Troublesome mini-ITX

Show off your quiet rig.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Post Reply
nagi
Posts: 125
Joined: Tue Aug 31, 2010 6:48 pm
Location: Outside the box

Troublesome mini-ITX

Post by nagi » Sat Jan 02, 2016 9:03 am

Hi guys,

As described in this topic, I've begun to build a mini-ITX machine for work (coding) purposes. The components were selected so that when this gig is over, I'll be able to just put it in my much older storage server.

Image

The components are the following:
Case: RAIJINTEK Metis Window, silver
Motherboard: ASROCK Z97E-ITX/ac
CPU: Intel i7-4790K (@stock)
CPU cooler: Scythe Fuma
Memory: HyperX 16GB FURY DDR3 1866MHz CL10 KIT HX318C10FBK2/16
SSD: Samsung 850 EVO 250GB
PSU: Seasonic X-400

Idle temps were in the 35°C (delta+10-12) region, and noise levels were acceptable (apart from problem #1). However, I had to completely disassemble the case to get the build in, and I also excerted a considerable amount of pressure from the bottom of the case to the CPU cooler heatsink (the mobo is oriented upside-down), which is what I think caused Prime 95 temperatures to jump to 100°C. The positioning also makes the bottom 2.5" drive bays unusable.

The problems & observations:
  1. 1. Both fans (SY1225SL12M-CJP) with the Fuma exhibit a very annoying clickiness on low RPM. AFAIK they are simply recolored Scythe Slip Streams, but I've never had problems like this with their older revisions (I have 2 Mugen 2 rev.B in other builds).
  • 2. It seems the motherboard will only control PWM fans on its CPU fan header. Or connecting through the scythe-provided PWM splitter to it confuses it to believe the fan connected needs to be controlled with PWM and eschews DC control. The mobo does have one other fan control which works OK with my 3-pin Nexus.
  • 3. Unfortunately I only had one Nexus fan in stock, and the local supplier decided to call it quits. I might be able to buy some Noiseblocker Multiframe S-Series M12-S2 120mm fans, if I'm lucky. Getting some NOCTUA NF-S12B redux-700 fans should be easy, but those are 3-pin only, and not regarded as well as the Noiseblockers in the SPCR reviews.
  • 4. As said, the CPU socket is too high, so I had to rotate the fuma by 90°, which means that instead of a fan-duct with great air-flow (in through the PSU, through the CPU cooler and out the back) I have a hackjob. And the Noctua fan does not like sitting flat, and is a bit noisier in this orientation.
  • 5. If put in 90°, the fan clips of the Fuma will interfere with the back case fan, so I had to remove it.
  • 6. While the 90°C orientation did drop the Prime 95 temperatures from 100°C spikes down to 70 max, (both at 100% fan speeds, most definitely not silent or even quiet) the heatpipes put some mechanical pressure on the memory. So far so good though.
  • 7. The X-400 exhibits a white-noise buzzing in low load (idle). When the CPU ramps up with Prime 95, this stops completely as far as I can tell. (Since the fans ramp up at that point, I don't have much time to observe the quiality of noise.)
  • 8. To whomever decided to put the BIOS reset button next to the HDMI output on the motherboard: I will find you.
The plan for now:
Exchanging the mobo for another is a no-go. I went through our local supply, and this is the one board that has the features I'll need either at start or down the line. (6 SATA ports on the mobo, not Gigabyte, integrated wifi + bluetooth, DP, HDMI and DVI out with at least some being 4K capable, mini-ITX size)

Get either a NOCTUA NH-U9B SE2 or a NH-D9L. These should be able to fit in and provide at least adequate cooling. I was thinking about a rotated Mugen 4, buuuut first, that would leave only one fan in the entire build, and second, Noctua fans are much less of a gamble, and sound better in my experience.

I'm not sure which Noctua would be better though. SPCR reviewed the D9L, but not the U9B SE2 (only the U9B that had one heatpipe more but was bundled with only one fan.) The SE2 has two fans, but I'm not sure where I could put the second one: I have cabling between the heatsink and the PSU, and on the other side, there is already the 12cm fan on the case back, which could cause some nasty interference at least, and/or subpar cooling performance. But the D9L is 50% more expensive, and the SE2 is already more expensive than a Mugen 4. I'm thinking about going with the SE2 and adding some cardboard air ducts. And some LED lighting so all can appreciate the horrible end result.

short Scythe Fuma review:
Split tower design. Cools truly great on open air. Should be also great in a regular ATX case. You get a PWM splitter and 2 fans. 150mm tall, so should fit most cases. Unfortunately the fans are noticeably clicky on low rpm (~<750, higher the wind noise drowns it out), and getting the cooler in 90° can interfere a bit with the memory module in the first slot, unless you have reduced height ones.

