Best CPU for hqplayer

[Quadman]

16 minutes ago, Sevenfeet said:

The FlightSim rig cannot upsample PCM to DSD512 on the hardest filters (like poly-sync-xtr)

Back on page 2 of this thread Miska posts a screen shot of his 6950x rig running DSD512 with the poly-sinc-xtr filter and CPU utilization is around upper 30%.  My guess is you have cuda offload check marked in HQP settings, Uncheck it or grey box it and you should be able to do DSD512 on XTR filter.  The GPU is not fast enough to handle this task and is most likely causing the stuttering at 512 with full XTR.

 

1 hour ago, sbenyo said:

I decided to invest in it both for audio and for other uses so the system is not idle when I am not listening to music. I think that if there are other uses t can justify the investment.

 

 Do you use a dual boot system with your PC?  When I build audio PC's, I run some of the scripts found on this forum and use Audiophile optimizer to reduce windows processes, essentially making the PC audio only.  The SQ jump you can get doing this will blow away XTR vs XTR 2s sound comparisons.  Look into dual boot so you can still use PC for non-audio and then use other boot, tweaked to reduce windows process for music only.  You will be very happy with that decision.

[shadowlight]

2 hours ago, Quadman said:

One disadvantage of the AMD CPU line is no built in internal graphics, you need a GPU to get video. An inexpensive video card can accomplish this just fine.  Intel comes with internal graphics so no external GPU is needed.  I value L3 cache so much that this did not sway me from AMD.

 

Is this really a disadvantage if you are using the embedded version?  For the initial build you can leverage usb serial cable and terminal program.

[sbenyo]

I am not using dual boot for now. It sounds like a good option.

How are you using it? Do you have two separate installations of Windows one regular and one optimized?

 

I am using Fidelizer with max settings. It supposed to do similar optimizations as AO. Maybe it's not the same and it's worth trying AO.

I did not see any issue using it for both Audio and other applications so far so I am not sure if dual boot is really necessary.

 

 

 

 

[Quadman]

6 minutes ago, shadowlight said:

For the initial build you can leverage usb serial cable and terminal program.

Is the average person really going to do that?  It's only a disadvantage if you get the CPU and expect graphics out of the box like intel.

[Quadman]

13 minutes ago, sbenyo said:

I am not using dual boot for now. It sounds like a good option.

Currently I do not use dual boot for audio PC as I only use it 100% for audio.  If I did I would actually use 2 small SLC SSD's (32 or 64GB) with windows OS on each (then externally power each SSD with linear PSU's).  One for Normal use and one highly optimized for audio only, I would add easyBCD to the normal SSD which makes it super easy to dual boot.

 

I have run AF for many years currently running 8 pro, it is an excellent program and runs well with AO.  AO and AF together is another very nice step up in sound quality over just AF.  I just set up another SSD with a clone of my preAO SSD, it has AF on it and I want to compare the SQ differences between no AO vs with AO.  

 

If you run AO and set maximum optimizations then you will have to do dual boot if you plan on using PC for normal stuff.

[Sevenfeet]

41 minutes ago, Quadman said:

Back on page 2 of this thread Miska posts a screen shot of his 6950x rig running DSD512 with the poly-sinc-xtr filter and CPU utilization is around upper 30%.  My guess is you have cuda offload check marked in HQP settings, Uncheck it or grey box it and you should be able to do DSD512 on XTR filter.  The GPU is not fast enough to handle this task and is most likely causing the stuttering at 512 with full XTR.

 

 

Miska is running 10 cores (20 multithreaded).  I'm doing 6/12.  I'm not able to do it with this machine (CUDA or not).  The extra cores may be key in his situation.

[bibo01]

Please allow me to insist.

How come nobody seems interested in delid? Has anyone tried it?

[Quadman]

30 minutes ago, bibo01 said:

How come nobody seems interested in delid? Has anyone tried it?

