DIY Project High Performance Audio PC with high quality wiring

[Zauurx]

On 1/23/2021 at 5:28 PM, StreamFidelity said:

- Receive Side Scaling (RSS)
RSS is enabled by default for best networking performance. It is best to set the maximum number of RSS queues to be equal to the maximum number of RSS CPUs.

 

I did it, without having read this post, for my 2 intel cards integrated in my Gigabyte motherboard.
On the other hand, I chose to limit to 1 RSS queue on the entry card and 2 queues that which goes to the NAA player.
HQPlayer uses core 0 and 3 (ASDM7EC). Only core 1/2 and 4/5 handle network interrupts.
Choice validate after LatencyMon tests

 

Name                                            : Ethernet
InterfaceDescription                            : Intel(R) Ethern
Enabled                                         : True
NumberOfReceiveQueues                           : 1
Profile                                         : Conservative
BaseProcessor: [Group:Number]                   : 0:2
MaxProcessor: [Group:Number]                    : 0:2
MaxProcessors                                   : 1
RssProcessorArray: [Group:Number/NUMA Distance] : 0:2/0

Name                                            : USBridge
InterfaceDescription                            : Intel(R) I211 G
Enabled                                         : True
NumberOfReceiveQueues                           : 2
Profile                                         : Conservative
BaseProcessor: [Group:Number]                   : 0:4
MaxProcessor: [Group:Number]                    : 0:5
MaxProcessors                                   : 2
RssProcessorArray: [Group:Number/NUMA Distance] : 0:4/0  0:5/0

 

[StreamFidelity]

13 hours ago, Zauurx said:

On the other hand, I chose to limit to 1 RSS queue on the entry card and 2 queues that which goes to the NAA player.

 

Interesting. This is somewhat in line with Taiko Audio's philosophy of assigning precisely defined tasks to specific cores. Did you do it with Process Lasso?

[Zauurx]

On 3/10/2021 at 8:27 AM, StreamFidelity said:

 

Interesting. This is somewhat in line with Taiko Audio's philosophy of assigning precisely defined tasks to specific cores. Did you do it with Process Lasso?

 

Yes with ProcessLasso but you can do almost the same with taskmanager.

And I use other utilities ..
Have you noticed how the core 0 is always more in demand ?

[StreamFidelity]

23 hours ago, Zauurx said:

Have you noticed how the core 0 is always more in demand ?

 

In fact, Core 0 is the busiest. The settings in the Task Manager are only of a temporary nature, as they have to be set again after each restart.

 

For the Solarflare SFN8522 NIC the manual says:

 

"RSS Base Processor
For a Solarflare network adapter, the RSS base processor is 0 (zero), which means it starts processing on CPU core 0 which is also the default processor for all other general Windows processes and will likely be the default for all other network adapters in the server.
To avoid this unnecessary contention, set the adapter RSS base processor to another processor on the NUMA node the Solarflare adapter is assigned to."

 

In HWMonitor it was easy to see that from Core 6 onwards there was a lower load. I set this value. It seems to be working. 👍

 

 

[Zauurx]

Yes, and it is especially LatencyMon that intrigued me with this imbalance of interruptions between cores.
I tried to distribute but windows itself is not flexible.

[StreamFidelity]

Easter greetings from Berlin 😉

 

 

 

[Topk]

5 hours ago, StreamFidelity said:

Easter greetings from Berlin 😉

 

 

 

So you left the clear thick lid on in the end? Looks nice!

[StreamFidelity]

44 minutes ago, Topk said:

So you left the clear thick lid on in the end? Looks nice!

 

Thank you. Yes, the acrylic glass with a thickness of 10mm breaks the light very nicely.

[Nenon]

For those aiming at the lowest possible latency, why not try the Cisco Nexus 3548 (or 3524) switch. It's the industry standard for low latency (190-250 nanoseconds) switching without going to layer 1 switches. 

[StreamFidelity]

Another report for the latency minimizer. 😁

 

A tip from this forum brought me to the Solarflare XtremeScale X2522 NIC. The front looks almost like the Solarflare SFN8522 NIC.

 

 

The back then shows a few more chips.

 

 

As usual, the firmware was not up to date. I did that right away.

 

 

The BIOS menu was a surprise. The firmware can be set according to the purpose. I chose Ultra low latency.

