination of OS, cpu, and bios tweaks to get these things running without interruption.  So......

 

Let's discuss these issues here!  Please be specific about how you run them (what hardware and settings seem to make a difference), what OS and variant you are running, and any plain speaking bios and/or OS tweaks (turning off cores, overclocking, etc) you are using to get uninterrupted service. 

 

I am starting this thread for another reason, too:  I need a more powerful server environment than my current i7-6700k and while my audio room gets built I can do something hardware-wise to prepare to enjoy HQPlayer 4.  Thx

[ppy]

OK. I will repeat my post from another forum thread.

My configuration allows me to successfully play PCM48-192 -> DSD256, ASDM7EC, poly-sinc-ext2

Roon server + HQPlayer Embedded 4.11.1 on Debian 10 (minimal install)

MB ASUS Prime H310I-Plus

CPU i5-9600K

RAM Hyperx HX426C15BK2/8

In the BIOS:

the parameter is set - "Sync All Cores"

"Intel(R) SpeedStep(tm)" - disabled

 

[Miska]

My hardware documented here:

 

It can do DSD256 with the EC modulators. Multicore set to "auto".

 

Embedded on Ubuntu Server 18.04.

[dean70]

Core I5 2500k overclocked to 4.6ghz can run 44.1k  poly-sync-ext2 ASDM7EC DSD128 (integer multiple) w/o CUDA. Have GTX970 & now updated drives to run CUDA. Should do DSD128 (48k) after offloading convolution to gpu

 

Running Windows 10 with all the bloated crapware running 😁

[ericuco]

Ubuntu Studio 18.X running HQP Desktop & Roon Core on i5 with 4 cores, sorry not sure other hardware specs.

 

I have been using the ASDM7+f512 modulars at DSD256 rate. All four CPU's running about 50%.

 

With the new EC modulators, I have to move down to DSD128 with the DSD5EC & ASDM5EC, CPU's running about 60 - 65%. With ASDM7EC, about 70 - 75% with no stuttering.

 

Tried DSD5EC at DSD256 rate but got lots of stuttering. CPU's around 80 - 85%.

[rando]

If I might interject. 

 

This subject gained speed due to @JTS mentioning a fairly standard tune of his i9-9900k video editing server resulted in considerably better results than those with very similar hardware were claiming.  Not wanting to go completely off topic, we quickly established the nature of changes he effected.  

 

Based on a very slim sample of respondents it was clear

7 hours ago, ted_b said:

the huge HQPlayer threads are getting bogged down in hardware discussions, being that folks are having a hard time finding the right combination of OS, cpu, and bios tweaks to get these things running without interruption.  So......

 

So a good place to start might be establishing a more in depth picture of what a fairly standard BIOS profile for a video editing server resembles that matches with the small snippet of details he provided.  (I can toss off some links or more direct info later.  Puget Systems articles would be a good place to dig around in the meantime.)

 

 

So far as I have seen, and I did just recheck, this is most informative description of Intel BIOS and other settings impacting HQP EC modulators for the better.  Before choosing a direction to move conversation forward it's probably good to reiterate what I mentioned in the HQP thread.  

 

4 hours ago, rando said:

There seems to be a sizeable group unaware of how to tune their gaming optimized mb's in general.  Without the strain of EC modulators making clear wasted potential. 

 

You could easily do serious harm if you don't know what you are in doing in BIOS or an ASUS/Gigabyte/etc desktop program.  There is a very specific balance that needs to be maintained and there can be a lot of testing after rechecking to assure stability if you starting pushing the envelope.  Establishing a set of terminology across manufacturer BIOS, tackling the utility and connection to other parts and the whole of each setting, and very importantly what if any gaming derived calculator or process is trustworthy or even relevant is going to be big and messy.  People are going to need to keep their cool and accept setbacks.  Or else I really doubt the mostly silent members who would hold the greatest impact on this being a fruitful endeavor will fail to appear.