If your temperatures of the CPU you use are fine under HQP 512 load and run at or below 50C (or so) then there is no reason to delid.  Delidding is primarily designed for extreme overclockers who want to push their CPU's to the highest limits possible.  Also AMD, at least for 7 and thread ripper, solders their heat spreader to the chip making delidding very risky.  Intel's CPU generally use thermal compound between silicon chip and heat spreader and are much easier to delid.  Here's a nice article on SAFE delidding (don't try with AMD).

 

https://www.pcgamer.com/meet-the-overclocker-who-made-delidding-your-cpu-idiot-proof/

[n2it]

7 hours ago, Miska said:

If it runs fast enough already, you just gain significantly more heat by overclocking. So maybe not worth if you don't need it...

 

 

Agree it will create more heat (i.e. the memory will run at a higher voltage), but it shouldn't add (much) to the CPU heat - except a little bit from the proximity to the CPU. 

 

7 hours ago, sbenyo said:

Does anyone know if it's recommended or beneficial to OC only memory from 2400Mhz to 3200Mhz keeping the CPU at base clock?

My DDR4 is currently running at 2400Mhz and supports 3200Mhz. I wonder if I can gain anything from it.

If 3200Mhz is not the optimal setting, is there a "sweet spot" frequency?

 

 

 

Given that you can control the heat, faster memory will mean faster memory access (and less CPU wait time).  A threadripper - which in reality is like running dual CPUs - the faster memory will help with latency when the L3 fills and the core needs to access the memory on either of the 2 far cores.  The faster the memory, the faster the access to the far memory. 

 

That being said, I don't know what exact constraints are, so you might experiment to see if this actually reduces CPU utilization by reducing the memory wait times after using up the L3 cache.  

 

@sbenyo I assume you are running the memory in quad channel?  Also - I assume you are in "creator mode" and not "game mode". 

 

[shadowlight]

3 hours ago, Quadman said:

Is the average person really going to do that?  It's only a disadvantage if you get the CPU and expect graphics out of the box like intel.

 

Just another option if you go down the AMD route and do not want to spend the extra 50 bucks for cheap video card when using with embedded version.

[bibo01]

2 hours ago, Quadman said:

If your temperatures of the CPU you use are fine under HQP 512 load and run at or below 50C (or so) then there is no reason to delid.  Delidding is primarily designed for extreme overclockers who want to push their CPU's to the highest limits possible.  Also AMD, at least for 7 and thread ripper, solders their heat spreader to the chip making delidding very risky.  Intel's CPU generally use thermal compound between silicon chip and heat spreader and are much easier to delid.  Here's a nice article on SAFE delidding (don't try with AMD).

 

https://www.pcgamer.com/meet-the-overclocker-who-made-delidding-your-cpu-idiot-proof/

I agree about Intel vs AMD for delid. In my experience of PC playback with HQPlayer there is always a need - sometimes "potential" - for sqeezing the most out of a CPU, because one sooner or later wants to try DSD512, convolution, multichannel or a combination of either of them

[Miska]

5 hours ago, n2it said:

Agree it will create more heat (i.e. the memory will run at a higher voltage), but it shouldn't add (much) to the CPU heat - except a

 

Sure, but of course the RAM module will run hotter and it adds to total heat bill of the case (just like everything else too). For example my Xeon workstation manual says that if all 8 memory slots are populated, extra RAM fan needs to be added to ensure enough ventilation in order to maintain stability under all conditions and ensure component lifetime. So observing RAM module temperatures is useful and depends a lot on overall computer case air flow, because there's usually not dedicated fan for RAM modules (but such of course can be added).

 

5 hours ago, n2it said:

Given that you can control the heat, faster memory will mean faster memory access (and less CPU wait time).  A threadripper - which in reality is like running dual CPUs - the faster memory will help with latency when the L3 fills and the core needs to access the memory on either of the 2 far cores.  The faster the memory, the faster the access to the far memory. 

 

That being said, I don't know what exact constraints are, so you might experiment to see if this actually reduces CPU utilization by reducing the memory wait times after using up the L3 cache.

