A novel way to massively improve the SQ of computer audio streaming

[elan120]

5 minutes ago, Miska said:

 

Has someone tested a lower latency memory? At least HyperX (Kingston) Impact DDR4 offers faster CAS latency 15 for DDR4-2666 speed.

 

I originally use Hyper X Impact DDR4 2400 with CL14, and Apacer DDR4 2400 clearly surpass it in SQ, and now Apacer DDR4 2666 is another step up.

[Miska]

16 minutes ago, elan120 said:

I originally use Hyper X Impact DDR4 2400 with CL14, and Apacer DDR4 2400 clearly surpass it in SQ, and now Apacer DDR4 2666 is another step up.

 

OK, trying to think of what the technical factor could be.

 

For me, the server memory speed is one of the most critical factors in addition to CPU clock speed to allow best possible algorithms.

 

While almost all my NAA devices have RAM soldered directly to the CPU board. Some even have RAM and CPU stacked same way as has been done on mobile phones for a very long time (CPU on the motherboard and RAM on top of the CPU, lower side of the CPU attaching to the mother board has larger BGA array while on the upper side BGA array runs around the chip itself. This makes the whole package extremely compact).

 

I have just one NAA that has SO-DIMM memory, which is LogicSupply (now called OnLogic) CL100, but it has Transcend memory modules, from the picture seems to have Hynix chips:

https://www.onlogic.com/eu-en/ts1gsk64w6h/

I have never opened the device, so don't know for sure.

 

[auricgoldfinger]

30 minutes ago, Miska said:

 

Has someone tested a lower latency memory? At least HyperX (Kingston) Impact DDR4 offers faster CAS latency 15 for DDR4-2666 speed.

 

 

I was previously using this memory:  https://www.amazon.com/gp/product/B014R8JUWO/ref=ppx_yo_dt_b_search_asin_title?ie=UTF8&psc=1

 

The Apacer RAM sounds far better.

[Bricki]

2 hours ago, luisma said:

Currently have this NUC7i7BNH, on specs doesn't support ECC ram so I guess ECC could be ruled out, anyone using a ram chip which makes a difference in sound with industrial specs non-ECC?

Yes, the RAM in my comparison is non ECC... All Intel "i" series and lower processors do not support ECC... Xeon does and I think most AMD processors do as well.

[Nenon]

I have no scientific explanation why all these RAM chips sound different, but they do. I have three different types of Apacer RAM and one set of G.Skill DDR4-3200 CL16 with me at the moment, and every single one of them sounds different. 

Two are ECC and the other two are non-ECC.

1. The G.Skill DDR4-3200 CL16 sounds worst of all although it is the the lowest latency.

2. The Non-ECC Apacer DDR4-2666 CL19 sounds a lot better. 

3. ECC Apacer DDR4-2400 CL17 sounds different than #2. I can't tell which one is better. It's a matter of taste.

4. ECC Apacer DDR4-2666 CL19 is my favorite at this time. 

 

My guess is that the component quality matters as well as the execution, quality of service, etc. Apacer is doing a good job with their industrial wide range temperature RAM, and it is a no brainer to me. 

 

The most interesting post I have found about RAM is from Emile who designed the Taiko Extreme server. You can read it here:

https://www.whatsbestforum.com/threads/taiko-audio-sgm-extreme-the-crème-de-la-crème.27433/page-2#post-556159

 

I am tired of testing RAM for now and not planning to do it for a while. But I highly recommend to people to try the industrial Apacer RAM. Unfortunately the new model, and the one I like the best, is not available anywhere online. A friend who works directly with Apacer did a favor with the group buy. He said he bought a few extra pieces. If anyone missed the group buy, ping me on PM, and I will check if there are any left. 

 

All the reports have been extremely positive. The feedback about what people heard in their systems was very useful. I learned a lot from this. Thank you for all of the feedback. People tried it on NUCs, Intel consumer-based computers, AMD-based computers, Xeon-based, running different operating systems, tweaked differently. And everyone liked it so far. 

