Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[AspiringAL]

As the title suggests, I am going in a different direction with my HTPC then "minimalist" set-ups that are common for dedicated media rigs.

I use my HTPC for music, movies, gaming, video editing, graphic arts, basic animation, and surfing.    In many cases, I am playing music while I do surfing and computer arts.

My current set-up is:
Intel Core Duo Quad Q9550 overclocked to 3.4 GHz. 
Video is handled by a HD 6970 Graphics Card
Audio DAC #1 is M-Audio Fast Track Pro
Audio DAC #2 is On-board Realtek (for multi-channel)
Audio DAC #3 is Twisted Pair Buffalo III kit which is not yet built due to finding a reliable way to stream 8 channels

My JRiver benchmark is 2956 with this rig.

This system is running several custom built speakers I have made.  I use a combination of PEQ and high FET convolution (65536/channel) made in rePhase, as the high FET seems to create a more 3D soundstage, subjectively speaking.   

When using my single driver, full-range stereo speakers (2 channels), I get a SSE real time score around 25 X
When using my line arrays and subs (4 channels), I get a SSE real time score of 19 X for music and 12 X for Blu-ray MKV files.  CPU usage is 22% for Blu-ray playback
When using my Quad surround system (8 channels), I get a SSE real Time score of 8 X for Blu-ray
All video playback is done using Red October HQ. 

For games and surfing, I do not use convolution, as the latency is way out of hand.  Currently, We do not play any hard core games, but I keep the possibility in mind. 

My current system does well for my current situation.  Most of the time, we watch movies and listen to music in stereo, as I am still building the amps and speakers that will replace the current ones (which will be reassigned to the back  )  Again, I am planning for the possibilities, as I definitely prefer 5.1 (4.4 in my case) playback for movies and am heading that way.

Looking at the SSE scores above, esp considering 8 channel playback of music while doing graphic arts, I can see out growing this system soon.

The two CPU's I am considering for the upcoming build is the Haswell-E i7-5820 or the Skylake i7-6700.  The skylake is faster, yet the Haswell-E has more cores and more PCI-e lanes. 

I am considering PCI lanes, as it seems the best way to integrate my Buffalo III DAC is to extract the I2S streams from an on-board sound card, like the Asus Xonar.  I am curious to know if having dedicated CPU lanes to the DAC would help with latency.  I am open to other DAC options, as the Buffalo Build is turning out to be more then what I expected.  I am better at woodworking then I am at soldering. 

I would like to improve convolution latency, but I have a feeling it is based on FETs.  Does a faster CPU or DAC help?

The Skylake rig is about $200 cheaper then the Haswell-e, but I am willing to pay the extra $200, esp considering the next Intel 2011 v3 chip set will fit the X99 motherboard.

The question is, do I need the Haswell-e, considering the multi-tasking I do and the high FET convolutions, or would the speed of the Skylake overcome that?  

I would also like to play around with MadVR, which to my understanding is a CPU task?

I am in no rush to upgrade the HD 6970 graphics card, as it seems it is more of the CPU bottle necking my current system.  I do have a second HD 6970 in my office computer (way overkill) that I can cross-fire in the new system to buy time for a graphics upgrade.

I have a Fractal design R5 case, a Consair HX750i PSU, and two Be quiet 140 silent wings fans to help keep the noise down.  As the Skylake is "only" a 85 watt chip -vs- 140 of the Haswell-e, I am sure the Skylake will be a quieter system. 

Thanks in Advance for your suggestions. 



 

 

[mattkhan]

The latency of the filter is a function it's " length"and the sample rate, 64k filter at 48kHz has a delay of ~2/3s. To reduce the latency you need to shorten the filter (reducing frequency resolution and hence increasing the minimum frequency at which it is feasible to correct phase), run at a higher sample rate (which increases CPU load) or use a partitioned convolver (which jriver isn't and which would consume as much CPU as you have available). In the absence of any of those options, increasing CPU horsepower would enable more paths (ie higher channel count) but otherwise gives you nothing I think. The main benefit is enabling additional users, e.g. streaming to other clients.