 

HQPlayer is quite heavy on memory access and speed of memory is critical, but OTOH, if things are fast enough already, pumping more speed out of CPU and RAM will usually just increase heat production. Wait times usually don't heat up the CPU, because it is idling.

 

When things are quite not running fast enough, then getting faster RAM and running it at higher speed may help. But compared to caches, it is slow, so amount of CPU cache is still very critical.

 

[sbenyo]

8 hours ago, n2it said:

 

Agree it will create more heat (i.e. the memory will run at a higher voltage), but it shouldn't add (much) to the CPU heat - except a little bit from the proximity to the CPU. 

 

 

Given that you can control the heat, faster memory will mean faster memory access (and less CPU wait time).  A threadripper - which in reality is like running dual CPUs - the faster memory will help with latency when the L3 fills and the core needs to access the memory on either of the 2 far cores.  The faster the memory, the faster the access to the far memory. 

 

That being said, I don't know what exact constraints are, so you might experiment to see if this actually reduces CPU utilization by reducing the memory wait times after using up the L3 cache.  

 

@sbenyo I assume you are running the memory in quad channel?  Also - I assume you are in "creator mode" and not "game mode". 

 

I run in quad mode. I have 4x8GB. I also have MSI x399 which has x-boot utility or changing modes. I use game mode. I think it's what you ment. I don't see any difference between game mode or other modes regarding CPU utilization or temperature.

 

I will have to play a bit more with it to see how memory speed between 2400-3200 is affecting temperature and if there is a "sweet spot".

I understand the impact of it is not significant but if it's possible to gain something and still be ~50c or below it's good.

 

BTW, I don't see too much effect of the Noctua CPU fans speed. I will try to stop them completely and see if temperature is affected but from what I could see having them at low or high RPM does not affect the CPU temperature too much.

[astrotoy]

Not being as knowledgable as those on the thread, I had my computer builder determine the components for my PC.  He consulted with Jussi/Mishka and chose the i7-6950X.  I wanted to be able to upconvert mostly DSD64 and a few PCM192/24 albums to DSD256 for 5.1 (6 channels) using HQP.  He recommended Roon with HQP for which I am very happy.  The mch conversion works great with the processor, but only with the poly-sinc-2s filter. I haven't tried any similar filters, except for closed form, which also works fine.

 

For my stereo files I did a lot of experimenting and really like the sound of the -xtr filter without anything else.  It absolutely will not work upconverting and decoding mch, but works flawlessly, as Jussi/Mishka reported earlier, with stereo files.   I have had the computer for more than 10 months.  Back in March of 2017, my computer builder bought the processor for $1650, and I thought that with Moore's Law still in force, that the chip would be a lot cheaper.  But looking on Amazon it is just about the same price as a year ago. Has congress repealed Moore's law? 

 

Larry

[salaryman]

Hi, I'm not a newbie to this forum but this is my first post.

 

Just want to let you know my testing on using some old used Xeon to play around hqp DSD512 with poly_sinc_xtr_mp.

The test run with new instsalled Win10 Pro with no optimizers yet,  to a winsvr 2016 NUC NAA and T+A DAC8 DSD.

 

1) Xeon e5-2670 C1 (8C16T) + X79 + DDR1333 ECC 4G x4

2s okay, non-2s not ok

 

2) Xeon e5-2660 v2 (10C20T) + X79 + DDR1333 ECC 4G x4

2s okay, non-2s not ok

 

3) Xeon e5-2683 v3 (14C28T) + X99 + DDR2133 8G x2

non-2s ok

 

Set #3 would be a cheap alternative which cost about 35% of new TR set.

[sbenyo]

Quote

 

Nice. It is very interesting to see the Xeon tests. If it works it probably wins for the cheapest option.

Processor itself it not that cheap but I guess everything else is.