Only one person told me that he prefered the older Apacer model. But everyone else who had the old version and tried the new Apacer RAM prefered the new one. Not a single person reported negative feedback. And everyone who has sent me feedback reported an improvement over the previously used non-Apacer RAM. That's telling me whatever Apacer does, it is a universal upgrade so far. 

One other thing that was consistently reported was that it gets better over time. I was trying to keep track of people's impressions and statistically it seems like the RAM settles in about 70-100 hours. I thought I heard improvement over time but was not careful and that was the one thing I was not sure about. 

[Chopin75]

1 hour ago, Nenon said:

I have no scientific explanation why all these RAM chips sound different, but they do. I have three different types of Apacer RAM and one set of G.Skill DDR4-3200 CL16 with me at the moment, and every single one of them sounds different. 

Two are ECC and the other two are non-ECC.

1. The G.Skill DDR4-3200 CL16 sounds worst of all although it is the the lowest latency.

2. The Non-ECC Apacer DDR4-2666 CL19 sounds a lot better. 

3. ECC Apacer DDR4-2400 CL17 sounds different than #2. I can't tell which one is better. It's a matter of taste.

4. ECC Apacer DDR4-2666 CL19 is my favorite at this time. 

 

My guess is that the component quality matters as well as the execution, quality of service, etc. Apacer is doing a good job with their industrial wide range temperature RAM, and it is a no brainer to me. 

 

The most interesting post I have found about RAM is from Emile who designed the Taiko Extreme server. You can read it here:

https://www.whatsbestforum.com/threads/taiko-audio-sgm-extreme-the-crème-de-la-crème.27433/page-2#post-556159

 

I am tired of testing RAM for now and not planning to do it for a while. But I highly recommend to people to try the industrial Apacer RAM. Unfortunately the new model, and the one I like the best, is not available anywhere online. A friend who works directly with Apacer did a favor with the group buy. He said he bought a few extra pieces. If anyone missed the group buy, ping me on PM, and I will check if there are any left. 

 

All the reports have been extremely positive. The feedback about what people heard in their systems was very useful. I learned a lot from this. Thank you for all of the feedback. People tried it on NUCs, Intel consumer-based computers, AMD-based computers, Xeon-based, running different operating systems, tweaked differently. And everyone liked it so far. 

Only one person told me that he prefered the older Apacer model. But everyone else who had the old version and tried the new Apacer RAM prefered the new one. Not a single person reported negative feedback. And everyone who has sent me feedback reported an improvement over the previously used non-Apacer RAM. That's telling me whatever Apacer does, it is a universal upgrade so far. 

One other thing that was consistently reported was that it gets better over time. I was trying to keep track of people's impressions and statistically it seems like the RAM settles in about 70-100 hours. I thought I heard improvement over time but was not careful and that was the one thing I was not sure about. 

I am sure mine is still burning in (APacer  ECC 8 GB), as the moment I take it out of the PC the sound gets worse and it takes hrs before the sound gets back to previous great level. So I gather tt is still quite unstable after few days of constant playing it. I agree it is a pain in the butt to do RAM comparison. It is hard to take them out of the slots and also the RAM cools down so then the sound gets worse when you put them back in etc.... so really hard to do A/B comparison unless you have 2 identical PC running at the same time. 

[rickca]

Comet Lake desktop lineup from Intel has leaked.  This is still 14 nm, the 10 nm desktop timeframe is unknown.

So we will have 10 core/20 thread i9-10900K as the new top of the heap.  Really there's nothing all that exciting here over the current 9th generation desktops.

 

https://www.theinquirer.net/inquirer/news/3084095/intel-comet-lake-s-desktop-processor-lineup-in-full

[Miska]

4 hours ago, rickca said:

Comet Lake desktop lineup from Intel has leaked.  This is still 14 nm, the 10 nm desktop timeframe is unknown.

So we will have 10 core/20 thread i9-10900K as the new top of the heap.  Really there's nothing all that exciting here over the current 9th generation desktops.

 

https://www.theinquirer.net/inquirer/news/3084095/intel-comet-lake-s-desktop-processor-lineup-in-full

 

For me it would be big news if they manage to keep 9900K-like clock speeds and again add another two cores. Like it's been 7700K -> 8086K -> 9900K(S)... Percentage wise the increase would get smaller, but still be relatively big.

