I’m toying to build a roon server to replace my i7-7700 w10 machine.
Looking at max 65w TDP for Fanless operation so max will be i9-9900 or i7-9700
512gb m.2 32gb ddr4 and onboard gfx and HDPLEX 200w LMPS. Storage on Nas currently but will migrate a lot of the library to local 4tb ssd drives in the pc as time and funds permit.
This machine will also run fidelizer and hqp. An option for HD and 4K video out to tv Win10 os etc for movies maybe too.
It will also run jriver and maybe audirvana too.
Anything else to consider! The MOCK will probably be optioned on a 2nd M.2 bootable option
Oh running on Bonn N8 and EtherREGEN networking plus a xxxxxRendu of some sort via usb to Denafrips terminator
Although mine is i7 8700 based and a pure MOCK machine, most aspects of what you propose are identical to my m-ITX build. The only thing I’ll add are consider the Jcat Net Femto card to get away from the on board NIC. Or perhaps an optical nic might offer gains without the expense. I would also add that if you really want to put that processor to work, 200 watts may not be enough. I feed mine with a 160 watt silent adaptor supplied by a fixed 12 volt feed from my HDPlex. I have to limit this to two cores for analysing in Roon so with HQPlayer I’d look at the full ATX supply.
I admit I haven’t worked mine out but anecdotally I ran into problems trying to employ more than two cores. Essentially the machine became unresponsive. It may be that the thermals picked up too quickly not allowing the passive heat dissipation to do its job, causing the CPU to throttle. Either way it is something to be aware of but during normal use it has been fantastic and the SQ has been really eye opening with the tweaks.
I use the fc8 too. Great case!
If I were to do anything differently it would probably be to pay more attention to the quality of the thermal compound used and it’s application.
I’d also question the fanless case constraint. Stick the computer in a box near your router far away from the stereo system and use an AIO to keep it cool is my preference.
I’m worried that you may not see a lot of actual substantive changes in HQP capability with various filters and modulators etc. unless you take a step up towards a faster clock speed, which usually needs more cooling than you’ve got in mind.
AMD is kicking Intel’s butt again at the moment imo.
Well the jriver part is for video to a tv in the listening area so that’s a hard one to avoid being in the room…and silent is as much about fan air noise as electrical noise
@jussi_laako Jussi can you maybe lend some insight here as to how the 2 CPU’s I am thinking about might cope…even up to DSD1024 if that is even doable?
Or should Roon and HQP be on different systems…does HQP work fine with Fidelizer too?
Will HQP work with JRiver or Audirvana too?
Can it work thru NAA on OpticalRendu or other Rendu models?
So far i9-9900K has been the only one giving reliable DSD1024 output. But you may still need to carefully select modulator to make sure it keeps up.
I know nothing about Fidelizer, never used it. But usually anything that messes with the OS can potentially damage HQPlayer’s performance.
I have never used JRiver. But HQPlayer Embedded works with Audirvana, since HQPlayer Embedded can operate as a UPnP Renderer.
Since NAA is a software module I’ve made explicitly for HQPlayer and it works only with HQPlayer, sure. DSD1024 likely doesn’t work through Rendu though.
Challenge with these bigger CPUs are not as much filters, if you are using -2s variants. But the challenge are modulators. GPU can help with filters, but not with modulators. With my i7-7700K that is only quad-core, using RTX 2080Ti helps because it frees up CPU cores. But with 8-core CPU like i9-9900K(S) it matters less. So, while the GPU is useful for getting wider selection of filters available for those rates, modulators are the critical aspect here and thus CPU performance.
i7-7700K with help of GPU, and i7-8086K without can do DSD1024 using DSD5 modulator. But ASDM5 gives a drop-out every now and then.
If you look at benchmark figures, possibly single thread performance of LAME MP3 encode benchmark is most representative of relative CPU performance for HQPlayer-type workload.