Pictures time! (click for bigger)

Looking nice & big
Image

The 12V cable is a "bit" stressed.
Image

You can see the fins pressing against the rubber mounting gromets at the bottom of the case
Image

The fan-"duct" cooling path I first tried. It could work if the CPU socket was about 15-20mm more distant from the top edge of the motherboard. (also, Hi!)
Image

Another shot, with all but side panels fitted.
Image

My cabling in this case follows the school of the tentacle monster.
Image

lodestar
Posts: 1683
Joined: Fri Aug 05, 2005 3:29 am
Location: UK

Re: Troublesome mini-ITX

Post by lodestar » Sat Jan 02, 2016 10:41 am

If you can source them locally it might be worth changing the existing Scythe fans for a pair of Noctua NF-S12B Redux 1200 PWM fans. Using the existing PWM Y cable these will run from the PWM CPU header in the 400 to 1200 rpm range without clicking. It could give you the lowest idle speeds without compromising on the maximum fan speeds available when the system is under stress and/or higher ambient temperatures. I don't know what your BIOS options are for the CPU fan but selecting a Quiet profile if there is one might assist in reducing idle speeds to a minimum. As it's an ASROCK board the supplementary Fan-Tastic software might help. I think this is one of those boards where the 4 pin chassis fan header is in fact just a normal 3 pin header with a non-functioning extra pin.

ggumdol
Posts: 154
Joined: Mon Aug 05, 2013 1:01 pm
Location: Scandinavia

Re: Troublesome mini-ITX

Post by ggumdol » Sat Jan 02, 2016 12:37 pm

Random thoughts:

1. Don't ever buy fans smaller than 120mm. The same applies for CPU coolers. They produce unbearably high frequency noise. You can understand this fact easily if you studied physics properly.

2. I tinkered with Asrock Haswell & Ivy Bridge (Sandy Bridge) motherboards for a long time though I'm currently using Asus H97M-E for my main rig. As far as my experience goes, all Asrock motherboards have true PWM headers if they are designated so in their respective manuals. They have absolutely the best BIOS fan control functionality, incomparable to any other motherboards, perhaps even now. To be fair, I had some issues related to USB front header with Asrock motherboards. Also, I prefer the Crystal Sound of ASUS to the Asrock equivalents. I couldn't understand your point exactly but I have had no issue whatsoever in controlling PWM fans with PWM headers in Asrock Haswell motherboards. All Noctua fans were controlled with finest ever granularity, not to mention full-range PWM duty cycles from 0% to 100%.

3. I would't have bought such a mighty i7 Haswell CPU for such a compact ITX build. I reckon you will resolve the best part of those issues simply by replacing your i7 with a decent i5. I haven't ever felt my i5-4430 is sluggish.

4. When it comes to measuring noise levels of fans, it seems necessary to rethink about how it is measured. Unlike how the noise is measured, I think most people are far more sensitive to high-frequency part of fan noise rather than to its low-frequencey part. In this regard, I think Noctua fans are usually superior to other fans due to its reticent noise profile (i.e., smaller high-frequency component).

nagi
Posts: 125
Joined: Tue Aug 31, 2010 6:48 pm
Location: Outside the box

Re: Troublesome mini-ITX

Post by nagi » Sat Jan 02, 2016 3:05 pm

@lodestar: Hmm, I might, thanks for the tip, I thought all Noctua NF-S12B's were only 3-pin.

Yes, AsRock BIOS is capable of a quiet mode, although right now I'm using speedfan, as it allows more data points to be set and use ,and a few other extras. I did not test the chassis fan header with a PWM fan, so yeah, it might be a 3-pin in disguise, although this is AsRock, not Gigabyte. :D

@ggumdol: Thanks for you thoughts.

1. I agree that generally 120mm is the way to go, and 99% sub-120mm fans are rubbish and noisy, but if I can't fit a good tower that uses a 120mm fan in a decent airflow configuration, there isn't really any point of recommending to only buy 120mm. With that said, I did meet a very few (2) 80mm fans that had very good noise profile when run on low rpm. And let's not forget that my 250W dissipating 290X video card is cooled by 3 92mm fans, which create an acceptable level of noise during gaming.