 

[RickyV]

I am wondering if the "wide temperature" think would also work for motherboards? Supermicro has wt boards.

[seeteeyou]

Indeed, industrial offerings would give us different ranges as follows

 

https://fit-iot.com/web/product/fitlet2-build-to-order/

Quote

 

Commercial temperature range 0°C to 45°C

Extended temperature range -20°C to 70°C [TE] $21.40

Industrial temperature range -40°C to 85°C [TI] $107.00

 

 

BTW, Apacer also announced the "latest the greatest" recently that would require specific motherboards

 

http://www.perfectron.com/products/embedded/ATX/AB20.html

http://www.perfectron.com/products/embedded/EBX/index.html

https://industrial.apacer.com/en-ww/NEWS/The-First-XR-DIMM-DRAM-Module-with-RTCA-DO-160G-Certification-From-the-Leading-Supplier-of-Industrial-grade-Memory

https://industrial.apacer.com/en-ww/DRAM/DDR4-XR-DIMM

 

[Chopin75]

10 hours ago, seeteeyou said:

Indeed, industrial offerings would give us different ranges as follows

 

https://fit-iot.com/web/product/fitlet2-build-to-order/

 

BTW, Apacer also announced the "latest the greatest" recently that would require specific motherboards

 

http://www.perfectron.com/products/embedded/ATX/AB20.html

http://www.perfectron.com/products/embedded/EBX/index.html

https://industrial.apacer.com/en-ww/NEWS/The-First-XR-DIMM-DRAM-Module-with-RTCA-DO-160G-Certification-From-the-Leading-Supplier-of-Industrial-grade-Memory

https://industrial.apacer.com/en-ww/DRAM/DDR4-XR-DIMM

 

I suppose if one needs to play music next to the volcano, in the desert,  or in a space shuttle it will continue to play music quite nicely! 

[seeteeyou]

Not sure if that were meant to be humorous or sarcastic, though I simply wanna keep extracting information everywhere so that whatever manufacturers on this thread could take some notes. Pretty much like how I pointed out Marcin's RAM adventures back in August and now everyone is able to reap the benefits thanks to @Nenon (as well as his buddy)

 

https://audiophilestyle.com/forums/topic/30376-a-novel-way-to-massively-improve-the-sq-of-computer-audio-streaming/page/589/?tab=comments#comment-980943

 

As always it's just a matter of stringent quality that would enable Apacer to meet the requirements of various industrial standards. If they were able to survive an insane level of vibration, then it's gonna be piece of cake to handle electromagnetically-induced acoustic noise and vibration (we should know better when it comes to EMI / RFI these days) or micro vibration mentioned here

 

https://www.audioshark.org/vibration-isolation-devices-26/installed-new-isolation-platform-my-music-server-7804.html

 

Just like what @nvitorino mentioned in this podcast below, lots of small improvements really can add up to major improvements

 

https://www.audiostream.com/content/bits-are-bits-or-are-they-nuno-vitorino-innuos-audiostream-podcast-no-2-0

http://feeds.soundcloud.com/stream/526930206-user-766769873-audiostream-podcast-no2-with-nuno-vittorio-of-innuos.mp3

 

Maybe even having a different connector (i.e. XR-DIMM instead of DIMM or SO-DIMM) alone might get us one step closer to the RAM inside SGM Extreme? There's still a fairly wide gap between the SQ of flagship models from Antipodes / Aurender / Innuos and Taiko Audio so far, hopefully having double-regulated PSUs from Dr. Sean Jacobs as well as (much) better RAM etc. could narrow that gap in the future.

 

Even SSDs from Apacer seemed to be quite different but not sure if that were relevant for music servers

 

https://industrial.apacer.com/en-ww/Technology/CoreVolt

https://industrial.apacer.com/en-ww/Technology/CorePower

[shahed99]

@Nenon Thank you for facilitating purchase of Apacer RAM. I am eagerly waiting for mine.

