Best CPU for hqplayer

[sbenyo]

As far as I know the 7700K cannot do XTR but only XTR -2s. If this is the test for CPU usage and temperature it is not the same as doing XTR which is much more CPU intensive. Just to compare, I will try the Prime 95 CPU test to see what results I get.

 

[Quadman]

4 minutes ago, sbenyo said:

As far as I know the 7700K cannot do XTR

Correct as I stated in my first post but the poly-sinc non-2s filters run the CPU in the upper 20% range and prime 95 pegs it at 100% for the full hour that I set.  I use prime 95 for initial set up knowing that the CPU will never get pushed as hard as it does there (except for the initial filter building stage, which usually runs no longer than 1 minute) and if the temps I see are good after fan tweaking and I am satisfied with the noise levels, then I remove prime 95 from the PC.  Stability is very important as I have had some intel CPU's crash (blue screen) during the filter building stages, upping the core voltage usually solves this but then you dance with correct voltage vs. heat.

[n2it]

12 hours ago, sbenyo said:

I think it's still a bit high. It would be great to see comparisions with other Threadripper users (either 1950x or 1920x) using HQP with same Noctua cooler or anything else.

 

Wish I could help ... I don't have any devices that can do native DSD512 and I don't have HQPlayer (although it looks like I could use the trial - not sure what limitations there would be).   I am not sure if a DSD512 dac is within my price range right now. 

[sbenyo]

Thanks. I hope someone will come with additional Threadripper temperature results for HQPlayer. I actually did a few more stress tests even loading CPU to 100% for sometime. It still looks like HQP is winning in raising the temperature. Everything else I tried did not get as high as HQP.

 

[bibo01]

Hi guys,

 

Has anybody got experience with deliding which could be useful in terms of CPU temeperature?

"Deliding is the process of removing the IHS (Integrated Heat Spreader) from the CPU which aim to drop CPU temperatures. Since Intel Series 3, 4, 6 and now 7, Intel uses a black silicone type to glue the IHS to the PCB, but the die barely touches it. And Intel uses a poor kind of thermal paste too, so there's very poor contact between the die and HIS, resulting in bad heat dissipation, thus a hot CPU."

 

There are various guides and videos about it on the net:

https://www.tweaktown.com/guides/8039/cpu-deliding-quick-guide/index.html

https://www.ekwb.com/blog/what-is-delidding/

 

[Quadman]

2 hours ago, sbenyo said:

It still looks like HQP is winning in raising the temperature.

That shouldn't be.  Stress tests, especially, peg your CPU at 100% for almost the entire test.  100% CPU utilization creates more heat than 30-40% utilization with HQP XTR filters.  Unless you use a GPU and set CUDA offload (checked) and the GPU creates so much heat that it raises the ambient temperature inside you case so much that the CPU runs much hotter.  But I think you said you have to grey the cuda box which means cuda is then only used for convolution and I don't think you do that so heat from GPU probably not an issue.  What does task manager show for GPU utilization while you play music?  If you do use CUDA offload then you need a GPU fan control program like MSI afterburner to control the GPU fan to keep GPU cooler.

 

37 minutes ago, bibo01 said:

got experience with deliding which could be useful

I have read up on deliding but have never had the courage to go forward with it in fear that I might trash the CPU and at $350-$800 for these things is a bit much for my risk tolerance.  I agree that it could have a very positive effect on CPU temps tho.  Meantime I make sure to use excellent thermal paste, as thin a layer as possible and good CPU coolers to keep temps down.  I never use the stock thermal paste that comes with most coolers, even Noctura which is pretty good.  They just apply it way to thick to do justice.

[bibo01]

1 hour ago, Quadman said:

I have read up on deliding but have never had the courage to go forward with it in fear that I might trash the CPU and at $350-$800 for these things is a bit much for my risk tolerance.  I agree that it could have a very positive effect on CPU temps tho.  Meantime I make sure to use excellent thermal paste, as thin a layer as possible and good CPU coolers to keep temps down.  I never use the stock thermal paste that comes with most coolers, even Noctura which is pretty good.  They just apply it way to thick to do justice.

There are shops that apply delide for you as you buy a CPU. If the price is not too high, let them take the risk. They also have the necessary tools to do it.

I am seriously considering it for my next system.

[Solstice380]

Here is a Ryzen 1950x Noctua cooler comparison with some good load and temp data. 