2. The issue was controlling a 3-pin DC fan from the 4-pin PWM CPU fan header. It might have been some quirk in the system. I know they are good, as I've had other AsRock boards before this one.

3. Well, to each their own, I wanted this particular model because it does make a difference in some of my work.

4. I'm measuring with my ears, as the configuration will be for my own use.

Frankly, what I want is silence in idle, and something less than a jet engine on full tilt. As my desktop is an i7-4770k, I know cooling a haswell is quite hard, and not exactly possible fully silent on 100%. As most of the time the CPU won't be running full tilt for enough time to engage the jet engine, I would accept this.
Last edited by nagi on Sat Jan 02, 2016 3:13 pm, edited 1 time in total.

edh
Posts: 1621
Joined: Mon Mar 17, 2008 1:49 pm
Location: UK

Re: Troublesome mini-ITX

Post by edh » Sat Jan 02, 2016 3:08 pm

ggumdol wrote:Don't ever buy fans smaller than 120mm. The same applies for CPU coolers.
OK, a strong view there. Bear in mind that if you can't fit a larger size then you don't have a choice. In this instance there are other options involving big tower coolers for 120mm fans but be careful about blanket statements. All graphics cards with fans have smaller fans than 120mm on them, unless you have a big aftermarket solution. So your 'don't ever buy' isn't entirely the truth.
ggumdol wrote:I would't have bought such a mighty i7 Haswell CPU for such a compact ITX build.
Not necessarily. A stock 4790k has a TDP of 88W, in a system with no graphics card this shouldn't be so much of a problem. I have an i5-6400 with a GTX950 in the same case and CPU temps are less than 50C under stress testing.

Having looked at the pictures in more detail I can't help but feeling that a Ninja 4 would work OK. The width is 7mm less so the clips would not be a problem and fans could be arranged in any direction. It will be a little further south than in the system I have but it would be better than what you have now. You could also mount a second fan on the top so that the airflow is diagonal (in at the bottom and front, out at the top and back) but this may be noisy.

You also could rearrange the cables around the PSU to gain a bit better ventilation.

One other difference you have is the windowed sidepanel. This of course means no side ventilation. You could swap left and right panels over to remedy this.

ggumdol
Posts: 154
Joined: Mon Aug 05, 2013 1:01 pm
Location: Scandinavia

Re: Troublesome mini-ITX

Post by ggumdol » Sat Jan 02, 2016 3:27 pm

edh wrote:
ggumdol wrote:Don't ever buy fans smaller than 120mm. The same applies for CPU coolers.
OK, a strong view there. Bear in mind that if you can't fit a larger size then you don't have a choice. In this instance there are other options involving big tower coolers for 120mm fans but be careful about blanket statements. All graphics cards with fans have smaller fans than 120mm on them, unless you have a big aftermarket solution. So your 'don't ever buy' isn't entirely the truth.
I will admit it was a bit sweeping statement. Each person seems to have a different acoustic acuteness. I can't bear a single sub-120mm fan rotating except when gaming. They are simply inefficient in terms of fluid mechanics. However, I don't have this insistence for HTPCs, which is why I don't (maybe can't) turn off the sub-120mm fans (Arctic) on my GTX750Ti in my HTPC during idle state.
edh wrote:
ggumdol wrote:I would't have bought such a mighty i7 Haswell CPU for such a compact ITX build.
Not necessarily. A stock 4790k has a TDP of 88W, in a system with no graphics card this shouldn't be so much of a problem. I have an i5-6400 with a GTX950 in the same case and CPU temps are less than 50C under stress testing.
You are here comparing i5-6400 with i7-4790k. (i) Haswell vs. Skylake; (ii) i5 vs. i7. Did you really think i5-6400 is comparable to i7-4790k in thermal aspects? As far as my experience goes, the official TDP figures are more like upper-bounds of thermal dissipation rather than factual values. I think most i5 CPUs from Haswell or later are completely fit for (compact) ITX builds such as in Metis chassis. When you move on to i7, it's an entirely different story.

edh
Posts: 1621
Joined: Mon Mar 17, 2008 1:49 pm
Location: UK

Re: Troublesome mini-ITX

Post by edh » Sun Jan 03, 2016 5:28 am

ggumdol wrote:You are here comparing i5-6400 with i7-4790k. (i) Haswell vs. Skylake; (ii) i5 vs. i7. Did you really think i5-6400 is comparable to i7-4790k in thermal aspects?
Did you read what I wrote? I have a GTX950 in there too. That has a 95W TDP. The OP has no graphics card in the system. Therefore I have a higher total TDP in the system. Saying that you just can't make an 88W CPU work in a small case is not true, you can do but it is harder.