 

Were there a pile of current top shelf consumer hardware at my disposal I'd have dug in and then reported back if largely unrecognized performance with certain chips or combinations of hardware were feasible.  The idea of putting a bunch of heads together with highly evolved HQP machines sounds very interesting.  Quite often an older processor will reveal a highly sought after quality that allows it to keep in the fight.  Until the six core 8000 series Intel were released it was possible to throttle the snot out of an i5-2400 system (similar to bolting a whole ton of parts on a bulletproof engine block) and get acceptable performance.  Anyways, I might revisit the idea of getting an Asus ROG Maximus XI Gene mATX with double-capacity DIMM if it looks like that might be a worthwhile experiment.  🥳

 

 

[k6davis]

Core i7-9700K. It seems to like EC modulators. 😀

 

It will easily upsample PCM 24-192 => EC DSD256 x 48 with poly-sync-ext2. With any OS. Without a GPU. Without high CPU utilization. No overclocking or special settings or special cooling required. 

[ted_b]

17 minutes ago, k6davis said:

Core i7-9700K. It seems to like EC modulators. 😀

 

It will easily upsample PCM 24-192 => EC DSD256 x 48 with poly-sync-ext2. With any OS. Without a GPU. Without high CPU utilization. No overclocking or special settings or special cooling required. 

What mobo?

[k6davis]

Just now, ted_b said:

What mobo?

 

I'm using the ASRock Z390 Phantom Gaming ITX (review), because it's the most highly regarded ITX board. 

[JTS]

Puget is a great place to start for solid builds. They test the hell out of stuff for video production work.

 

I am running W10 Pro on an i9-9900k on an ASUS WS Z390 PRO with 64Gb ram and a GTX1080. I have since simplified my overclock even further than previously mentioned on the HQ player thread.

 

I leave Turbo on and run all the cores at 4.7GHz (I found that the voltage required for 4.8 ran things a touch too hot for my tastes when rendering video). I run a power profile that allows the CPUs to throttle down when not in use (I used the balance power profile as a starting point). I have a middling GTX 1080 in there and can run the EC modulators with any of the filters in DSD256. It sounds glorious through my RME-ADI2 DAC in DSD direct (->Freya->Neumann KH120As)

 

In my experience, overclocking does not have to be super complex. I approach it as simply as I can. Don't take this as instructional, but as an example - there are many ways up the mountain.

 

Many people get into overclocking their cache and their ram etc. Across multiple machines (Like 5 that I run in the studio for rendering and animation work) I have found that research is the most important part - The big questions are what are other people running their cpu at and at what voltage with what cooler???

 

I have three machines in 2U cases for rendering that I have actually undervolted a bit to get temperatures suitable for the coolers that fit in the 2U cases. One thing that you will find in your research is that the defaults on a given MOBO are often inefficient. They dump gobs of power into your CPU which makes a bunch of heat without any return on speed increases.

 

For the overclocking process, I start by using default voltage and settings with everything except the CPU. Then I go through a process of balancing speed vs voltage vs. tempuratures. I have used this process to pretty good success over a number of years.

 

I cannot stress enough the value of using a great cooler. I don't run water as I have found loud and way too fiddly and unreliable. I run Noctua coolers. My i9-9900k uses the big 15" bastard with dual fans. (I might add that my workstation is in a quiet-ish isolation cabinet 8' from my desk.)

Here is how I (rather crudely) approach overclocking. I anticipate that there will be others with more refined methodologies that may chime in. Again I offer this up as an example rather than instruction.

 

My goal is always stable then fast then cool as I use my machines principally in a busy work environment.

 

- Check out the overclocking recommendations of your particular MOBO manufacturer. I mostly use ASUS and they have excellent tutorials on overclocking. Research research.

 

- Check out the thermal limits of your CPU on your manufacturers website. You want to be well under this as you work. For example, the 9900k limit is 100C. It's a hot CPU to begin with. I work so that full load temperatures are around 80-85 and any temperature spikes don't go far beyond 90.