 

I became a RAM believer when I build my dedicated music server using all Paul Pang components in early 2015. I recall when I was waiting for Paul Pangs special RAM (non ECC) to arrive, I was using regular consumer grade RAM. Once I replaced them with PPA RAM, the improvement in sound quality was immediate, noticeable and quite significant. I still have those although they are DDR3 RAM. I'm curious what Paul Pang has up its sleeve lately.

[Chopin75]

2 hours ago, seeteeyou said:

Not sure if that were meant to be humorous or sarcastic, though I simply wanna keep extracting information everywhere so that whatever manufacturers on this thread could take some notes. Pretty much like how I pointed out Marcin's RAM adventures back in August and now everyone is able to reap the benefits thanks to @Nenon (as well as his buddy)

 

https://audiophilestyle.com/forums/topic/30376-a-novel-way-to-massively-improve-the-sq-of-computer-audio-streaming/page/589/?tab=comments#comment-980943

 

As always it's just a matter of stringent quality that would enable Apacer to meet the requirements of various industrial standards. If they were able to survive an insane level of vibration, then it's gonna be piece of cake to handle electromagnetically-induced acoustic noise and vibration (we should know better when it comes to EMI / RFI these days) or micro vibration mentioned here

 

https://www.audioshark.org/vibration-isolation-devices-26/installed-new-isolation-platform-my-music-server-7804.html

 

Just like what @nvitorino mentioned in this podcast below, lots of small improvements really can add up to major improvements

 

https://www.audiostream.com/content/bits-are-bits-or-are-they-nuno-vitorino-innuos-audiostream-podcast-no-2-0

http://feeds.soundcloud.com/stream/526930206-user-766769873-audiostream-podcast-no2-with-nuno-vittorio-of-innuos.mp3

 

Maybe even having a different connector (i.e. XR-DIMM instead of DIMM or SO-DIMM) alone might get us one step closer to the RAM inside SGM Extreme? There's still a fairly wide gap between the SQ of flagship models from Antipodes / Aurender / Innuos and Taiko Audio so far, hopefully having double-regulated PSUs from Dr. Sean Jacobs as well as (much) better RAM etc. could narrow that gap in the future.

 

Even SSDs from Apacer seemed to be quite different but not sure if that were relevant for music servers

 

https://industrial.apacer.com/en-ww/Technology/CoreVolt

https://industrial.apacer.com/en-ww/Technology/CorePower

Hmm 

the Apacer M2 is interesting:

 

https://industrial.apacer.com/en-ww/SSD/PV310-M280

 

Faster than the optane M10 that many on this forum have been using for euphony setup. But does faster means better ? And is it a noisy SSD ?  Good is that it has huge memory and it also runs as ECC so would it work better than non-ECC ??

 

[auricgoldfinger]

I've attached the specifications for the Apacer PV310-M280 flash drive in case anyone is interested in learning more about it.

PV310-M280_APPxxxG1DN-01VTD_Spec_rev1.5.pdf

[Energy]

Let's say one decided to get the Apacer SSD with "Built-in hardware ECC". With operating read/write speeds of over 500MB/s, it makes you wonder if going any higher would be beneficial for audio purposes. If speeds matter, the reading speed should be more important as that is mainly what is being used to validate the digital bits. Another thing to think about is whether the built-in hardware ECC in Apacer SSD's is any different than the ECC Engine used for Low-Density Parity-Check (LDPC) used in their M2 lineup. 

 

Having pointed out those things, which would sound better?

1. Apacer SSD (Built-in hardware ECC) with 520MB/s read speed

2. Apacer M2 (LDPC ECC) with 2660MB/s read speed

 

Furthermore, what if someone powered the SSD with an external linear power supply's 5V?

Powering the M2 on the other hand with a 3.3V seems less likely as its attached to the motherboard.

 

Seeing that power consumption on the Apacer M2 goes as high as 1,725 mA, for those barely meeting power requirements from maybe running an AMD 3950X DSD upsampling server, this wouldn't be very ideal.

[dminches]

7 minutes ago, Energy said:

Powering the M2 on the other hand with a 3.3V seems less likely as its attached to the motherboard.

 

Is there a sound advantage to using a PCI card for the M2 as opposed to putting it in the M2 slot on the motherboard?

