Passive cooling on CPU w/ integrated GPU in HTPC possible?

[njekt]

I'm building a HTPC. I'd like to do so without using any fans whatsoever - passive radiators only! I'm trying to minimize the noise introduced in the conductors of my soundcard (which is an ASUS Essence STX II 7.1 + daughter board), this means minimizing all local electric, magnetic and electromagnetic fields as much as possible (read fans, among other things).

Furthermore, FYI:
· I will not be including a dedicated GPU (will be using integrated graphics);
· onboard audio and network will be disabled (not sure if these affect processor demands);
· other components in my case are a bluray drive and a SSD;
· PC will be built inside a horizontal ATX case (no benches despite their superior cooling);
· I'm impartial to socket type and MoBo form factor so long as a MoBo exists with that socket and one PCIexpress v1.0 1x slot.
· the maximum simultaneous demand on the processor would be that of playing a video while upscaling it from 480->1080 on the fly on Windows8.1 with uTorrent and XMBC running.

I would appreciate that those with experience on relevant objectives comment with whatever knowledge they can offer, hopefully maybe even confirm such an objective is possible and recommend a CPU they've had success with.

Thanks for your time!

 

[anbello262]

I'm quite sure that's completely doable. Last year Tom's made an article about a fanless PC ( http://www.tomshardware.com/reviews/silent-pc-gaming-performance,3435.html ) , and with this year's highly efficient new products, I have no doubt you will achieve this. I can't help you very much, though, but I can recommend going with a low end intel CPU, like a Pentium or maybe even an i3, with a big passive heatsink. Pentium processors would have no problem with upscaling video, so it would be more than enough. A low end intel integrated GPU should be enough for your needs.
Another option would be an AMD low end APU, but those are usually hotter than intel's.

Seasonic's X series have fanless options among their low wattage PSUs, so that might be useful as well.

 

[njekt]

Thanks Anbello. Its amazing what can be missed if googling a not exactly relevant set of words. The article you linked definitely proves my objective is obtainable.

With regards to your comment on Seasonic PSUs, you meant the low end of the platinum series (400w,460w,520w). Reviews on the 400w and 520w exposed a major flaw which prevents their use for my objective. PCIexpress slots are powered by a 12V rail and on this series of PSUs 12V rail ripple is nightmarish- HardOCP measured a 30mv ripple (deviation, not peak to peak) at 300w load on the 400w PSU. A 0.25% variation! In the worst case scenario this could add as much as -52.04db of noise. Of course there are going to be countermeasures against 12V rail ripples on the Essence STX II. But even with countermeasures, that PSU will be the greater source of noise as the Essence STX II DAC has THD+N of just 0.0000631% (-124db) -- 4000 times less. For a 300w load, a PSU with a ~4mv ripple on the 12V rail should be obtainable.

edit: Grammar

 

[anbello262]

Oh, didn't really know that about those PSUs o.o That's quite interesting...
I can't help you much more, but at least there is evidence that this can be done. If you can find a better PSU, then it's all ready =D

 

[mlga91]

Most passive coolers are huge chunks of metal, you'll need to make sure that it will fit in your case, as for the power supply, take a look at this, maybe can be of help:

http://www.jonnyguru.com/modules.php?name=NDReviews&op=Story&reid=207

 

[hunter315]

Doing a full silent PC is a bad idea, all "passively" cooled components still require air movement to cool them, this either needs to be forced(a fan) or convection(intake bottom vent top), with your horizontal case you won't have either of these options.

As for your concern about the PSU, thinking 30 mV of ripple on the 12V rail shows you have clearly lost all actual sense when it comes to this issue. 30 mV is actually damn good for a PSU, top end ones can stay below 20 mV but those are the really big ones with lots of filter capacitors. You also need to realize that the sound card is going to take that power in and step it down to voltages it can use, it is going to have capacitors on the input and capacitors on the output, the noise from the PSU isn't even going to make it through that. It also won't have ANY impact on SOUND quality, the ripple on a PSU is generally 100 kHz and up, sometimes its up around 1 MHz, you aren't going to be hearing anything over about 18 kHz so it really doesn't matter.

Also, what makes you think that "minimizing all local electric, magnetic and electromagnetic fields as much as possible (read fans, among other things)" will actually have an impact on anything? Odds are your sound card is output a digital signal to a digital receiver so it really doesn't matter.