Right now a 70C load temp is not ideal but it is not ridiculously hot. Clearly something was wrong with the other CPU orientation that caused the 100C temp. I would expect that with work this could be brought under 60C.

Things that could be done:
- Swap the CPU cooler for one that fits better, eg. Ninja 4 (I wouldn't do this on a whim, try the other things first and do some maths)
- Modify the rear panel so that the rear fan is mounted lower down, bringing it in line with the centre of the CPU socket so that it correctly cools the CPU cooler
- Change the airflow round so that the air is brought in from the underside of the case (cut a hole) and either out the back or further up the case
- Move some of the cables out of the way so that better ventilation comes through the PSU
- Swap the side panels round so that the ventilated panel is on the CPU side

xan_user
*Lifetime Patron*
Posts: 2269
Joined: Sun May 21, 2006 9:09 am
Location: Northern California.

Re: Troublesome mini-ITX

Post by xan_user » Sun Jan 03, 2016 7:54 am

under load you are using mostly hot air from the psu to cool the CPU.

simple fix would be to reverse fan direction.

edh
Posts: 1621
Joined: Mon Mar 17, 2008 1:49 pm
Location: UK

Re: Troublesome mini-ITX

Post by edh » Sun Jan 03, 2016 9:04 am

xan_user wrote:under load you are using mostly hot air from the psu to cool the CPU.

simple fix would be to reverse fan direction.
Have you calculated the thermal rise through a PSU before? It's very small. Assuming a very low 30m3/hour airflow (a 120mm fan doing 300-400rpm lets say), that's 10g of air per second going through the PSU. The thermal capacity of air at room temperature is 1.005 J/kg/K. Let's say that with a 90% + efficiency PSU you have 10W of waste heat going into the air then what is the temperature rise on the air flowing through the PSU? 1C, that's all. Very easy to calculate. This will equate to less than 1C rise on the CPU. What people get confused about is that if the PSU is used as an exhaust from a system it pulls in hot air and incorrectly assume that the hot air coming out is entirely caused by the PSU whereas the PSU has only added 1C to it.

I run a similar setup and it works well. Reversing the fan direction would give positive pressure but the Metis is not really set up for it and laminar flow can not be achieved.

nagi
Posts: 125
Joined: Tue Aug 31, 2010 6:48 pm
Location: Outside the box

Re: Troublesome mini-ITX

Post by nagi » Sun Jan 03, 2016 9:24 am

edh wrote:Things that could be done:
- Swap the CPU cooler for one that fits better, eg. Ninja 4 (I wouldn't do this on a whim, try the other things first and do some maths)
- Modify the rear panel so that the rear fan is mounted lower down, bringing it in line with the centre of the CPU socket so that it correctly cools the CPU cooler
- Change the airflow round so that the air is brought in from the underside of the case (cut a hole) and either out the back or further up the case
- Move some of the cables out of the way so that better ventilation comes through the PSU
- Swap the side panels round so that the ventilated panel is on the CPU side
That sidepanel idea is really thinking outside the box, thanks. :) I actually wanted to buy a window-less version, but that was only available in gold.
Before I start cutting away at the case, I'll try with another cooler.

xan_user wrote:under load you are using mostly hot air from the psu to cool the CPU.

simple fix would be to reverse fan direction.
I have to agree with edh here.
The X-400 does not get hot - or even warm - under this kind of load. I have another one in my gaming desktop, which draws many times this, and frankly, on idle that PSU is cold (room temperature) to the touch. On prolonged 400W load, its casing gets maybe 5°C degrees warmer. The air through it even less. And that's with minimal ventillation (~400rpm on a 120mm fan).

