SSD Performance Degradation Caused by File Fragmentation Tested

[semente]

@yamamoto2002 is it possible to measure the current draw of reading from RAM?

[PeterSt]

9 minutes ago, semente said:

@yamamoto2002 is it possible to measure the current draw of reading from RAM?

 

I hope Yamamoto will say "sure", but I think this is way too complex, already because it involves the OS itself. However, various WMI Instrumentation resources are at our disposal to show derivates of it and eliminate the irrelevant (depending on how smart we are at setting the WMI stuff).

 

On the large scale we will be able to see it (use a power meter) but this will be too rough if you not first set up all there is to do to this regard.

But for example, if I had to load a 350MB CD into memory, then with contiguous memory it could go in one second (this includes the source as a bottleneck - e.g. Network), while with not contiguous memory it may take 10 seconds. Mind you, this is only the writing part.

Reading I don't know because nobody reads 350MB CDs as fast as possible from RAM in the audio environment, but this could be tested of course. I expect the reading time to be 20 times faster at least (wild guess but alas).

 

Before anyone starts working on this: please don't. It is useless. No audible differences will come from this anyway.

Well, that's what they say, right ?

[manueljenkin]

1 minute ago, PeterSt said:

 

I hope Yamamoto will say "sure", but I think this is way too complex, already because it involves the OS itself. However, various WMI Instrumentation resources are at our disposal to show derivates of it and eliminate the irrelevant (depending on how smart we are at setting the WMI stuff).

 

On the large scale we will be able to see it (use a power meter) but this will be too rough if you not first set up all there is to do to this regard.

But for example, if I had to load a 350MB CD into memory, then with contiguous memory it could go in one second (this includes the source as a bottleneck - e.g. Network), while with not contiguous memory it may take 10 seconds. Mind you, this is only the writing part.

Reading I don't know because nobody reads 350MB CDs as fast as possible from RAM in the audio environment, but this could be tested of course. I expect the reading time to be 20 times faster at least (wild guess but alas).

 

Before anyone starts working on this: please don't. It is useless. No audible differences will come from this anyway.

Well, that's what they say, right ?

Yes, the internal probes for checking power consumption are mainly for just basic diagnostic purposes. I don't know how accurate they are and how fast it samples. If one were to insert a high quality ADC for each of these, each component would probably become 2-3x the price they currently are 😅

[PeterSt]

7 hours ago, March Audio said:

So, what audio playback software doesn't load the file into memory before playback?  An SSD will read the file in a fraction of a second.  How does this power consumption that happened prior to playback affect the sound?

 

As I told elsewhere. Investigate how the Windows OS works (I said literally the same thing upon the same question). But let me rephrase the question so goal post will stay where they are (this is a thread in the objective area).

 

How does this power consumption that happened prior to playback keep recurring during playback?

 

Btw, I don't have the idea to act as a teacher here. I am not denigrating anything either.

[idiot_savant]

With DDR, probably impossible to say - it’s constantly being refreshed ( because otherwise it forgets stuff ), and the CPU is more or less constantly reading it anyway

https://en.m.wikipedia.org/wiki/Memory_refresh

 

your friendly neighbourhood idiot 

[PeterSt]

8 minutes ago, manueljenkin said:

If one were to insert a high quality ADC for each of these, each component would probably become 2-3x the price they currently are 😅

 

... And the point I was trying to make is that you won't be able to separate the "activity" in there from the normal OS operations. Unless ... unless you made the memory contiguous and know where it sits as well (and then read from one DIMM - the correct one).

Quite undoable, although everything can be done (IMO).

[PeterSt]

2 minutes ago, idiot_savant said:

With DDR, probably impossible to say - it’s constantly being refreshed ( because otherwise it forgets stuff ), and the CPU is more or less constantly reading it anyway

 

that

[idiot_savant]

Whoops! I can’t read

 

 

your friendly neighbourhood idiot 

[PeterSt]

4 minutes ago, idiot_savant said:

Whoops! I can’t read

 

Finally you could. 😉

[manueljenkin]

13 minutes ago, idiot_savant said:

With DDR, probably impossible to say - it’s constantly being refreshed ( because otherwise it forgets stuff ), and the CPU is more or less constantly reading it anyway

https://en.m.wikipedia.org/wiki/Memory_refresh

 

your friendly neighbourhood idiot 

Sorry to be pedantic, it doesn't forget, the charges leak (its stored in a capacitor like structure that happens to have leakages).

[March Audio]

1 hour ago, PeterSt said:

 

Yes, and if someone pus forward a very well conducted test - just that - nobody should jump in "but that is not audible anyway !".

Did I read wrongly ?

 

And so it happens.

No one did that.

[idiot_savant]

And the sum effect if you don’t refresh is…?

 

 

Your friendly neighbourhood idiot 

[March Audio]

34 minutes ago, PeterSt said:

 

As I told elsewhere. Investigate how the Windows OS works (I said literally the same thing upon the same question). But let me rephrase the question so goal post will stay where they are (this is a thread in the objective area).