 

[njekt]

@mlga91: Thanks for the warning. I have taken that into consideration, Peltier devices are a great way to accelerate heat flow from the CPU to the radiator. While the space around where the CPU will be in my case doesn't seem as if it would restrict my heatsink options, if I need to go smaller with a smaller heatsink, I'll incorporate Peltiers. That power supply project does spark my interest however, thanks.

@hunter315: This is the first time I've EVER had to say this to a moderator: Please obey the rules on the forum when posting threads or responding to others posts. Please refer to [urlExt=http://www.tomshardware.com/forum/id-2083609/read-forum-rules-styling-posts.html]Read First: Forum Rules & Styling Posts[/urlExt] for the list of forum rules that members are required to obey. I'd like to bring your attention to a few specific rules from that list:

- Under the section labelled "Please do...", the 4th bullet: "Keep criticism constructive! If you disagree with an opinion, explain why, but never attack the person."
- Under the section labelled "Please don't...", the 7th bullet: "Be rude or impolite. Civility is essential on Tom's, and remember that behind each user is a real person. Personal attacks (ad hominem) and insults are not allowed."
- And lastly, under the section labelled "Specific Rules", the 1st bullet: "Harassment, threatening, embarrassing or insulting other users, including sending unwanted messages, attacking race, religion, gender, sexual orientation, etc."

Until such time you come to realize that these rules were put in place for everyone's benefit, I would ask you to not interrupt any more threads or any threads farther.


Even with all that being said, if you do truly respect the words in your signature, please allow me to refer you to the follow links and explain the significance of them: (these are for the benefit of all readers as they will show hunter315 is completely wrong in every one of his claims. anyone who happens upon this thread looking for information relating to this topic won't be misled by the information hunter315 stated.)
- [urlExt=http://www.hardocp.com/article/2013/07/16/seasonic_x400fl_platinum_400w_power_supply_review/8#.VDeae-IbC9c]HARDOCP - DC Output Quality - Seasonic X-400FL Platinum 400W Power Supply Review[/urlExt]. Click any of the images in the link and notice two things: 1) the text to the bottom right of each graph, the timescale is described: 2ms/Div, and 2) that the period of the ripple is 4 divisions wide and that there seems to be about 100 minor ripples on top of the major ripple. The significance: Since the frequency of a wave is equal to 1 divided by the period of that wave, and since we discovered the period of the wave is 8ms (4Div*2ms/Div=8ms), then the frequency of the major ripple is f=1/0.008=125hz - an audible frequency. Similarly, to find the significance of the ~100 minor ripples inside the major ripple we will also need to calculate their frequency. If the period of the major ripple is 100x the minor ripples then the period of the minor ripples is 8ms/100=80us and therefore the frequency of the minor ripples is f=1/0.00008=12500hz - a very, very audible frequency!
- As noted in my Topic post, the sound card I'm using is an ASUS Essence STX II 7.1. From the [urlExt=http://www.asus.com/Essence_HiFi_Audio/Essence_STX_II_71/]Official Site[/urlExt], it is very clearly noted and displayed that this card outputs 8 preout signals (7.1) through RCA ports. They are generated by quite an amazing DAC. My second post describes the incredible THD+N of the DAC and my priority of preserving its potential output on this card. The significance of this is that my mention of a DAC and concern of preserving its potential tells you that not only does it output analog signals, but I'm also incredibly concerned about preserving the analog signals.
- And lastly, a site detailing [urlExt=http://www.sengpielaudio.com/calculator-thd.htm]THD to dB - convert percent % to decibels dB[/urlExt]. In post two I stated that the soundcard would obviously have countermeasures against the noise/ripple being fed to it by the PSU, but I doubted it would be smoothed down to an acceptable degree. There is a calculator at the bottom of the link that will demonstrate the significance. Adding together the THD+N decibal ratings of the PSU and Soundcard found in my second post, you'll see that the sum of their noise equals that of the PSU. In other words, the quality of the final signal is mostly a function of the quality of the most noisy signal. If you plugged in -120 db as a dummy number replacing the PSUs value just to see how dependant the final signal is on the most noisy signal you get a value of -118.545 db. Anyway as I've already expressed, I did recognize this as the worse case scenario. I would also like to note that when the input voltage gets "stepped down", the the ratio of ripple to desired supply will remains the same as it was before being stepped down. There are other countermeasures incorporated into this sound card to reduce unwanted components in the power supply.

Also, you claimed I was to use "all "passively" cooled components", as my first post clearly states I said "passive radiators". That's a big difference.