The other issue is that while exhausting through the back fan is clean, (the fan exhaust cannot go elsewhere) exhausting through the PSU is a bit more problematic and turbulent, possibly even decreasing the cooling performance.

ggumdol
Posts: 154
Joined: Mon Aug 05, 2013 1:01 pm
Location: Scandinavia

Re: Troublesome mini-ITX

Post by ggumdol » Sun Jan 03, 2016 11:58 am

Well, I did plough through your reply a few times and it didn't come to my mind that GTX 950 might make such a big difference. I also fiddled with many GPUs and CPUs. My experience however dictates the contrary. The existence of GPUs makes only cosmetic differences because hot parts of those GPUs are not physically connected to CPUs (i.e., conspicuous absence of conductive heat transfer).

Having said that, I agree the likes of GTX 950 might make 1-2 Celsius difference in CPU temperature.
edh wrote:
ggumdol wrote:You are here comparing i5-6400 with i7-4790k. (i) Haswell vs. Skylake; (ii) i5 vs. i7. Did you really think i5-6400 is comparable to i7-4790k in thermal aspects?
Did you read what I wrote? I have a GTX950 in there too. That has a 95W TDP. The OP has no graphics card in the system. Therefore I have a higher total TDP in the system. Saying that you just can't make an 88W CPU work in a small case is not true, you can do but it is harder.
Although I would personally be wary of 70 something Celsius CPU temperature, I think it's probably completely safe temperature by any standard and I suppose 70 Celsius will be observed only during load states, which will happen on rare occassions. At the same time, I think, given your investment on CPU coolers and etcetra, you would rightfully expect more satisfying thermal performance.
edh wrote:Things that could be done:
- Swap the CPU cooler for one that fits better, eg. Ninja 4 (I wouldn't do this on a whim, try the other things first and do some maths)
- Modify the rear panel so that the rear fan is mounted lower down, bringing it in line with the centre of the CPU socket so that it correctly cools the CPU cooler
- Change the airflow round so that the air is brought in from the underside of the case (cut a hole) and either out the back or further up the case
- Move some of the cables out of the way so that better ventilation comes through the PSU
- Swap the side panels round so that the ventilated panel is on the CPU side
Since you are intent on using i7 Haswell for your work, I think the most influential change might come from a new aftermarket cooler outperforming your current one. Other issues sound merely cosmetic to me, partly because you don't have a discrete GPU fuming massive heat and the Seasonic PSU is incredibly efficient so that there will be no significant heat injected by the PSU into the chassis. My two suggestions are as follows:
1) Try to apply good thermal paste (not necessarily top-notch one) as evenly as possible. I remember there was a YouTube video on applying thermal paste evenly by using a rectangular glass.
2) Make sure the mounting system of the CPU cooler is as secure as possible. Subjectively speaking, I think Noctua mounting system is unrivalled and I like their secure mounting system a lot. Maybe you could consider a Noctua cooler, by doing so you can automatically get a Noctua PWM fan which rotates as low as 150 RPM (in Asrock Haswell motherboards) with infinitesimal increments to the maximum rotation speed.

edh
Posts: 1621
Joined: Mon Mar 17, 2008 1:49 pm
Location: UK

Re: Troublesome mini-ITX

Post by edh » Sun Jan 03, 2016 12:40 pm

ggumdol wrote:My experience however dictates the contrary. The existence of GPUs makes only cosmetic differences because hot parts of those GPUs are not physically connected to CPUs (i.e., conspicuous absence of conductive heat transfer).

Having said that, I agree the likes of GTX 950 might make 1-2 Celsius difference in CPU temperature.
You are very confident in your own views aren't you? :wink: In a small case like this you will get far more than 1-2C difference by adding a graphics card given proximity and also the blockin of top ventilation from the CPU.
ggumdol wrote:Since you are intent on using i7 Haswell for your work, I think the most influential change might come from a new aftermarket cooler outperforming your current one.
I've already suggested this to the OP. It might give more than a couple of degrees on it's own but it needs to be one that fits correctly. I'm already using a Ninja 4 in such a case but on a motherboard where it lines up almost perfectly with the rear exhaust:
viewtopic.php?f=14&t=68832
ggumdol wrote:Other issues sound merely cosmetic to me
No. The position of the rear fan versus the CPU cooler is enormously influential here. Don't just dismiss others suggestions and then quote something about physics or fluid mechanics. For the record I have a degree in physics so be careful what you pick to argue about. :lol:

ggumdol
Posts: 154
Joined: Mon Aug 05, 2013 1:01 pm
Location: Scandinavia

Re: Troublesome mini-ITX

Post by ggumdol » Sun Jan 03, 2016 12:59 pm

Maybe we can agree to disagree. We are not really helping the OP as there is no discrete graphics card there. We seem to have had enough about peripheral issues. Please don't let me know your degree or CV. I don't want to know.
edh wrote:
ggumdol wrote:My experience however dictates the contrary. The existence of GPUs makes only cosmetic differences because hot parts of those GPUs are not physically connected to CPUs (i.e., conspicuous absence of conductive heat transfer).

