Is this $40 card a SCAM? - PCIe Power Filtering

[ColinLTT]

 

Are you a distinguished audiophile who uses their PC for high-end audio reproduction? Or perhaps you just want the best possible sound to come out of your speakers… so can you smooth out the power delivery of your PCI-e bus with a $40 power filter card?
 

Buy : GRADO SR80e
On Amazon (PAID LINK): https://geni.us/NZ7AB
On Newegg (PAID LINK): https://geni.us/f58onf
On B&H (PAID LINK): https://geni.us/sqJp

 

Buy: AMD Radeon
On Amazon (PAID LINK): https://geni.us/JFON7
On Newegg (PAID LINK): https://geni.us/AAskAat
On B&H (PAID LINK): https://geni.us/SZ7U6

 

Buy: Power Supplies
On Amazon (PAID LINK): https://geni.us/6KymQCl
On Newegg (PAID LINK): https://geni.us/MtvYhG
On B&H (PAID LINK): https://geni.us/NkYBDk

 

Buy: AC Power Filter
On Amazon (PAID LINK): https://geni.us/6Zk0KAs
On Newegg (PAID LINK): https://geni.us/QcJj4
On B&H (PAID LINK): https://geni.us/lJ85

 

Buy: FiiO E10K
On Amazon (PAID LINK): https://geni.us/ANA2i
On B&H (PAID LINK): https://geni.us/ymKKf4
 

[Duckson McQuack]

What a nice match of PSU and motherboard

[Parideboy]

The most unpredictable segway I've ever seen

[ColinLTT]

2 minutes ago, Parideboy said:

The most unpredictable segway I've ever seen

cutest segue ever 

[Derkoli]

Funnily enough, We actually had a discussion about a very similar product over at the audio sub-forum a while ago.

[Hunter259]

That only kinda makes sense if it was on something with a shared bus like PCI but PCIE not at all.

[Nystemy]

A few things that makes a card like this useless is mainly due to:

1. A well made audio section already has its own local capacitors.
2. The amount of noise a capacitor can filter is both dependent on the amount of capacity in the capacitor, but also on the Equivalent Series Resistance (ESR) of the capacitor and the wiring going to it. (More resistance = more noise. (I am also greatly oversimplifying things here...))
3. Having these "extra" capacitors out on a PCIe card adds a lot of additional wiring between the capacitor and the thing we want to isolate from noise. Making it rather ineffective. _{(Not to mention the contact resistance of a PCIe connector isn't all that impressive...)}
4. The PCIe connector only has 12 volt and 3.3 volt rails going to it. (A lot of Audio DACs tends to use the 5 volt rail....)

In short, this card logically shouldn't do much at all to the quality of our audio.
Also:

9 minutes ago, Hunter259 said:

That only kinda makes sense if it was on something with a shared bus like PCI but PCIE not at all.

Yes, the data bus on PCI-express is not a multi drop bus like PCI, but the power section of PCI-express is still shared between all slots. And noise can and will propagate from one slot to the others. After all, we are only interested in the power rails, not the PCIe lanes themselves. (But as stated above, the rail of interest isn't even used by PCIe....)

Adding in a card like this will in general have little to no effect on system noise.
And if one really wants to cut down on the noise, then it is easier to just get an external DAC regardless. (As stated in the video...)

Also @ColinLTT Linus mentioned in the video that "Low or unbalanced power, like we have here in our building." Why doesn't LMG hire a contractor to fix that issue? It can rather easily be done with a line interactive "UPS" solution. _{(It doesn't actually have to be a UPS, but line interactive UPS does power conditioning to regulate the output voltage to what one desires, making things like brownouts and power flickering a thing of the past.) https://en.wikipedia.org/wiki/Uninterruptible_power_supply#Line-interactive}

 

[TehDwonz]

1 hour ago, ColinLTT said:

 

Are you a distinguished audiophile who uses their PC for high-end audio reproduction? Or perhaps you just want the best possible sound to come out of your speakers… so can you smooth out the power delivery of your PCI-e bus with a $40 power filter card?
 

