For a while I had a problem to stream certain samplerates to my DS.
The trouble was with 48Khz, 96Khz and 192Khz.
These format's only played when I made a cold start with the AV-Streamer (computer streamer) one minute without power.
It than could play 48 and 96Khz for 15 minutes after 15 minutes the music started to skipp and stopt completely in minutes.
192Khz could play with some background noise.
Other formats like 44.1Khz 88.2Khz and 176.4Khz just played 100% fine.
Last Thursday I started a mission to solve the problem and so I reinstalled Windows 8.1 pro 64bit on the AV-Streamer, but without luck. After this I doubted the I2S Surround bridge, because my other PC with a I2S stereobridge just worked fine. Because I do not have a spare bridge that fits the AV-Streamer I had nothing to try.
After that I looked into the other hardware and found out that only 48, 96 and 192 had a problem the memory is 2400Mhz.
2400.000.000 could be divided by 192, 96 and 48Khz the outcome will be whole numbers. So probably it had something to do with synchronisation of certain frequencies where memory became instable. When 2400Mhz is dived by the other formats you will get fractions. So with this in mind I lowered the frequency of the memory to 2133Mhz into range where I get fractions on all samplerates. The problem was solved and an even better sound for the samplerates that already worked fine.
I still doubt the conclusion, but I don't believe in coincidence either. After this I brought the frequency of the processor back from 3700Mhz to 2200Mhz and lowered the voltage from 1.38 volt to 1.2V. The memory is brought back from 2400 to 1066Mhz. The result is stunning better sounstage more definition and even more rest. The soundstage became wider 1 meter in both directions and deeper. The separation of voices and instruments is better all in a certain perspective, but more than enough to sit on the couch to deep in the night.
The understanding of all circumstances that could influence our systems is very annoying. It's not only what's in the music files. I mean is it right that the soundstage became this big or should it be more compact. Is this the information from the music file or is it digitally added.
My computer is a Pink-Faun AV-streamer with a linear powersupply. It uses a 3.7Ghz AMD processor and 16GB 2400Mhz Ram.
The settings I made were all done in the Bios. Today I installed a program to measure the processor temperature and installed a program to do some stresstests. I started at automatic settings 3.7Ghz for the processor and 2400Mhz for the RAM. I made a program and went trough it step by step first a stresstest and than a listening session. With decreasing speed settings you can hear rest in music getting better by every step, but decreasing the cpu voltage keeps the processor much cooler and bigger steps are made definition and soundstage became better.
With maximum setting the processor temperature went up to 65 degrees Celsius, now see the temperature in the screen picture’s.
Off course with normal streaming conditions the CPU temperature will never reach higher than 1 or 2 degrees above the room temperature and te load is less than 1%. I think the real bennefit of the lower voltage and frequency is less leakage and stress for the transistors inside the processor.
The 36 degrees Celsius from the picture above is at full load. The test duration was 25minutes before I stopped it and it never exceeded 36 degrees. I ended up with 2Ghz for the processor 1.235V. The memory now runs on 1333Mhz.
The soundstage with my formal settings 1.2V and 2.2Ghz for the processor and 1066Mhz for the memory was slightly bigger. I like my new settings better because the soundstage is a bit more intimate while definition is slightly better.
Reduce your CPU voltage as much as you can. This cannot harm your computer. I run mine at 1V, but at this voltage I am not able to run multi cores. I just run one core and when I go back to multiple cores it just does not sound as good. If you prefer multi core operation you can pare down the voltage a little at a time until the system becomes unstable, then go back up to where it is stable. I do not use hyperthreading or turbo boost (Yes, an Intel). Perhaps 1.25v is as low as possible with your chip, but it is worth a try, even if you do not use all your cores.
I must look into the bios again to see if it is possible to just run only one core with my system! My system is an audio and video streamer so I need some speed hear and there.
Indeed undervolting will not harm a system unless people make big steps down. The voltage than could be to low to even startup your system to manage the bios again.
I suspect that if you are using it for video that you will not like the one core performance, but it would be interesting to see if you get a change in sound quality. I also use the CPU’s video and run in low resolution mode, again, not good for your setup.
I got some new settings. I,m back again on 1.2V (normal is 1.38V) for the processor now running 1,8Ghz (normal is 3,7Ghz). Memory is back on 1066Mhz but now on 1.5V (normal is 2400Mhz with 1.65V)
The bigger soundstage is back as I had earlier with 1.2V for the processor. When I run a stress test it seems only 1 core is working at 100%. The others stay near 1 or 2 %. The processor temperature stay’s at full load at exactly the room temperature which is 19 degrees Celsius.