 

[Someone Somewhere]

Text from report abuse (as written):

Up until this moderators post, this thread was proceeding in well mannered and in directions relevant to the topic post.

hunter315 then brought nothing but misinformation and rudeness. Disciple him.

I, quite frankly, see nothing major wrong on his part. Maybe a little misinformation, but that's marked as a guess.

- Under the section labelled "Please do...", the 4th bullet: "Keep criticism constructive! If you disagree with an opinion, explain why, but never attack the person."

Where didn't he explain his criticism?

Under the section labelled "Please don't...", the 7th bullet: "Be rude or impolite. Civility is essential on Tom's, and remember that behind each user is a real person. Personal attacks (ad hominem) and insults are not allowed."

I don't see any incivility, with the possible exception of "shows you have clearly lost all actual sense when it comes to this issue". Which is very mild as insults go, and seems well-founded.

- And lastly, under the section labelled "Specific Rules", the 1st bullet: "Harassment, threatening, embarrassing or insulting other users, including sending unwanted messages, attacking race, religion, gender, sexual orientation, etc."

Only possible insult noted above.

Also, relevant to the rest of the thread:

a PSU with a ~4mv ripple on the 12V rail should be obtainable

Um. You're dreaming. I hooked up a dead 12V battery (all I had lying around) to an oscilloscope and even that had about 10mV peak to peak. It's simply not doable from mains.

I would also like to note that when the input voltage gets "stepped down", the the ratio of ripple to desired supply will remains the same as it was before being stepped down.

If you're dealing with a straight transformer, no filtering, and AC. None of which are accurate here. Buck converters will be used, which typically drop way more noise than that in. Expect lots of filtering after them.

Click any of the images in the link and notice two things: 1) the text to the bottom right of each graph, the timescale is described: 2ms/Div, and 2) that the period of the ripple is 4 divisions wide and that there seems to be about 100 minor ripples on top of the major ripple. The significance: Since the frequency of a wave is equal to 1 divided by the period of that wave, and since we discovered the period of the wave is 8ms (4Div*2ms/Div=8ms), then the frequency of the major ripple is f=1/0.008=125hz - an audible frequency. Similarly, to find the significance of the ~100 minor ripples inside the major ripple we will also need to calculate their frequency. If the period of the major ripple is 100x the minor ripples then the period of the minor ripples is 8ms/100=80us and therefore the frequency of the minor ripples is f=1/0.00008=12500hz - a very, very audible frequency!

Using 25 kilosamples per second prevents you picking up any noise above 12.5kHz. 12.5kHz is also very easy to filter out from DC. I'd worry far more about the 125Hz.

 

[Lutfij]

Until such time you come to realize that these rules were put in place for everyone's benefit, I would ask you to not interrupt any more threads or any threads farther.

I would appreciate that those with experience on relevant objectives comment with whatever knowledge they can offer, hopefully maybe even confirm such an objective is possible and recommend a CPU they've had success with.

Hunter was offering assistance into the matter and you'd also need to realize who would benefit from this thread the most with said responses.

Furthermore, FYI:
· onboard audio and network will be disabled (not sure if these affect processor demands)

It will only reduce the overall systems power and depending on your mobo and the network card bundled, the Intel NIC uses very little CPU overhead and apparently is server grade tech watered down for mass use.

I'd like to do so without using any fans whatsoever - passive radiators only!

In order to do that you'd need to go for a very large radiator space/profile - probably a 1080 or less however we don't know your hardware specs yet and the intended TDP of your system. In which case you'd find all relevant information in the watercooling sticky(located in my sig) Even if you did tone down all the noise in your system, you won't be able to eliminate the coil whine coming from your PSU be it a seasonic fanless unit or even their hybrid units furthermore there is the matter of the pumps noise/speed/flow and overall heat dissipation of the loop.

Hope this helps.

 

[njekt]

Where didn't he explain his criticism?
-"Also, what makes you think that "minimizing all local electric, magnetic and electromagnetic fields as much as possible (read fans, among other things)" will actually have an impact on anything?"
-"30 mV is actually damn good for a PSU, top end ones can stay below 20 mV but those are the really big ones with lots of filter capacitors." Where is his proof on this? My link below shows otherwise. Also, where did I mention that I'm partial against using large capacity PSUs?