Did you see the DIMM versions of that? Ignoring their intended use, I wondered if they might help with extreme RAM overclocking situations? ie, on a 4 DIMM board, with 2 actual DIMMs, then 2 of the Totally Not Dodgy filter/capacitor DIMMs in the two remaining slots. Could be interesting to test. I wasn't willing to risk killing my motherboard over it though...

 

edit: these: https://www.aliexpress.com/item/4000885276971.html

Forget the Hi-Fi and filtering nonsense - I am thinking from a "preventing voltage drop" kinda angle, as well as potentially terminating the shared address/data bus to the DIMMs.

[bloodaxe222]

will ABC book for Little Gamer be available in lttstore.com??

[soldier_ph]

Before I watch the video: Yes, yes it's a scam if it isn't Highend.

[Nystemy]

6 minutes ago, TehDwonz said:

Did you see the DIMM versions of that? Ignoring their intended use, I wondered if they might help with extreme RAM overclocking situations? ie, on a 4 DIMM board, with 2 actual DIMMs, then 2 of the Totally Not Dodgy filter/capacitor DIMMs in the two remaining slots. Could be interesting to test. I wasn't willing to risk killing my motherboard over it though...

Technically speaking, there is potential for a better overclock.
One of the main limits to how fast a DDR memory bus can run is partly down to the amount of noise on the memory voltage rail.
One reason some motherboards with 2-3 phase memory VRMs tends to perform better. _{(With more phases, one can get reduced voltage ripple, and in turn have an easier time filtering out the noise.)}

I would though not suspect that it would have any major impacts to be fair. The distance is far too great for it to have much of any effect to be fair. So at that point, it is likely better to add more capacitors to one's RAM modules. _{(And that is even more sketchy than adding in a card stuffed full of capacitors...)}

[TehDwonz]

10 minutes ago, Nystemy said:

A few things that makes a card like this useless is mainly due to:

The fact that there is a version for your HDD should tell you all you need to know
https://www.aliexpress.com/item/4001051365842.html

 

[TehDwonz]

2 minutes ago, Nystemy said:

Technically speaking, there is potential for a better overclock.
One of the main limits to how fast a DDR memory bus can run is partly down to the amount of noise on the memory voltage rail.
One reason some motherboards with 2-3 phase memory VRMs tends to perform better. _{(With more phases, one can get reduced voltage ripple, and in turn have an easier time filtering out the noise.)}

I would though not suspect that it would have any major impacts to be fair. The distance is far too great for it to have much of any effect to be fair. So at that point, it is likely better to add more capacitors to one's RAM modules. _{(And that is even more sketchy than adding in a card stuffed full of capacitors...)}

Yeah, I was also thinking about distances. The reason we use the furthest 2 RAM slots for best results, it that the data/addr lines, as well as clock, need to be terminated right? Otherwise you can get reflections. But, if you put the dummy DIMMs in the furthest slots, you could potentially get away with using the closer DIMM slots for the actual RAM - reducing the signal path length. It'd be marginal at best, but I'd be interested in seeing a competent RAM overclocker test it out.

[ColinLTT]

53 minutes ago, Nystemy said:

Also @ColinLTT Linus mentioned in the video that "Low or unbalanced power, like we have here in our building." Why doesn't LMG hire a contractor to fix that issue? It can rather easily be done with a line interactive "UPS" solution. _{(It doesn't actually have to be a UPS, but line interactive UPS does power conditioning to regulate the output voltage to what one desires, making things like brownouts and power flickering a thing of the past.) https://en.wikipedia.org/wiki/Uninterruptible_power_supply#Line-interactive}

 

Just stating the facts when it comes to the building power - the majority of our PC's have either individual UPS's or otherwise.  

[Nystemy]

8 minutes ago, TehDwonz said:

Yeah, I was also thinking about distances. The reason we use the furthest 2 RAM slots for best results, it that the data/addr lines, as well as clock, need to be terminated right? Otherwise you can get reflections. But, if you put the dummy DIMMs in the furthest slots, you could potentially get away with using the closer DIMM slots for the actual RAM - reducing the signal path length. It'd be marginal at best, but I'd be interested in seeing a competent RAM overclocker test it out.