 

- get a cpu monitoring software package like HWmonitor so you can see the temperatures of things as you work.

 

- I use the bios setting, not software settings in a package like Intel extreme tuning or Asus tuning utilities. It required more rebooting, but I like working with the bios.

 

- Set a target speed based on research

- Set a cpu voltage based on research

 

- Perform a reasonable stability test (AIDA or intel for example run it for a while - 15 minutes or so so everything has an opportunity to come up to temp)

- If you crash then increase voltage by .01 and repeat

- If it does not crash and I like the temps, and I want more, I'll go higher with the clock and the voltage and repeat.

- Once I feel like I'm in a good spot I move on to harder stress tests

 

- I like Prime 95 !!!WITH AVX TURNED OFF!!! (AVX will heat up your machine to the extreme - see their docs on how to turn AVX off)

- remember that prime 95 is the extreme of what your box will need to deal with...I have NEVER seen any temps like it in my video work, even when CPU rendering CGI.

- If crash then increase voltage by .01 if your temps are ok or decrease the clock by one step and repeat

- Once I have it balanced, I will let prime 95 run overnight. If it crashes, I run through the steps again.

 

Then...use the machine a bunch. You may get a BSOD when doing a bunch of stuff at the same time and you have a choice here, reduce clock speed or increase voltage.

 

There are many concerns about overclocking that you will find online. Remember that there are gobs of gamers doing this all the time and that your CPU will throttle down if you hit the thermal limit to avoid damage. In my opinion, if you leave all the safety mechanisms on in the bios (which are the default) you would need to work hard to damage your CPU.

 

The greater concern is stability. It can take weeks to get it all dialed in so that you're system is rock solid. I run my machines 24/7. My old, but still killer, overclocked (4.2Ghz on a tiny Nocuta in a 4U case) 5960x is almost 4 years old and is happy as can be.

 

In any event, I think for a chip like the 9900k, allowing all the cores to operate simultaneously without any time limit at 4.7 or 4.8 makes a huge increase in performance over what the original spec allows. My overclock is a gentle one by many standards, but it is enough to give me the performance boost and stability I want.

 

It rocks as a production machine and, as I mentioned, it runs any EC modulator with any filter at DSD256 without any complaints.

[joelha]

While I have HQPlayer, I've mothballed my servers in favor of using Roon on my Synology NAS.

 

Rather than have to set up a server again, I'm wondering if Miska has sample clips of any tracks employing EC modulator upsampled to say DSD128.

 

Hearing that compared with the original track would be a quick, easy way for me (and perhaps others) to get a sense of what the latest from HQPlayer sounds like.

 

Joel

 

 

[ted_b]

Joel,

A free trial of HQPlayer Pro would allow you to record offline that yourself...the only downside is 60 seconds max per sample.

[rando]

Getting confirmation dropouts while playing w/EC modulators were from repeated thermal throttling, among other worrisome oblique recognitions, make me cautious about relaying info.  The majority here are probably nearer the opposite end of the spectrum, ie weighing (or soon will be) 1% improvements out of impeccably detailed comprehension and looking for firmware updates to fix things further.

 

Back to that initial concern, heat management with active cooling.  Thermal paste and coolers (CPU & GPU) really need to be sorted out early and often(!) when you are at 80%-90% of TDP on +100w processors regularly.  Surface prep and awareness of paste conductivity!  So much more.  This really is a huge determinant, but we're in the software forum.  

 

@JTS overview hit a lot of key points.  Expand all that to encompass RAM (DIMM and the xGB inside video cards) as well as GPU processor and the full picture starts coming into view.  Restraint is a noble quality.  Meaning knowing well enough to leave options alone because in a vast number of cases they are... kludges to safe fail against gamers or provide for a dire need of some incessant bugger.  