I don't see any incivility, with the possible exception of "shows you have clearly lost all actual sense when it comes to this issue". Which is very mild as insults go, and seems well-founded.
-If his insult is so well founded, why do you agree he was spreading misinformation? also, if his insult is so well founded, I wonder why he isn't so eager to reference other material or show some math? I'm bringing tonnes of reference and math to the table.

Only possible insult noted above.
- Wrong. He asked a rhetorical question meant to tell me that my concerns couldn't possibly make a difference.


Also, relevant to the rest of the thread:
- No it isn't. This is a thread about CPUs, that is why I posted it in the CPU section. If you'd like to argue about PSU, theres a section for that. If you'd like to talk about hunter315s behavior, theres other channels for that too, like the private channel I opened up between us! Nothing about any post later than mlga91's is relevant to the threads topic. This thread has been completely hijacked.... ironically, been done so by moderators at that! Moderators are meant to curb poor behavior.

a PSU with a ~4mv ripple on the 12V rail should be obtainable

- =173885&specId=8028&tcId=200]http://uk.hardware.info/productinfo/benchmarks/21/power-supplies?products[]=173885&specId=8028&tcId=200. As is clearly shown, ripple approaching 4mv is out there and easy to find. Its not quite 4, but its a lot better than 30mv.

If you're dealing with a straight transformer, no filtering, and AC. None of which are accurate here. Buck converters will be used, which typically drop way more noise than that in. Expect lots of filtering after them.
- Prove this statement. I support all of my statements with math or references.

Using 25 kilosamples per second prevents you picking up any noise above 12.5kHz. 12.5kHz is also very easy to filter out from DC. I'd worry far more about the 125Hz.
- Wrong, again.If there was a 12.501kHz ripple that at t=0 had an increasing amplitude currently equal to 0, then sampling at 25kHz your second sample (at t=1/25000hz) would have an amplitude slightly below 0 and would be decreasing. The noise is still noticed by your sampler, its just the wave will no longer appear to be at the same frequency nor beginning at the same phase. In this particular case the frequency will appear very small, will appear to begin with a phase of 180 degrees and the amplitude will be very small and negative. a(t)=a(0)*sin(2*pi*t/T)=a(0)*sin(2*pi*(sample_num(t)/25000hz)*(1/(1/12501hz))). It is not like noise all of a sudden becomes undetectable if higher than 12500Hz when sampling at 25000Hz.

 

[njekt]

Hunter was offering assistance into the matter and you'd also need to realize who would benefit from this thread the most with said responses.

Hunter offered an answer to a question no one asked and offered incorrect information with absolutely no reference or backing to his claims. He just said (paraphrasing) '30mV is good. 20mV is very good. You have lost your mind.' Didn't show me any proof. In my previous post responding to someones something there is a link showing that ~4mV is close to whats out there.

Furthermore, FYI:
· onboard audio and network will be disabled (not sure if these affect processor demands)

It will only reduce the overall systems power and depending on your mobo and the network card bundled, the Intel NIC uses very little CPU overhead and apparently is server grade tech watered down for mass use.

Now this is helpful information. Thank you for your contribution.

I'd like to do so without using any fans whatsoever - passive radiators only!

In order to do that you'd need to go for a very large radiator space/profile - probably a 1080 or less however we don't know your hardware specs yet and the intended TDP of your system. In which case you'd find all relevant information in the watercooling sticky(located in my sig) Even if you did tone down all the noise in your system, you won't be able to eliminate the coil whine coming from your PSU be it a seasonic fanless unit or even their hybrid units furthermore there is the matter of the pumps noise/speed/flow and overall heat dissipation of the loop.

This thread isn't about the PSU, its about the CPU. But I do have the room in the case to do whatever I want.
I did mentioned the only components in my system other than CPU this topic was meant to determine and the MoBo I'll fit to it is the ASUS Essence STX II 7.1 + Daughterboard, Bluray Writer, SSD. I did not mention the memory, but obviously that is hardly an issue with power or fields and needs to be fit to the MoBo anyway.
CPU cooling will be a few Peltiers in thermal series between the CPU and Heatsink if the heatsink can't handle the thermal load itself. By thermal series i mean the following:
[CPU] [Cold Side Hot Side] [Cold Hot] [Cold Hot] [Heatsink]
When Peltiers are placed in thermal series like this their total deltaT from end to end is far greater than than that of a single Peltier even if the Peltier series is only consuming the same amount of power than the single Peltier. The overall effect of this is to accelerate the transition of heat from the CPU to the heat sink.
I do realize that in the end I'll still have the fields coming off the PSU, theres almost nothing I can do about that. But that doesn't mean I should completely ignore all the other sources of noise.