 

Using the two furthest away slots is logical in a daisy chained setup.
Since the longer traces will have a larger effect on our signal than the shorter "traces" of our unused connector.

In general, a motherboard that only has one slot per channel is going to have the best signal quality. (If it has equally well routed traces as our multi slot per channel board.)

But distance is a funny thing when it comes to electrical signals.
If we were to add on a trace that is exactly 1/4 the length of our signal of interest, then the reflection will come back 180 degrees out of phase, and thereby cancel out our signal. (Preventing our signal from passing.)

 

_{(The calculation is 15.09 divided by the length of our trace in cm, and this gives us the frequency that gets blocked by that trace, on FR4 PCB material. This is called a quarter wave stub, a very effective notch filter. (And for everyone asking why we divide 15.09 by the length of our trace, that is because 15.09 is the propagation speed of a signal in RF4, its 50.3% the speed of light. And the speed of light is 30 cm per ns. (this is accurate to within 0.06923%)))}

Now, the very very short piece of metal that forms the pin in our connector isn't going to result in much interference, if it's 4 mm long, then it will only start effecting signals up at around 9.43 GHz.

But if we were to use the other slot in our daisy chain instead, leaving the far off slot unused, then we also will likely have about 10 mm of PCB trace to contend with. Bringing our rough estimate up to 14 mm of "trace". In other words, it will block signals at 2.695 GHz, this is much much lower.

Also, do note, 2.4 GHz RAM might struggle here, even if its "well bellow" 2.695 GHz, since our signal's frequency content is actually larger than just 2.4 GHz, our signal isn't just spending time transitioning from one state to the next, but it also idles around at each state for a bit. This means that our transitions are actually faster than 2.4 GHz.

Now, some motherboard vendors go with Y configurations instead.
But that just means that regardless of what slot we use, will would then have about 9 mm of trace to reflect signals with... Ie, it would block stuff at 4.19 GHz. (Likely still good enough for 3200 MHz RAM to be fair.) Though, the advantage with the Y configuration is that memory timings becomes much easier to work with. Since the length difference to the two modules becomes the same, not that this technically has much effect on memory timings.... But as fast as we have two modules, both connectors are terminated properly and will not reflect sufficiently for it to interfere with our data.

[Murasaki]

Ladies but mostly gentlemen, the PCI/PCI-E registry cleaner with 8 types of capacitors.

[Nystemy]

9 minutes ago, ColinLTT said:

Just stating the facts when it comes to the building power - the majority of our PC's have either individual UPS's or otherwise.  

Guess that is the downside of 120 volt mains.
More current, more voltage drops, and generally higher costs for the extra copper.

Though, living in Europe myself where 230 _(+/-20) volts is the standard, we have a bit more sparking to contend with instead. _{(It is though not all that problematic, and its nice to be able to draw 2500+ watts from almost any outlet without issue...) Though, inrush current on modern power supplies can trip breakers if the power supply is poorly designed, or one simply has "too many" of them on the same circuit. (After all, the energy stored in a capacitor is equal to E = F*V²/2 So doubling the voltage, means it will have to charge 4x more energy into the caps (that also becomes even more deadly if one touches them), not to mention that the inrush current is I = V/R, so the inrush current is also twice as big.... And with the higher mains voltage, we could make due with a capacitor of about a quarter of the size...)}

[sub68]

this is for the audiophile that cares about there setup too much

the same person that buys special cables just to power there units

 

[tehj15]

First off I agree with the consensus this is snake oil.

 

It looks like it has a 1x slot on the reverse side, is that so you can install it in a 1x slot or to hookup a daughter board?

 

So I don't think this will/would work regardless but in theory wouldn't you have to install it in the top slot?

[TempestCatto]

A power conditioner actually works: https://www.sweetwater.com/store/detail/M8x2--furman-m-8x2-power-conditioner

 

Though, this is the kinda thing you mainly use when inside a building with horrid power. For example, it prevents that awful hum you can get  in your audio system when a refrigerator kicks on.

[SolarNova]

Still rocking a Asus Xonar D2X soundcard here

 

[Luscious]

4 hours ago, ColinLTT said:

Just stating the facts when it comes to the building power - the majority of our PC's have either individual UPS's or otherwise.  