Madvr is a GPU consumer BTW not cpu

[AspiringAL]

Thanks mattkhen,

I had a feeling the latency was tied to the filter length.  Thankfully, the latency does not effect my listening enjoyment of music or movies.  With Games and Online streaming, the audio is already compromised so much, that I do not mind running my system with the IIR filters provided in PEQ, still way better then their passive counter parts (at least for my speakers).

Now knowing madVR is a GPU task, I should give it a try as the HD 6970 should have no problems. 

Thanks for clearing things up.

[AspiringAL]

I have been doing a lot of research today to see if I can find a way to get an 8 channel signal to my Buffalo III kit.  All the options so far are well outside my comfort zone. 

I am looking into the Asus Essence STX 7.1.  I am thinking I can run my back speakers and subwoofers off of that, and maybe run my Buffalo III off the SPDIF for my main two channels.  In this case, I will build the Buffalo III as a two channel DAC, which I have all the parts for.  I will need to do additional research on this, to make sure I can align the clock of the Buffalo to the Asus and that JRiver will "see" the SPDIF as the front channels.  It is a shame I can not do this with the Realtek, I will look into it...

[AspiringAL]

I figured I would chime in to update where my current thoughts are on the computer and DAC upgrades...

As my current PC seems to work very well for the 2.2 system I am currently running, I am planning on focusing more on getting my DAC situation sorted out.  But I can say, I am leaning more towards a budget X99 platform, as I like the expansion options it has.  This will depend on my eventual DAC solution also, whether or not I will using PCI-e or USB.

With the Buffalo III DAC kit, I have the following options:

#1:  Build it as a two channel DAC and input my audio from the on-board SPDIF.  To get my 2.2 system, I would build an analog active cross-over board to hook-up between the DAC and Amps.  I could still use convolution to phase correct the cross-over.  This has one advantage of needing only 2 channels of convolution, but I am adding an extra circuit to my audio chain and this set-up will not work for my future multi-channel system.

#2:  Same as above, except using the new AMANERO USB RECEIVER available from Twisted pair.  It sounds like this is the "Cat's Meow" as far as getting a low jitter, bit perfect signal to my DAC.  Once again, with the kit as is, I am limited to two channel, and this costs $238.00 (which I believe would be a steal if it was multi-channel).  But there may be ways of using parts of this kit for a multi-channel solution. 

#3:  Buying an Asus Essence STX 7.1 and tapping the I2S lines using TELEPORTERS available from Twisted pair.  Here I could build my Buffalo III as a 2, 4 or 8 channel kit, using the Asus as the front end.  If I wanted to go all out on quality with my Buffalo, I could build that as 2 channel and, tap off 2 I2S lines off the daughter board I2S and still be able to use the "better" output stages on the main stereo channels of the Asus for my back speakers.  If I was willing to go 4 channel with my Buffalo now (and build a second 2 channel buffalo later), I would consider replacing (not even buying) the daughter board altogether.  My concern with this option are the Asus drivers and bottle-necking the 32 bit inputs of the Buffalo.  I do not seeing a point of JRiver going from 64 bit to 24 bit (for the Essence) back to 32 bit at the Buffalo. 

#4:  Is scraping the Buffalo III build all together and going the Exasound or MUTO routes, both of which cost more $$$$.  They have the same chips, but the Buffalo is really open ended on output stages, which is what drew me to it in the first place.

The big consideration with any of these options is the drivers. 

There are a lot of loose ends regarding high-end multi-channel DACS for PCs especially with the DIY Buffalo III.  It seems JRiver is ahead to its time!  So as the answers come to me, I will continue to post my findings. 

I just hope, as I move along on this journey, that this thread turns out to be helpful for someone!