 

http://www.legitreviews.com/air-versus-water-cooling-amd-ryzen-threadripper-noctua-coolers_198402

 

[sbenyo]

2 hours ago, Solstice380 said:

Here is a Ryzen 1950x Noctua cooler comparison with some good load and temp data. 

 

http://www.legitreviews.com/air-versus-water-cooling-amd-ryzen-threadripper-noctua-coolers_198402

 

Thanks for the link. I actually saw it and repeated the exact same test. It's downsizing a 4K 60fps video to 1080p 30fps.

This works for about 10 mins loading CPU to almost 100%

 

My CPU starts at idle time with a bit higher temp. They show 28c and I see 30-35c.

As for temperature on load, I was close to what they show but not as high. It was most of the time between 50-55 with peaks to around 60c.

 

Surprisingly this loads CPU to 100% and it has less effect on temperature than HQPlayer loading CPU only to 30%-40%.

 

I am still not sure what HQP is doing to Threadripper CPU that cause the temperature to be ~55c with frequent peaks above 60c.

I am also not sure if this is a problem with my cooling solution or not. If I compare it to the link above, it seems Threadripper gets hot but it's probably OK.

 

I still need to see other HQPlayer comparisions or more details on expected temperatures to be sure if my cooling solution is good enough or not.

 

[Quadman]

44 minutes ago, sbenyo said:

I am still not sure what HQP is doing to Threadripper CPU that cause the temperature to be ~55c with frequent peaks above 60c.

Impressive results from the Noctura coolers.  Do you keep your CPU at a set frequency?  or do you allow it to adjust as the load requires?  If the latter that is why your temps with HQP may be higher than under 100% load.  HQP is running higher frequencies and higher voltages than the CPU does under 100% load as the review above showed.  As long as you can keep peak temps below 68 you should be okay.

 

I use an HTPC case for my 1800x so I need a lower profile cooler and the case does not accommodate water cooling so I have been running a noctura NH-9DL cooler and it does a fantastic job of keeping my CPU cool.  I also set my frequency multiplier at 3.8GHz and core voltage at 1.32v.  Noctura makes good stuff.

 

 

[sbenyo]

I am not sure how I can control the CPU frequency. As far as I know I use it on default set frequency (3.5GHz).

 

Checking it using HWInfo may show I might be wrong. It does seems frequency is changing  and even reaching ~4.0GHz.

This may explain what I see. I will have to further check it to see if this is true and if I can control it.

 

If this is the cause for the higher temps and if this can be resolved it will be great!

 

 

 

 

[sbenyo]

I looked into the MSI x399 BIOS settings and found that OC is disabled and can be only set physically on the MB.

I checked deeper and found there is an advanced setting "Core Performance Boost" set to auto. I disabled it.

 

Now CPU frequency does not exceed 3.5GHz and temperature sits nicely at ~45c!

'

 

I am also happy with it as CPU fans don't have to get to full speed which keeps everything dead silent.

 

I am very pleased now. I thought the CPU is getting too hot. It turns out it overclocked itself as high as 4.0GHz.

CPU is now working a bit harder moving from 30% to 70% and sometimes even to 90%+  but everything still works perfectly so I don't need to Core Boost for DSD512 XTR. Before CPU would not get higher than 30%.

 

Quadman - Thanks for giving me the right hint where to look for. I am still surprised BIOS default setting is set like this even though OC is disabled.

 

I will also run the video stress test again to see how CPU and heat performs now.

[Miska]

This AMD CoreBoost is similar to Intel's TurboBoost. It can boost clock frequency of some of the cores when other cores are on lighter load. Amount of boost it can give depends on load of each core, you can get highest boost when only single core is loaded. In practice it aims to keep the total CPU package within power limits while allowing performance boost on some cores when others are under lighter load. So this feature is not considered overclocking, but just standard feature of the CPU (and can be disabled if necessary).

 

[sbenyo]

18 minutes ago, Miska said:

This AMD CoreBoost is similar to Intel's TurboBoost. It can boost clock frequency of some of the cores when other cores are on lighter load. Amount of boost it can give depends on load of each core, you can get highest boost when only single core is loaded. In practice it aims to keep the total CPU package within power limits while allowing performance boost on some cores when others are under lighter load. So this feature is not considered overclocking, but just standard feature of the CPU (and can be disabled if necessary).