Having said that, I agree the likes of GTX 950 might make 1-2 Celsius difference in CPU temperature.
You are very confident in your own views aren't you? :wink: In a small case like this you will get far more than 1-2C difference by adding a graphics card given proximity and also the blockin of top ventilation from the CPU.
If I were the OP, I would simply disable "hyperthreading" in Asrock BIOS (I remember there is an entry in Asrock BIOS) and replace the non-PWM fan on the CPU cooler (Fuma?) with a proper PWM fan (as you can guess, I will go for a Noctua one). You can enable hyperthreading whenever you work. Sorry being not really helpful but I can't come up with a better idea for cooling Haswell i7 in the minimalistic chassis.

edh
Posts: 1621
Joined: Mon Mar 17, 2008 1:49 pm
Location: UK

Re: Troublesome mini-ITX

Post by edh » Sun Jan 03, 2016 1:44 pm

ggumdol wrote:Please don't let me know your degree or CV. I don't want to know.
Don't worry, I won't lose sleep.
ggumdol wrote:If I were the OP, I would simply disable "hyperthreading" in Asrock BIOS
...and in what way will this help cooling? There is some misinformation around about this, TDP is not influenced by hyperthreading. For software development hyperthreading can be beneficial so don't assume that the OP does not need this.
ggumdol wrote:replace the non-PWM fan on the CPU cooler (Fuma?) with a proper PWM fan
The Fuma has two PWM fans. OK, the Glidestream isn't the best PWM fan out and yes I have replaced mine after only a month and a half with a Noctua but let's at least get the facts right shall we? The OP already has PWM fans and cooling is not great for now so changing the fans themselves for quietness is not the top priority.
ggumdol wrote:Sorry being not really helpful but I can't come up with a better idea for cooling Haswell i7 in the minimalistic chassis.
Totally possible. Don't be so dismissive until you have tried it.

xan_user
*Lifetime Patron*
Posts: 2269
Joined: Sun May 21, 2006 9:09 am
Location: Northern California.

Re: Troublesome mini-ITX

Post by xan_user » Sun Jan 03, 2016 2:04 pm

edh wrote:Have you calculated the thermal rise through a PSU before?
I will admit, I have not. but with a lowly gtx750 and i3 running on a 500 watt sfx-l silverstone, the PSU (removed from case) is dumping out 28.5c in a 22.7 room.

ggumdol
Posts: 154
Joined: Mon Aug 05, 2013 1:01 pm
Location: Scandinavia

Re: Troublesome mini-ITX

Post by ggumdol » Sun Jan 03, 2016 2:08 pm

edh wrote:
ggumdol wrote:If I were the OP, I would simply disable "hyperthreading" in Asrock BIOS
...and in what way will this help cooling? There is some misinformation around about this, TDP is not influenced by hyperthreading. For software development hyperthreading can be beneficial so don't assume that the OP does not need this.
Wow, you are utterly misinformed here. Hyperthreading does affect power consumption. That's actually the very culprit why i7 is harder to cool than i5 and so many people here opt for i5. Have a look at:

http://www.tomshardware.com/reviews/hyp ... 584-9.html
edh wrote:
ggumdol wrote:replace the non-PWM fan on the CPU cooler (Fuma?) with a proper PWM fan
The Fuma has two PWM fans. OK, the Glidestream isn't the best PWM fan out and yes I have replaced mine after only a month and a half with a Noctua but let's at least get the facts right shall we? The OP already has PWM fans and cooling is not great for now so changing the fans themselves for quietness is not the top priority.
I hear Scythe Glidestream fans generate clicking noise in low rpms. As you say, if OP doesn't care about it, it's not the top priority. I didn't bother to handle all issues here because the OP seems to experience an inexplicable or strange problem regarding Scythe fan splitter. I think the OP would better pinpoint the cause of this issue before purchasing any fan.