The visual result is an audio system complete with two loudspeakers and a sub, but the music is behind the system with no connection to the system 0%.
The audio result is. The voice of Chie Ayado which always stand on 1,5 meter high is now on 1,2 meter high. Normaly she sings 2 meters behind the loudspeakers, but now she stands further away. Because she stands further away the realism is much better. In many cases when I created a bigger soundstage the voices and instruments lost body (Sound became thinner). Not now the soundstage widened and caned the same amound for dept. So nothing has lost it just can be compansated by increasing the volume a bit.
I just want to listen for weeks and do nothing else the realism is realy great. It seems far less important to make room corrections, because the influences of the room are hard to detect. The acoustics of the recording environment are much more clear and changes with every track.
For the record my room is partly corrected. Screen behind the system carpet on the floor and an acoustic ceiling.
Unfortunately, I cannot under clock my CPU since it is locked. This is what led me to undervoltage it in the first place as an alternative. You are able to do both and I have to think that the combination is complimentary. It certainly is in terms of power consumption and the associated heat dissipation! I might have to look into an unlocked i7 now. I haven’t looked recently, but at the time that I set up my BIOS to run lower CPU voltages not much info was on the web about doing either undervolt or under clocking as a way to get better sound. Hopefully your experiences will spur some more interest in this. I’m all ears!
Yes indeed it is a combination that did it for me. I think 60% is due to the lower voltage of the CPU and the memory. 10% for the lower frequency of the memory and 30% for the lower frequency of the CPU.
It would be fun to rate a couple of tested streaming devices with and without tweak settings from 0 to 100%. I will start another thread for this soon!
Dropping the voltage made a big improvement in the sound of my server. I have gone back and forth and prefer the black background and deep soundstage of the 1v setting vs 1.5v.
Are the commercial devices tweakable through BIOS? Most seem to be low power systems to start with. I don’t know if there is much room to play with parameters or if the BIOS even allows that in those units. My friend has a DigiBit Aria which I could try to tweak while he is out on vacation… I bought low voltage memory but discovered later that my mobo doesn’t support the lower voltages. Maybe I need a new mobo AND an Intel K processor.
I don’t know the specific processor you’re talking about. I have and had several computers with different types of processors. It seems that the faster the processor the better the sound, but even better when a fast processor works a half the speed and a quarter of the power.
I will test my other system next week it has an asus Z87 deluxe mainboard with 1150 socket with Intel I7 processor 3,4Ghz and also 16Gb ram. With the Asus many many settings are possible.
I’m thinking of the Atom based systems used in some of the oem units or other, similarly low powered embedded processors. I have an i7 4770 CPU, non - K, which has the clock function locked. Anxious to hear if the results are repeatable on the other combo.
What you appear to be implementing is I think commonly referred to as underclocking or downclocking.
What is the difference between this and deliberately using slower components in the build such as the Antipodes servers which also have great SQ (some say the best)?
Excellent question and I do not know at this point whether starting with a lower powered processor or underclocking/undervolting a more advanced processor to produce audio is more advantageous regarding sound quality. Certainly the smaller cpu/mobo combinations are much cheaper for oem use and can sound quite good. Sometimes you just have to play with this stuff to find out.
Some argue a powerful CPU loafing sounds better than a less powerful CPU working near capacity.
My guess is that whatever configuration produces the less noise will sound the best. Thus lower clock speed with a large data path, etc. may sound better.
Having a CPU/motherboard that’s expecting more heat and not producing that heat will save transitions thru different CPU speeds and will not engage the fan. All probably good things. Turning off Intel Speed Step will probably also help - or having a CPU that over powered enough that the CPU is always running at top speed.
In my formal Pink Faun Audio Streamer a Intel Atom processor was used the cooler was passif and the powersupply linear. The mainboard also was an Intel. I never tried to lower the voltage or the cpu speed. The Streamer with the Atom processor sounded great, but it can’t compete with the AV-Streamer using the AMD 3,7Ghz processor. I liked the audiostreamer better than my normal Intel PC using a 3,4Ghz processor, but not on every aspect.
So I’m not sure if I’m a fan of the Atom because the faster Intel PC 3,4Ghz had some aspects that I liked better while the Atom streamer has a linear powersupply and a passif cooler
I bought low voltage memory but discovered later that my mobo doesn’t support the lower voltages. Maybe I need a new mobo AND an Intel K processor.