 

[Lutfij]

1| You may need to step back and re-read all the info given henceforth. You're holding the cat by its tail and not by the neck frankly speaking.
2| You are interested in adding more peltier than is needed. The relevant information is found in the watercooling sticky and is also found in the below ambient cooling thread located at the top of the overclocking section.
3| This may be your thread but it is highly encouraged to keep thread on topic and to keep discussion to it as close as possible however in the event that the thread does take a different manner/form the thread section can easily be changed courtesy of Mods.
4| We asked for a total systems specs to guage which would be the nosiest component in your system.

You'd find it in your best interest to start a new thread on the overclocking section with the intended number of peltiers for use and hopefully understand that you're overthinking this through.

Anyways, I'm afraid I've hi-jacked your thread and will refrain from posting further.

My 2 cents though.

 

[njekt]

1| You may need to step back and re-read all the info given henceforth. You're holding the cat by its tail and not by the neck frankly speaking.
2| You are interested in adding more peltier than is needed. The relevant information is found in the watercooling sticky and is also found in the below ambient cooling thread located at the top of the overclocking section.
3| This may be your thread but it is highly encouraged to keep thread on topic and to keep discussion to it as close as possible however in the event that the thread does take a different manner/form the thread section can easily be changed courtesy of Mods.
4| We asked for a total systems specs to guage which would be the nosiest component in your system.

You'd find it in your best interest to start a new thread on the overclocking section with the intended number of peltiers for use and hopefully understand that you're overthinking this through.

Anyways, I'm afraid I've hi-jacked your thread and will refrain from posting further.

My 2 cents though.

1) Trust me I have no vested interest in this site anymore. And will have no vested interest in a site that makes arguments without proof or reference or measurement. In the real world, people who try to spout fact/arguments without aforementioned proof or reference or measurement are referred to as a word probably banned on this forum (hint: begins with the word for a male cow and ends with 'ers', in the middle is what female cows do a lot and the traits of being such can be described by the words arrogant and ignorant).
2) "a few Peltiers" is exactly what I said. Notice the indirect reference to exactly how many I will implement. I will implement whats necessary. Furthermore, different peltiers have different power requirements per different deltaT per surface area. I didn't mention any of the aforementioned, it is impossible to state that (undefined number) amount of peltiers would be too many as I didn't mention any of the aforementioned attributes of them either
3) I'm not the one who kept insisting on taking the thread off topic. That was done by the 3 moderators who responded to this thread (you included). And not done at all by the 2 people who weren't moderators that responded to this thread. This wasn't about overclocking, but rather CPU capability. Overclocking increases heat at a rate greater than processing capability/second. Hence why this thread was posted in the cpu portion of this forum.
4) And I provided it, twice, as much as I could. I can only provide as much information in my build that can be decided on considering the issue at hand. I can't pick a MoBo (socket type) or a set of Ram (standard or SO) without knowing the CPU.

lutfji: Thank you for agreeing to not screw up this thread any farther, and the admittance of guilt that you participated in screwing up this thread. However, I don't see any statement of remorse. This is kind of like saying you'd have no problem disrupting this forum again despite that you know that you have already (at least) once.

My 2 cents: Each of the 3 moderators that responded to this thread have failed to bring any evidence whatsoever (absolute, circumstantial or heretical) to reinforce their claims. I issued a challenge to each 3 to bring forth evidence in support of their claims, all have failed. Two of those moderators failed to respond to my challenge in any degree, one of those moderators is now trying to deflect that challenge. I, on the other hand, have supplied support through either reference or (interpolated or not) standard scientific principles. None of the opposition have brought forth an equal attitude of debate, just insult.

This thread can still be an informational resource of the knowledge I presented within. I beg those seeking this knowledge to read the whole thread and to notice who supplys resources or math to back up their claims before choosing which side of the argument to believe in.

Lastly, since it seems that trying to make any effort on this board, either inquisitional or informational, is met with hostility, I will not respond to any posts beyond this one. Do not take this as a forfeit but rather it not being worth it for me. All information I presented up to this point is correct as seem in the support I provided with it.

Toms used to be THE HW enthusiast site. WTF happened?

 

[Flaffz]

Toms used to be THE HW enthusiast site. WTF happened?

A douchebag came in acting all high and mighty, reacting on criticism as if he was facing death penalty in a courtroom.