edh
Posts: 1621
Joined: Mon Mar 17, 2008 1:49 pm
Location: UK

Re: Troublesome mini-ITX

Post by edh » Sun Jan 03, 2016 2:28 pm

xan_user wrote:I will admit, I have not. but with a lowly gtx750 and i3 running on a 500 watt sfx-l silverstone, the PSU (removed from case) is dumping out 28.5c in a 22.7 room.
So of that 5.8C difference I guess that 1C might be from the PSU and the other 4.8C might be from the air already pulled through the case? I can't tell without knowing the system but the power wasted by a modern PSU is very small.
ggumdol wrote:Wow, you are utterly misinformed here. Hyperthreading does affect power consumption. That's actually the very culprit why i7 is harder to cool than i5 and so many people here opt for i5. Have a look at:

http://www.tomshardware.com/reviews/hyp ... 584-9.html
No, that 16W maximum gap (less than 10%) could easily be explained by bottlenecks on other system components while hyperthreading is disabled. And if you were to look at energy consumption (no doubt you're smart enough to work out energy from power) then the time taken for each task as referenced on the different pages in that review (eg. http://www.tomshardware.com/reviews/hyp ... 584-8.html) more than counteracts the 16W. :P

Maybe look through what you cite for more than 5 seconds next time unless you want to be 'utterly misinformed'.

ggumdol
Posts: 154
Joined: Mon Aug 05, 2013 1:01 pm
Location: Scandinavia

Re: Troublesome mini-ITX

Post by ggumdol » Sun Jan 03, 2016 2:44 pm

It's an astounding revelation that we have been talking about "performance per watt" rather than "absolute power consumption". I give up here. You win!
edh wrote:
ggumdol wrote:Wow, you are utterly misinformed here. Hyperthreading does affect power consumption. That's actually the very culprit why i7 is harder to cool than i5 and so many people here opt for i5. Have a look at:

http://www.tomshardware.com/reviews/hyp ... 584-9.html
No, that 16W maximum gap (less than 10%) could easily be explained by bottlenecks on other system components while hyperthreading is disabled. And if you were to look at energy consumption (no doubt you're smart enough to work out energy from power) then the time taken for each task as referenced on the different pages in that review (eg. http://www.tomshardware.com/reviews/hyp ... 584-8.html) more than counteracts the 16W. :P

Maybe look through what you cite for more than 5 seconds next time unless you want to be 'utterly misinformed'.

xan_user
*Lifetime Patron*
Posts: 2269
Joined: Sun May 21, 2006 9:09 am
Location: Northern California.

Re: Troublesome mini-ITX

Post by xan_user » Sun Jan 03, 2016 3:02 pm

edh wrote: So of that 5.8C difference I guess that 1C might be from the PSU and the other 4.8C might be from the air already pulled through the case?
when i took the reading the PSU was sitting on my desk, outside of the case. it was pulling in only ambient room temp air. -measuring PSU heat only.


@ OP, how much lower are temps at load with case lid off?

nagi
Posts: 125
Joined: Tue Aug 31, 2010 6:48 pm
Location: Outside the box

Re: Troublesome mini-ITX

Post by nagi » Sun Jan 03, 2016 6:53 pm

Guys, please, less fighting!

1. I don't think replacing the Fuma fans would increase thermal performance in the load scenario. I did however just order a Noctua PWM fan, (Noctua NF-S12B redux-1200 PWM) mainly because I realized my own local stock of good 120mm fans is one Nexus. Maybe I'll use it on the back exhaust, or it will come handy when I'll actually use the Fuma.

2. Disabling HT would also decrease performance. I will not decrease performance until I exhausted my other options.

3. I've ordered the NOCTUA NH-U9B SE2, and I will experiment with high-tech airflow guides (cardboard).

4. Also, cable management, I'll try to remove as many cables as possible from between the PSU and the CPU cooler.
xan_user wrote:@ OP, how much lower are temps at load with case lid off?
Much. To reiterate, my fan curves reach 100% at 70°C, and is almost exponential in shape. Right now with the rotated cooler and shot airflow, this is reached. However, outside the case, the CPU maxes at around 65°C. While that might not seem much, fan speed is at 65-70%, which makes the world of difference.

ggumdol
Posts: 154
Joined: Mon Aug 05, 2013 1:01 pm
Location: Scandinavia

Re: Troublesome mini-ITX

Post by ggumdol » Sun Jan 03, 2016 10:21 pm

nagi wrote:3. I've ordered the NOCTUA NH-U9B SE2, and I will experiment with high-tech airflow guides (cardboard).
Hi nagi. I'm not as knowledgeable about aftermaket CPU coolers as other people in this forum but I think Scythe Fuma is a decent cooler apart from its clicking noise in low rpm states. Noctua NH-U9B SE2 is comparable to my Scythe Big Shuriken 2, which is far worse in terms of cooling performance than Scythe Fuma. You should go for coolers with sufficiently big heatsinks particularly for Haswell i7. To be honest, I think Scyhte Fuma is almost as good as any other aftermarket cooler and there is no apparent reason to change it unless you have spare money to shell out or you want to tinker with some Noctua fans.