 

Please anyone who has spent years streamlining these aspects of their fanless music pc.  Speak up.  Refinement is where the interesting individual solutions that will help here start popping up.  That and willingness to try turning your Threadripper into a middling horsepower chip like @Hammer  Who grasped the spirit of experimenting in directions you might not like learning from.

[ted_b]

13 minutes ago, rando said:

The majority here are probably nearer the opposite end of the spectrum, ie weighing (or soon will be) 1% improvements out of impeccably detailed comprehension and looking for firmware updates to fix things further.

Rando, your point here is over my head...sorry.  I am not getting it.  All I know is that "the majority HERE", on this thread, are hopefully either relaying their real life experiences with setting up an HQPlayer EC-capable server, or asking questions about same.  Whether they are also dealing with "impeccably detailed comprehension" (?) or looking for firmware fixes is, well, irrelevant to this...I think.  You say "restraint is a noble quality".  Um, ok.  Not sure, again, what that has to do with this thread.  Sorry if I'm being too thick.

[kelvinwsy]

In my situation, my 6700k is running on WinServer 2019 Core with Audiophile Optimizer 3.0. Overclocking is off! so is other Bios speed up settings. I find the music more serene and very very low on digital sound.

Of course if to run EC modulators on the 6700k requires turbo and other speed up/multicore modes, I am willing to try.

But fan noise I have an issue with as the server is within 6 feet of my HP listening position. Abyss AB1266 and HD800's are pretty open to external noises. I can hear the incessant hum of my AC compressor but I set this to ECON and higher temps say 26deg C and it only comes in an odd intervals.

Running the 6700k at 70-80 CPU utilisation rate invariably equals more fan noise. There may be gains to the HQplayer 4.0 SQ, but I have gained a lot on using resistors to detune all the fans in my PC, moving the additional fans onto a separate fan header etc and using fan power filters.

A bit hesitant to reverse direction. 

Eventually the 6700k (3 years coming on) will be obsolete. So biting the bullet with a 9900k and 2080 Super may be the way to get silent HQPlayer 4.0 server up and running without incessant fan noise.

Stressing the PSU (i use the supremely low distortion Corsair RMX550 - 8mv at the 200watt setting - its fan does not run) with this Turbo/speed up mode may not be optimal. I think i will have to swop in my Corsair HK850 in this case. Ummh... decisions, decisions.

Will try it this weekend. 

 

[rando]

Effectively I was buoying up the tech sense of members on AS "the majority HERE" while cautioning there can be risks tinkering with BIOS and/or settings.  What is vague or confusing was aimed at not alienating those who'd benefit from clicking around the linked articles page.  Multiple members have related really giving thermal protection a thorough drubbing in the main HQP threads.

 

If not being aware of poor stock settings is problematic.  Advising restraint above all else has been central to what JTS and I want to get across when beginning fixing them.  Small steps, slow and attentive, stability!  

 

 

 

 

13 hours ago, ted_b said:

folks are having a hard time finding the right combination of OS, cpu, and bios tweaks

 

13 hours ago, ted_b said:

Please be specific about how you run them (what hardware and settings seem to make a difference), what OS and variant you are running, and any plain speaking bios and/or OS tweaks (turning off cores, overclocking, etc)

 

@ted_b  You know more than you are letting on.  Why the Monaco server' gaming motherboard required commissioning a custom BIOS for example.  This topic is either bumping against that ceiling or hardly aware of it from far below.  Suggest another means of expressing "impeccably detailed comprehension" that makes sense if that one bothers you.  

[k6davis]

24 minutes ago, kelvinwsy said:

Eventually the 6700k (3 years coming on) will be obsolete. So biting the bullet with a 9900k and 2080 Super may be the way to get silent HQPlayer 4.0 server up and running without incessant fan noise.

 

I don't see the point in the 9900k for HQP. The 9700K will run every modulator with ease, costs less and runs cooler.

[jamesg11]

Thanks Ted, for this thread to focus on what clearly seems a major sonic step-up from Miska.