 

How does this power consumption that happened prior to playback keep recurring during playback?

 

Btw, I don't have the idea to act as a teacher here. I am not denigrating anything either.

 

It doesnt matter.  You are investigating the wrong thing.

 

A PC doesn't stop working.  Busses are still whizzing away, drives are still accessed, as is memory. CPU cycles up and down in load and clock speed. Dozens of processes are still running.  That will continue throughout playback whatever you do.

 

What matters is what appears in the dac output.

 

Trust me, you are not teaching anybody anything 😉

 

[PeterSt]

11 minutes ago, idiot_savant said:

And the sum effect if you don’t refresh is…?

 

Silence ?

:-)

[idiot_savant]

@PeterSt - the point is, it’s a memory that needs refreshing. If you don’t refresh it, it “forgets” what was in it. The fact this is by charge leakage is a detail, which I’m sure is going to somehow come into @manueljenkin big reveal. Which, unfortunately won’t help. 

 

 

your friendly neighbourhood idiot 

[PeterSt]

27 minutes ago, March Audio said:

It doesnt matter.  You are investigating the wrong thing.

 

A PC doesn't stop working.  Busses are still whizzing away, drives are still accessed, as is memory. CPU cycles up and down in load and clock speed. Dozens of processes are still running.  That will continue throughout playback whatever you do.

 

What matters is what appears in the dac output.

 

Trust me, you are not teaching anybody anything 😉

 

Funny, how you changed this post in two places at least. Funny=good in this case !

[PeterSt]

17 minutes ago, idiot_savant said:

The fact this is by charge leakage is a detail, which I’m sure is going to somehow come into @manueljenkin big reveal. Which, unfortunately won’t help. 

 

Why is that ?

[idiot_savant]

@PeterSt - well, let’s see what he has to say shall we?

 

 

 

 

your friendly neighbourhood idiot 

[yamamoto2002]

2 hours ago, semente said:

@yamamoto2002 is it possible to measure the current draw of reading from RAM?

 

For Intel Sandy Bridge or later model of CPUs, there is a RAPL MSR (Running Average Power Limit Model Specific Registers) CPU registers to read DRAM energy consumption.

[The Computer Audiophile]

8 hours ago, pkane2001 said:

My personal desire (and indeed, the charter of this forum) is that someone provide objective evidence that SSD reads, fragmented vs contiguous, have any effect on an otherwise well-designed, well-measuring DAC.

I’d also be interested in an effect on an otherwise poorly designed, poor measuring DAC. If an effect can be shown, it could help us understand more and could provide information for people who like certain DACs. 

[Racerxnet]

2 hours ago, yamamoto2002 said:

 

For Intel Sandy Bridge or later model of CPUs, there is a RAPL MSR (Running Average Power Limit Model Specific Registers) CPU registers to read DRAM energy consumption.

Is there a possibility to read the DRAM voltage variances in the bios? With the Bios mods for the Gigabyte boards this may be possible. 

[The Computer Audiophile]

We should also keep in mind all the variables involved here. Taking these measurements and transferring the results to other computers, SSDs, firmware versions on the computers and SSDs, SSD types SLC, MLC, TLC, is fraught with possible unscientific issues. 
 

It would be like saying you measured the way one DAC chip turned digital into analog, so they must all work this way. 

[yamamoto2002]

50 minutes ago, Racerxnet said:

Is there a possibility to read the DRAM voltage variances in the bios? With the Bios mods for the Gigabyte boards this may be possible. 

 

I don't know sorry. It seems there is a fixed voltage value (such as 1.2 V for DDR4 memory) that can be taken somehow.

 

BTW I found DRAM power consumption value on OpenHardwareMonitor.

 

[yamamoto2002]

1 hour ago, The Computer Audiophile said:

I’d also be interested in an effect on an otherwise poorly designed, poor measuring DAC. If an effect can be shown, it could help us understand more and could provide information for people who like certain DACs. 

 

I had had Roland UA-30 on Windows 2000 era. Sound pitch fluctuates when CPU becomes busy 😅

 

[manueljenkin]

35 minutes ago, The Computer Audiophile said:

We should also keep in mind all the variables involved here. Taking these measurements and transferring the results to other computers, SSDs, firmware versions on the computers and SSDs, SSD types SLC, MLC, TLC, is fraught with possible unscientific issues. 
 

It would be like saying you measured the way one DAC chip turned digital into analog, so they must all work this way. 

There's more. There seems to be more than one type of read and write circuits having completely different methods electrically. When you get into sd cards it gets even more complex, almost none of them follow a standard it seems.

 

The question would be, do the dacs that don't show measurable or audible differences with system tweaks, have actually better fidelity or are they just decorrelating the upstream noise while adding noise of their own (again sine squiggle charts and fft charts are inconclusive in this aspect).