 

 

The connection to the NAS and T + A NAA was established quickly. Unfortunately, the joy was short-lived. The card switched itself off after about 5 minutes. Unfortunately a heat problem. Despite the fan-optimized acrylic glass cover, operation is only possible without a cover. So it has been running stable for days now.

 

About the sound: Like the SFN8522 NIC, the sound is three-dimensional, more details can be heard. No sharpness and a deeply graduated bass. Of all these good features, with the X2522 Nic improves a bit. There are no worlds, but if you want the best ... 👍

[Gavin1977]

What SFP transceivers are you using with the X2522?

[StreamFidelity]

20 minutes ago, Gavin1977 said:

What SFP transceivers are you using with the X2522?

 

Transceiver Startech 1000BASE-EX SFP -SM LC-40KM 👍

[Exocer]

On 4/25/2021 at 11:29 AM, StreamFidelity said:

About the sound: Like the SFN8522 NIC, the sound is three-dimensional, more details can be heard. No sharpness and a deeply graduated bass. Of all these good features, with the X2522 Nic improves a bit. There are no worlds, but if you want the best ... 👍

Excellent report. Is there a specific license we should look for as well?

 

Thanks,

-Rob

[Topk]

On 4/25/2021 at 5:29 PM, StreamFidelity said:

Another report for the latency minimizer. 😁

 

A tip from this forum brought me to the Solarflare XtremeScale X2522 NIC. The front looks almost like the Solarflare SFN8522 NIC.

 

 

The back then shows a few more chips.

 

 

As usual, the firmware was not up to date. I did that right away.

 

 

The BIOS menu was a surprise. The firmware can be set according to the purpose. I chose Ultra low latency.

 

 

The connection to the NAS and T + A NAA was established quickly. Unfortunately, the joy was short-lived. The card switched itself off after about 5 minutes. Unfortunately a heat problem. Despite the fan-optimized acrylic glass cover, operation is only possible without a cover. So it has been running stable for days now.

 

About the sound: Like the SFN8522 NIC, the sound is three-dimensional, more details can be heard. No sharpness and a deeply graduated bass. Of all these good features, with the X2522 Nic improves a bit. There are no worlds, but if you want the best ... 👍

Is the X2522 more or less warm/analogue sounding compared to SFN8522?
 

Also can you confirm X2522 can’t be used with acrylic lid with holes? That could be a deal breaker for many. Thanks!

[StreamFidelity]

16 hours ago, Exocer said:

Excellent report. Is there a specific license we should look for as well?

 

Thank you. It's only one Type of the Solarflare XtremeScale X2522 NIC avaliable.

[StreamFidelity]

13 hours ago, Topk said:

Is the X2522 more or less warm/analogue sounding compared to SFN8522?

 

For me, X2522 has a little more analog sound character than SFN8522. But the difference is small. SFN8522 is already very good.

 

13 hours ago, Topk said:

Also can you confirm X2522 can’t be used with acrylic lid with holes?

 

Yes! At the moment, only an open case is an option. I will have to redesign the acrylic glass lid. 😉

[basillus]

On 4/25/2021 at 5:29 PM, StreamFidelity said:

Unfortunately, the joy was short-lived. The card switched itself off after about 5 minutes. Unfortunately a heat problem. Despite the fan-optimized acrylic glass cover, operation is only possible without a cover. So it has been running stable for days now

I don`t know the size of the x2522 heatsink, but the resolution can be to make a heatsink with heatpipes connected to the unused heatsink in Your cabinet.

Maybe this: https://www.aliexpress.com/item/33054815691.html or get one in copper or the best solution draw a heatpipe heatsink there fits the x2522 card and make “MoreSuns” to make a customsized heatpipe heatsink.    ;0)

[MarcelNL]

Great find on those heatsinks! I started looking but did not yet get to ALi ....

 

[mrkoven]

What is everyone's thoughts on the latest 500W ATX LPS by HDPlex? I am thinking of going this route instead of DC-ATX 400W + Keces P8. It has 2 DC output for USB cards and whatnot, so seems like a solid all-in-one solution for simplicity and convenience (?)

[Gavin1977]

I have one on order, delivery in May I hope.  I have been waiting for ages.  Apparently it's good, but I'm not aware of any reviews (or more importantly comparisons) out there.