 

[Energy]

26 minutes ago, dminches said:

 

Is there a sound advantage to using a PCI card for the M2 as opposed to putting it in the M2 slot on the motherboard?

 

 

PCI card to M2 risers make for a longer signal path and relies on two contact points of insertion.

M2 directly connected to the motherboard should sound better unless the chip that controls the PCI-E Gen 3/4 was put closer to the PCI-E card slots and doesn't have to go through the various PCB layers and noise upon reaching the M2 slot, but given multi-layer PCB motherboard designs and proper routing this shouldn't pose a problem.

 

At this point a riser card could be comparable to running an SSD over SATA cable, but in a lesser degree. It's more favorable to use a PCI-E to M2 because the distance isn't nearly as long as with making SATA connections (data/power). But one thing to note that can influence sound quality between the two is the power going into the PCI-E lanes loop through the motherboard (potentially picking up noise) opposed to power coming directly from the ATX PSU and into the SATA drive.

[Iving]

43 minutes ago, dminches said:

 

Is there a sound advantage to using a PCI card for the M2 as opposed to putting it in the M2 slot on the motherboard?

 

 

Been using Optane PCIe AIC for a long time. Ethernet card also in a CPU-direct lane. (No chipset as it were.) Local playback tho - no internet streaming. For storage no external hard drives. No SATA. No nothing. Just Optane PCIe AIC.

[dminches]

31 minutes ago, Energy said:

 

PCI card to M2 risers make for a longer signal path and relies on two contact points of insertion.

M2 directly connected to the motherboard should sound better unless the chip that controls the PCI-E Gen 3/4 was put closer to the PCI-E card slots and doesn't have to go through the various PCB layers and noise upon reaching the M2 slot, but given multi-layer PCB motherboard designs and proper routing this shouldn't pose a problem.

 

At this point a riser card could be comparable to running an SSD over SATA cable, but in a lesser degree. It's more favorable to use a PCI-E to M2 because the distance isn't nearly as long as with making SATA connections (data/power). But one thing to note that can influence sound quality between the two is the power going into the PCI-E lanes loop through the motherboard (potentially picking up noise) opposed to power coming directly from the ATX PSU and into the SATA drive.

 

I should have phrased my question better.  Is using an Optane PCIe instead of an Optane M.2 preferable? It seems like @Iving does this.

 

[Iving]

7 minutes ago, dminches said:

 

I should have phrased my question better.  Is using an Optane PCIe instead of an Optane M.2 preferable? It seems like @Iving does this.

 

 

Definitely!

 

But it's a NUC issue (full height PCIe cards and mobos with sufficient direct lanes etc). I have ethernet playback and my PC isn't a NUC. (I keep watching this thread!)

[dminches]

1 minute ago, Iving said:

 

Definitely!

 

But it's a NUC issue. I have ethernet playback and my PC isn't a NUC. (I keep watching this thread!)

 

I am not sure what you mean "it is a NUC issue."

 

I am running Roon server on a PC with a Supermicro mobo and Xeon E-2124G CPU.  Apacer RAM. Sonore ultraRendu endpoint.

 

 

[Iving]

Just now, dminches said:

 

I am not sure what you mean "it is a NUC issue."

 

I am running Roon server on a PC with a Supermicro mobo and Xeon E-2124G CPU.  Apacer RAM. Sonore ultraRendu endpoint.

 

 

 

sorry - just edited my post - it's about full height cases, nofan cooling for CPU, mobos with sufficient CPU-direct lanes etc - and my system is not USB it's Dante so I need/value a CPU-direct PCIe ethernet card ...

[Iving]

also Optane as PCIe AIC is relatively expensive - but not really all pros and cons considered - and i'm happy to feed my 280Gb Optane prn ...

[dminches]

1 minute ago, Iving said:

 

sorry - just edited my post - it's about full height cases, nofan cooling for CPU, mobos with sufficient CPU-direct lanes etc - and my system is not USB it's Dante so I need/value a CPU-direct PCIe ethernet card ...

 

I don't know what Dante is but I get your point.

 

I am not using any of the PCIe lanes since I am using the onboard ethernet controller.  Thus, I could use the Optane but I would need a ribbon cable since I am using a Streacom FC10 case.