Hopefully, there will be rapid use of the info to enable off-the-shelf players from the likes of small green computer etc - for those of us who don’t tinker much with hardware et al.

[emc_1984]

I can play 44.1-192k -> constant DSD256 11.2896Mhz with polysinc-xtr-2s & ext2 

OS: Windows Server 2019/Power plan: perfomance mode,  unusing services killed(DiagTrack, Print Spool et)

i5-8600T ( all cores in Turbo Boost 3.5Ghz)

Palit GTX 1050Ti

3200Mhz CL14 RAM

BIOS:  diable C-States , Virtualisation disabled

HQP Setting: Buffer 100-250mS, Multicore : Auto

I try this configuration with Ubuntu Server 18.04  HQP E and it shutter.

[Ales Prochazka]

HQPlayer Embedded

• Ubuntu Server 18.04.2 LTS (4.18.0-25-generic)
• MB GA-Z270-HD3P
• CPU i7-7700K (overclocked at 5GHz)
• 8GB RAM
• without CUDA

1. poly-sinc-ext2 + ASDM7EC + convolution ON: PCM 44,1 to 352,8kHz ---> SDM256 OK
2. poly-sinc-short-mp[lp] + ASDM7EC + convolution ON: PCM 44,1 to 352,8kHz ---> SDM256 OK
3. poly-sinc-xtr-mp[lp]-2s + ASDM7EC + convolution ON: PCM 44,1 to 352,8kHz ---> SDM256 OK

[kelvinwsy]

https://cpu.userbenchmark.com/Compare/Intel-Core-i9-9900K-vs-Intel-Core-i7-9700K/4028vs4030

I find from the above they run at the same TDP but the 9900k is 32-38% faster on Multicore Integer and Floating Point Computation speed. And it is only abouit 130USD higher in price. Might as well go to the 8 core 16 thread 9900k

 

[k6davis]

1 hour ago, kelvinwsy said:

https://cpu.userbenchmark.com/Compare/Intel-Core-i9-9900K-vs-Intel-Core-i7-9700K/4028vs4030

I find from the above they run at the same TDP but the 9900k is 32-38% faster on Multicore Integer and Floating Point Computation speed. And it is only abouit 130USD higher in price. Might as well go to the 8 core 16 thread 9900k

 

When it comes to HQP, there's nothing the 9900K can do that the 9700k can't. When it's time to upgrade for EC DSD512, they'll both be paperweights.

[asdf1000]

9 hours ago, k6davis said:

Core i7-9700K. It seems to like EC modulators. 😀

 

It will easily upsample PCM 24-192 => EC DSD256 x 48 with poly-sync-ext2. With any OS. Without a GPU. Without high CPU utilization. No overclocking or special settings or special cooling required. 

 

So the i7-9700 (non-K) should be fine then, if overclocking the i7-9700K is not required?

 

[asdf1000]

7 hours ago, JTS said:

In any event, I think for a chip like the 9900k, allowing all the cores to operate simultaneously without any time limit at 4.7 or 4.8 makes a huge increase in performance over what the original spec allows. My overclock is a gentle one by many standards, but it is enough to give me the performance boost and stability I want.

 

Can your i9-9900K actually do PCM 24-192 => EC DSD256 x 48 with poly-sync-ext2, without overclocking?

 

Or you haven't tried because you need your machine overclocked for your work?

[asdf1000]

On 8/4/2019 at 11:01 PM, k6davis said:

 

Actually, I'm not sure about that. 'K' basically means you can overclock the processor. The i7-9700K can do full EC DSD256 without overclocking, so the non-'K' variant should be able to as well.

 

It didn't cost much more to have the 'K', so I went for it anyway.

 

Isn't the "base" frequency of the 'K' variants higher?

 

In the case of i7-9700 vs 9700K its 3.00 vs 3.60 GHz base ?

 

This higher base frequency doesn't make a difference?

 

So if you went for the i7-9700 instead of 'K', there is a chance it wouldn't do what you can do now?