 

Thanks Miska. I didn't know about it. It does seem to affect CPU temperature significantly so I prefer not to use it if not needed.

I will have to check more into it to better understand it. It may be that even with this option, it may be still fine even if temperatures can get to 60c+. It is mentioned that it is not recommended to exceed 68 on AMD Threadripper.

CPU wise it seems to be better utilization but I am not sure it's good enough because of the temperature.

 

I also wonder how this works if memory boost is not possible. Can I get 4.0GHz clock speed on CPU with 2400Mhz memory speed or does it require memory to support it as well?

 

 

 

 

 

[Quadman]

1 hour ago, sbenyo said:

I checked deeper and found there is an advanced setting "Core Performance Boost" set to auto. I disabled it.

 

I see Jussi answered the core boost question.  Since I run the 1800x I have not researched the bios settings for the threadrippers as I currently do not a build planned using it.  But in my bios in the MIT section under Advanced frequency settings I manually enter a setting for CPU Clock control and I set mine at 100.1, and then under CPU clock ratio I set mine to 38 as I wanted to OC my stock 3.6GHz to 3.8GHz.  With the 100.1 clock setting and 38 ratio I get a speed just a hair over 3.8 GHz.  If I leave the clock at stock 100 then I get slightly under 3.8 and I hate that so thats why I slightly adjust the clock.  Your MB bios may be similar or different.

 

Since your CPU utilization took a big jump up and it is possible now you may get some stuttering, so I would manually adjust your clock to 100.1 and the clock ratio to say 37, just a small OC.  You may also need to up CPU core voltage a bit with this move (or leave on auto, but overvolting (heat) more likely), a hundredth of a volt can make a big difference.  Then run HQP and see what the utilization looks like, the goal being to get it closer to maybe 50% and also keeping heat lower than before.  That is that dance I mentioned earlier about stability and heat.  Check some forums about overclocking the 1920x to see if it is similar to what I describe above.

[Miska]

1 hour ago, sbenyo said:

I also wonder how this works if memory boost is not possible. Can I get 4.0GHz clock speed on CPU with 2400Mhz memory speed or does it require memory to support it as well?

 

Memory runs at it's own speed and there are layers of cache between RAM and CPU. So this boost is independent of memory bus which runs at constant speed.

 

[sbenyo]

Thanks for the valuable info!

 

I actually see my CPU running at ~30% CPU with HQPlayer with boost disabled. Currently I don't see any benefit in the boost. I also ran other performance tests and there is no significant difference.

 

I am sure I can play with it manually as suggested to tweak the performance just in the right balance between power and temp. I will play with it a bit to see what is the optimal settings.

 

[sbenyo]

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?

 

 

[Miska]

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

 

[sbenyo]

I don't really need it for now. The extra heat as I see now is just 2-3c which is not significant.

I do run the memory with higher voltage 1.35v instead of 1.2v and I don't know if this has any implications.

I should probably not use it if not needed unless someone has new information in favor of it.

 

[Quadman]

If your DDR4 supports 3200 then that is what I would run it at.  In Bios it looks like you have a XMP switch on the main screen, turn it on and that should jump your RAM speed up to its rated spec (you may need to reboot to see it). If that doesn't work it looks like you can also adjust XMP in the OC settings in bios.   I have 3000 Mhz in my 1800 and use XMP to get it to 2966.

[sbenyo]

I am able to run at 3200 without issues. I will keep it for sometime until I understand if it's worth it or not. It does seems to affect temp much but I have to run a bit more to be sure.

 

[Quadman]

This thread started as "Best CPU for HQPlayer" well that is a very broad question.  It maybe should have been worded "Best CPU for HQPlayer with DSD512 up sampling".  Are you determined to make a server that can play every HQP filter because you believe the non-2S sound significantly better, well then you need to step up to the AMD threadripper line beginning with the 1920x or even 1950x, then on the intel side Jussi has used the 6950x to accomplish this and some of the other high end intels may as well tho I have not read confirmations on them.  Regardless even with the cheaper 1920x a full build and optimized server your looking at around $1800 minimum before software costs, not cheap.