 

[louisxiawei]

On 3/10/2018 at 12:11 AM, salaryman said:

Just want to let you know my testing on using some old used Xeon to play around hqp DSD512 with poly_sinc_xtr_mp

 

3) Xeon e5-2683 v3 (14C28T) + X99 + DDR2133 8G x2

non-2s ok

 

Set #3 would be a cheap alternative which cost about 35% of new TR set.

Thanks for the post. I'm just wondering which music source/file you upsampled to DSD512 using polysinc-xtr-mp? Can you confirm whether it is 44.1/16 redbook file since the heaviest upsampling load is always to upsample 44.1/16 file to 44.1 x 512.

 

Although my 18-core i9 7980XE can upsample 44.1/16 to 44.1x512 with poly-sinc-xtr file without non-2s, the CPU utilization is around 45%,  clock speed is around 3.5 GHz.

 

The 14-core e5-2683 only has 3.0 GHz of max turbo frequency, I doubt a bit regarding its capability for the upsampling performance. If you have a solid confirm of such setting, then you will save many peoples' wallet who wish to apply poly-sinc-xtr filter at DSD512.

[Miska]

6 hours ago, louisxiawei said:

The 14-core e5-2683 only has 3.0 GHz of max turbo frequency, I doubt a bit regarding its capability for the upsampling performance. If you have a solid confirm of such setting, then you will save many peoples' wallet who wish to apply poly-sinc-xtr filter at DSD512.

 

Turbo frequencies don't really matter much when you have all cores loaded, because in those cases turbo cannot be really used. Xeons have smaller difference between max turbo and base frequency. Advantage of 2683 is it's 35 MB cache, 10 MB more than my 6950X.

 

But interesting that it works with just two memory modules, because 2011-3 socket has four memory channels and in this case only half of those populated which halves the memory bandwidth to RAM. Probably covered up by the much larger cache.

 

[louisxiawei]

1 hour ago, Miska said:

 

Turbo frequencies don't really matter much when you have all cores loaded, because in those cases turbo cannot be really used. Xeons have smaller difference between max turbo and base frequency. Advantage of 2683 is it's 35 MB cache, 10 MB more than my 6950X.

 

But interesting that it works with just two memory modules, because 2011-3 socket has four memory channels and in this case only half of those populated which halves the memory bandwidth to RAM. Probably covered up by the much larger cache.

 

But 2683 has only 2.00 GHz base clock, don't you think it's a bit low for task like poly-sinc-xtr at DSD512? My 7980 never runs below 3.5 Ghz with poly-sinc-xtr at DSD512.  Or are you suggesting that the 10 MB more cache size can compensate the base clock frequency? 

 

One thing I still don't understand is that I tried to upsample DXD file (352.8) to 48x512 with poly-sinc-xtr. The load should be significantly lower than upsampling from redbook file, but why is this setting still causing stutter? Which part cause the limitation? The cache of my 7980XE or the speed of the DDR4 3466 64GB RAM? 

[2a3set]

On 2018-03-09 at 11:11 AM, salaryman said:

Hi, I'm not a newbie to this forum but this is my first post.

 

Just want to let you know my testing on using some old used Xeon to play around hqp DSD512 with poly_sinc_xtr_mp.

The test run with new instsalled Win10 Pro with no optimizers yet,  to a winsvr 2016 NUC NAA and T+A DAC8 DSD.

 

1) Xeon e5-2670 C1 (8C16T) + X79 + DDR1333 ECC 4G x4

2s okay, non-2s not ok

 

2) Xeon e5-2660 v2 (10C20T) + X79 + DDR1333 ECC 4G x4

2s okay, non-2s not ok

 

3) Xeon e5-2683 v3 (14C28T) + X99 + DDR2133 8G x2

non-2s ok

 

Set #3 would be a cheap alternative which cost about 35% of new TR set.

That’s very nice, I couldn’t make it with e5-2680v2 (10c 2.8g with performance mode in kernel it actually runs 3.1g mostly) with non-2s/512, non-2s/256 all good, thinking about adding a tesla m2090 for  assisstence as low cost alternative,

 

I am using ubuntu 16 with naa on 2012r2+ao.