I was going to say any PC of value should be connected to a UPS as those condition the power quite well. I have a 120V 2200VA APS hooked up to my rig that uses a 1600W PSU. The big issue in my building is flaky voltages so the UPS definitely helps smooth out the power and even kicks in when a brief voltage drop happens. More importantly, my monitor, NAS, router/switch and land line are also run on a UPS so they are protected as well in case of a power outage.

 

Regarding audio, I use the optical output from my mobo to connect to a TEAC UD-503 desktop DAC that I can use with either my Sony XB700 cans or my Creative T40 Series 2 speakers. I find it's a really great setup given that connecting headphones can be an issue when the front panel audio connectors on a lot of cases are simply junk or the FPaudio cables going to your mobo are noisy.

[Egg-Roll]

The reason why it didn't work: Linus dropped it and broke it.

 

All seriousness tho I watched the video just hoping he dropped it to make that comment.

 

Also when is Linus going to actually write a book about himself like every other Youtuber?

[WereCatf]

This was, quite frankly, just a terrible and lazy video.

 

Coil whine doesn't happen because of bad power-quality or anything like that; it's caused by somewhat loose windings in the coil. The only way you can change that is by swapping the inductor for another one or changing the PWM-signal. Also, Linus is supposed to be able to tell the difference with his ears? You never bothered to actually check the output in any sort of an oscilloscope or audio-software or anything? Like...what the fuck, guys? Looking at Linus listening to stuff doesn't make for an informative or a particularly engaging video, but you'd have had something actually meaningful to show your viewers, if you'd bothered to actually check the signal properly!

[Nystemy]

1 hour ago, WereCatf said:

This was, quite frankly, just a terrible and lazy video.

 

Coil whine doesn't happen because of bad power-quality or anything like that; it's caused by somewhat loose windings in the coil. The only way you can change that is by swapping the inductor for another one or changing the PWM-signal. Also, Linus is supposed to be able to tell the difference with his ears? You never bothered to actually check the output in any sort of an oscilloscope or audio-software or anything? Like...what the fuck, guys? Looking at Linus listening to stuff doesn't make for an informative or a particularly engaging video, but you'd have had something actually meaningful to show your viewers, if you'd bothered to actually check the signal properly!

Coil whine can be caused by a few different factors.
The primary ones is general resonance, even if the wires are glued down, they can still vibrate. Not to mention that magnetic leakage out of the core itself can make nearby metals act a bit like a speaker. Though, usually it isn't the actual power switching we hear, since most VRMs work in the 100+ KHz range. Well outside of human hearing.
But what we can hear is when the VRM cycles between different loads. (Like each time it changes work load. For an example when it goes from mapping polygons to actually texturing, or when it goes from one frame to the next. This will create a tone that is some multiple of the FPS that our GPU produces, + harmonics of course.)

But "coil whine" can also come from capacitors. When one charges up a capacitor, the inner "plates" within it will attract each other by a tiny bit, and when the potential drops back down, it eases of that pressure. Now, this wouldn't be expected to be of any major effect in a GPU or PC in general. But in power electronics, it is a thing to be mindful of. _{(Though, with enough power, even conductors can sing. And some capacitor banks for pulsed lasers are specially designed to not explode when they discharge.)}

But yes, coil whine isn't indicative of bad power quality, and rarely a sign that there is something wrong with the product to be fair. All VRMs have coil whine, its just a question if one can hear it or not. The more loose the components are, the louder it will be. But thermal pads dampens the noise rather effectively. _{(And that is the main reason we want to see thermal pads connecting the inductors to a heat sink. Just to limit coil whine, not to cool the inductors, they can happily operate at 140 C and above.)}

In terms of noise checking, this can be done with an oscilloscope, but generally, one wouldn't use such for Audio. (Unless one is really cheap...)
Oscilloscopes aren't actually having all that great dynamic range, so testing audio equipment with them isn't something that they are good at. One can though measure power rail noise/ripple. (For measuring the output signal we would rather want to use a dynamic signal analyzer. (Much lower bandwidth than Oscilloscopes, but far superior dynamic range.))