 

If you just want to listen to HQP with DSD512 up sampling and are satisfied with the 2s filters then you can get by with a much more modest rig.  My experience is that the 2s and non-2s are not that much different sonically and you need a pretty revealing system to even discern the differences.  Because I can I do use the non-2s filters most of the time but switching back to the 2s version does not leave me yearning to switch back ASAP.  I have built 8 audio only PC's in the last 3 years starting with intel i7-4790 CPU thru AMD's Ryzen 1800x CPU.  All these processors can handle DSD512 with the 2s filters in HQP,  obviously the 1800x can do most of the non-2s filters as well but it certainly cannot do all the non-2s filters.  My experience has lead me to look for certain things in a dedicated music server.  Lately I am a big believer in more L3 cache as being beneficial for music playback and overall sound quality and that is the main reason I jumped from intel to the AMD line.  I have also found newer processors or one older generation are generally the best buys,  going back 2-4 generations the prices actually begin to rise.  My experience is with Intel's i3-i7 lines and AMD Ryzen line so I am not qualified to comment on other intel or AMD processor lines.

 

My recommendation for best bang for the buck processors would be as follows.  (All should be able to handle DSD512 up sampling with HQP 2-s filter family)

 

AMD Ryzen 5-1600x  retails for around $200.  It has 6 cores and 12 threads and a base speed of 3.6Ghz out of the box, and more importantly a whopping 16MB L3 cache and a 95W TDP rating.  Runner up Ryzen 5 1500x about $25 cheaper, same # cores and threads and L3 cache but drops base speed to 3.5 Ghz (still good) and lowers TDP to 65W, meaning it may work in fanless cases with careful cooling setup.

 

Fot intel I recommend the new i5-8600K processor it retails around $240, has 6 cores and 6 threads with a base speed of 3.6Ghz and 9MB of L3 cache with a 95W TDP rating.  The i5-7600k is so closely priced to the 8600k that I would not get it as its L3 cache is down to 6MB, but does offer higher base speeds of 3.8Ghz, core count drops to 4 as does threads.

 

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.

 

There is a lot more to a music server than just the CPU but this thread focuses on CPU's.  Just my 2 cents on the subject.

[sbenyo]

For me there is a noticeable and even significant difference in quality between XTR and -2s.

 

Is it worth the cost of a $1800+ PC? This is a matter of budget and other investments already done. If the setup is very expensive and capable of taking advantage of it, it may worth the price. It has to be in proportion to the quality and cost of the complete system.

 

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.

 

[Sevenfeet]

@Quadman, that sounds pretty reasonable, especially since many of us aren't on elastic budgets.

 

One of my limiting factors is that despite the number of computers in my house, most are old.  The newest is my wife's 2013 Macbook Air.  My daily driver is a 2008 Mac Pro and the music server is another 2008 Mac Pro.  Both Mac Pros are 8 core and do nicely for what they usually do, but they run into issues trying to get the most out of HQplayer.  For example, neither of them seemed to be able to push DSD512 to either of my NAAs, both of which are also elderly Macs.  For example, a 2009 Macbook hacked to run Win 10 is an NAA/Tidal machine for the Oppo 205 in my home theater rack.  And frankly, I wasn't sure if the DSD512 issue was a limitation of the server CPU bandwidth of the hardware in the NAAs.

 

The best machine I've had in my house in a year was a custom rig I built for a friend who uses it for a flight simulator with three Samsung 4K curved TVs.  That machine is a X99 motherboard with a i7-6950K with 6 cores (12 hyperthreaded) running at 3.60 Ghz (not overclocked).  It is liquid cooled with a Corsair rig doing that.  Anyway, my friend is having trouble getting the machine to update to the latest version of Windows 10 so it's back in my house for a few days, and it's giving me a chance to run HQPlayer with a higher performance platform.

 

My findings:

1.  Old Macs are just fine for NAA work at DSD512.  My problem is with the bandwidth of the Mac Pros in my house.  It's not a huge surprise since a mid-range 2017 iMac can outrun my machines but it's nice to validate the NAA issue anyway.

 

2.  The FlightSim rig can upsample DSD files up to DSD512 without breaking a sweat.

 

3.  The FlightSim rig cannot upsample PCM to DSD512 on the hardest filters (like poly-sync-xtr).  It can handle the -2s just fine.  This isn't a surprise...others have found the same "issue".  I was curious as to whether the CUDA offload would help in a machine with dual GTX 1080 cards...spoiler alert, it doesn't.  And HQPlayer only uses one card anyway.