[lmitche]

38 minutes ago, louisxiawei said:

But 2683 has only 2.00 GHz base clock, don't you think it's a bit low for task like poly-sinc-xtr at DSD512? My 7980 never runs below 3.5 Ghz with poly-sinc-xtr at DSD512.  Or are you suggesting that the 10 MB more cache size can compensate the base clock frequency? 

 

One thing I still don't understand is that I tried to upsample DXD file (352.8) to 48x512 with poly-sinc-xtr. The load should be significantly lower than upsampling from redbook file, but why is this setting still causing stutter? Which part cause the limitation? The cache of my 7980XE or the speed of the DDR4 3466 64GB RAM? 

You are changing base rate going from 352.8 to 48x512. How does 384 to 48x512 do?

[louisxiawei]

8 minutes ago, lmitche said:

You are changing base rate going from 352.8 to 48x512. How does 384 to 48x512 do?

48K to 48x512 has no problem. I know changing base rate is more demanding on the cpu load.

 

I can do 44.1 to 48x512 smoothly at any -2s filters and my CPU can even do 44.1 to 48x512 at poly-sinc smoothly without stutter, but initializing time is quite long.

 

I assume it must be related to the character of various filters, but from my point of understanding, hi-res files upsampling should be significantly lower than redbook file.

 

 

[lmitche]

6 minutes ago, louisxiawei said:

48K to 48x512 has no problem. I know changing base rate is more demanding on the cpu load.

 

I can do 44.1 to 48x512 smoothly at any -2s filters and my CPU can even do 44.1 to 48x512 at poly-sinc smoothly without stutter, but initializing time is quite long.

 

I assume it must be related to the character of various filters, but from my point of understanding, hi-res files upsampling should be significantly lower than redbook file.

 

 

Jussi will know the answer, but my guess is that every point on the curve needs to be "synthesized" when base rates are changed, where with the same rate family "interpolation" is less compute bound.

[Miska]

1 hour ago, louisxiawei said:

But 2683 has only 2.00 GHz base clock, don't you think it's a bit low for task like poly-sinc-xtr at DSD512? My 7980 never runs below 3.5 Ghz with poly-sinc-xtr at DSD512.  Or are you suggesting that the 10 MB more cache size can compensate the base clock frequency?  

 

If the clock rate is fast enough to run the modulator then it's enough.

 

For the filter, clock frequency doesn't matter if there are more cores and enough memory bandwidth (cache). GPU's being one example since they usually have clocks below 2 GHz, but lot of parallel cores.

 

1 hour ago, louisxiawei said:

One thing I still don't understand is that I tried to upsample DXD file (352.8) to 48x512 with poly-sinc-xtr. The load should be significantly lower than upsampling from redbook file, but why is this setting still causing stutter? Which part cause the limitation? The cache of my 7980XE or the speed of the DDR4 3466 64GB RAM?

 

It becomes memory bound, because the filter bank becomes massive when converting between rate families and high ratio. Cache is always much faster than RAM, so you would need huge cache.

 

I haven't got spare 3500€ yet to try out Titan V, but with it's high FP64 performance and HBM RAM it could potentially be good. But never know without trying... Then one could use something i7-8700K that has high clock rate but not so many cores for rest of the work.

 

[louisxiawei]

7 minutes ago, Miska said:

 

If the clock rate is fast enough to run the modulator then it's enough.

 

For the filter, clock frequency doesn't matter if there are more cores and enough memory bandwidth (cache). GPU's being one example since they usually have clocks below 2 GHz, but lot of parallel cores.

 

 

It becomes memory bound, because the filter bank becomes massive when converting between rate families and high ratio. Cache is always much faster than RAM, so you would need huge cache.

 

I haven't got spare 3500€ yet to try out Titan V, but with it's high FP64 performance and HBM RAM it could potentially be good. But never know without trying... Then one could use something i7-8700K that has high clock rate but not so many cores for rest of the work.

 

Thank you, that helps.