[StreamFidelity]

 

Test Taiko Audio 500W DC DC-ATX converter

 

Low residual ripples are beneficial in terms of sound. And the Taiko Audio 500W DC DC-ATX converter shines with a ripple noise of 0.10mV (millivolt). For comparison: the HDPLEX 400W DC-ATX converter has a ripple noise of 10mV (millivolt). The specifications of the Taiko Audio DC-DC ATX converter can be read here.

 

first impressions

 

The rather high weight surprised me, which is probably due to the sturdy copper plate.

 

Wherever you look, the workmanship and component quality are first class.

 

The gold-plated Molex pins are clearly visible, the contact quality of which is an excellent match for the fis audio PC cable.

 

I assume that there are wooden panels (Panzerholz?) In between to combat resonance, which would be typical for Taiko Audio.

 

The Taiko Audio 500W DC DC-ATX converter was easy to install because the four drill holes fit like before with the HDPLEX 400W DC-ATX converter.

 

Test constellation

 

First, I use the Taiko Audio 500W DC DC-ATX converter as it was previously the case with HDPLEX. This means that the CPU is supplied directly by a Keces P8. A circuit diagram:

 

first sound impression

 

For the first soundcheck, I like to use Shemekia Copeland. A nice rock blues.

 

If the instrument separation is bad and sharpness creeps into the electric guitars, I know that something is wrong. That was not the case here at all. I don't think Shemekia has sounded so good in my system. I usually wait for the components to burn in. But here the SQ increase is already so obvious. 👍

 

But of course I will continue testing. Next I want to connect the CPU to the Taiko Audio 500W DC DC-ATX converter.

[OAudio]

8 hours ago, StreamFidelity said:

And the Taiko Audio 500W DC DC-ATX converter shines with a ripple noise of 0.10mV (millivolt).

Hi @StreamFidelity,

 

That is an interesting writeup thanks for posting. Its sound as thought the DC ATX is working very well in your system.

 

The measurement you quoted is interesting, did you make it directly yourself ? If so I wonder if you can say what the conditions of the measurement were ? eg with a DVM an oscilloscope or analyser,  and what were the load conditions for the measurement ?

 

Thanks,

 

OAudio.

 

 

 

[StreamFidelity]

1 hour ago, OAudio said:

The measurement you quoted is interesting, did you make it directly yourself ?

 

No - You can find the measured values at Taiko Audio 😉

 

Taiko Audio 500W DC DC-ATX converter (specification sheet)

[OAudio]

52 minutes ago, StreamFidelity said:

 

No - You can find the measured values at Taiko Audio 😉

 

Taiko Audio 500W DC DC-ATX converter (specification sheet)

Thanks.

Nice ripple specs.

[StreamFidelity]

Today was a test with the full power supply from the Taiko Audio 500W DC DC-ATX converter. This means that only one Keces P8 is used.

The Keces P8 can deliver 20V / 8A and in my constellation has a maximum consumption of 5.5A if all 8 cores really run up to 4.2GHz. So there are enough power reserves available.

 

I played the first few songs and had trouble hearing any differences at all. But something was bothering me. After a few repositioning, the realization came to this song:

 

"No Sanctuary Here" by Chris Jones. The bass guitar is used right at the beginning. With the full Taiko ATX power supply, the bass is weaker and less staggered in depth. And the space is also narrower.

 

Too bad. I was already hoping to save a second Keces P8. What could be the causes?

 

With two Keces P8 I naturally have more power reserves. Here are the consumptions for comparison:

 

A) 2 Keces P8 20.2V 2.0A + P8 12.1V fluctuating 1.6A - 3.5A, max 12.1V / 5.5A with CPU full load

B) 1 Keces P8 20.1V fluctuating 2.1A - 4A, max 20.1V / 5.3A with CPU full load

 

Perhaps the performance of the Taiko Audio 500W DC DC-ATX converter would be better if a better LPS in the format of a Paul Hynes SR7 (19vdc @ 10A continuous, 40A transient) or Sean Jacobs DC4 (400VA) provided the power from the start.

 

Summary

 

While the replacement of the HDPLEX 400W DC-ATX converter by the Taiko Audio 500W DC DC-ATX converter was immediately audible, I had to prick up my ears when changing the CPU power supply. "Unfortunately" the direct CPU power supply from a Keces P8 with 12V is still better than the complete power supply from the Taiko ATX. That can change with a better LPS.