The circumstances you are facing seem very complicated to me and it's hard to pinpoint the culprit without experimenting it by myself. My first hunch is that you are not properly capitalizing on the dual towers of Scythe Fuma because there is no room for two fans in the chassis. In this case, the first tower is unlikely to help cool the CPU properly (You need a fan blowing onto it). That is, you are practically idling the first tower of Scythe Fuma. I suggest you try some single tower coolers if you must replace Scythe Fuma. If you are intent on opting for Noctua coolers, try the likes of NH-U14S (I'm a bit unsure if NH-U12S has sufficient prowess for Haswell i7 but you can give it a try). Or you do some research to find the best single tower cooler.

As a side note, Noctua NF-F12 and NF-S12x are designed for different purposes. The former "substantially" excels the latter in cooling heatsinks. It's not ideal to use NF-S12x on heatsinks.

nagi
Posts: 125
Joined: Tue Aug 31, 2010 6:48 pm
Location: Outside the box

Re: Troublesome mini-ITX

Post by nagi » Mon Jan 04, 2016 1:19 am

ggumdol wrote:Hi nagi. I'm not as knowledgeable about aftermaket CPU coolers as other people in this forum but I think Scythe Fuma is a decent cooler apart from its clicking noise in low rpm states. Noctua NH-U9B SE2 is comparable to my Scythe Big Shuriken 2, which is far worse in terms of cooling performance than Scythe Fuma. You should go for coolers with sufficiently big heatsinks particularly for Haswell i7. To be honest, I think Scyhte Fuma is almost as good as any other aftermarket cooler and there is no apparent reason to change it unless you have spare money to shell out or you want to tinker with some Noctua fans.

The circumstances you are facing seem very complicated to me and it's hard to pinpoint the culprit without experimenting it by myself. My first hunch is that you are not properly capitalizing on the dual towers of Scythe Fuma because there is no room for two fans in the chassis. In this case, the first tower is unlikely to help cool the CPU properly (You need a fan blowing onto it). That is, you are practically idling the first tower of Scythe Fuma. I suggest you try some single tower coolers if you must replace Scythe Fuma. If you are intent on opting for Noctua coolers, try the likes of NH-U14S (I'm a bit unsure if NH-U12S has sufficient prowess for Haswell i7 but you can give it a try). Or you do some research to find the best single tower cooler.

As a side note, Noctua NF-F12 and NF-S12x are designed for different purposes. The former "substantially" excels the latter in cooling heatsinks. It's not ideal to use NF-S12x on heatsinks.
Hi ggumdol.
I have no reason to believe the Fuma would perform on the level of other big towers by Scythe, like the Ninja or Mugen, even if used with only one fan on it. At least, in open air, without the case, it performed in that manner.

The problem really is the lack of space. The Fuma is too wide at 137mm. I might have ~125mm (if I give up the bottom 2.5" SSD slots) by my measurements, maybe 130mm tops, which effectively disqualifies coolers with 120mm fans. (I did say it required some force to close the bottom panel.) Unless I rotate them, which in turn compromises airflow. Unless I heavily mod the case, e.g. add a top fan grill. I'm not that confident in my metalworking skills.

The same is the problem with the 150mm wide NH-U14S. Plus, I did not mention this so far, but the case has an official top clearance of 160mm, whereas the U14S is 165mm tall. The NH-U12S might fit, with 125 width and 158mm height, but it would be a very, very tight fit.

Thanks for you ntoe on the Noctua fans.

ggumdol
Posts: 154
Joined: Mon Aug 05, 2013 1:01 pm
Location: Scandinavia

Re: Troublesome mini-ITX

Post by ggumdol » Mon Jan 04, 2016 2:43 am

Since I have NH-U12S on my main rig, I just did some tests with Prime 95. I chose the "maximum heat" test to maximize CPU temperature and ran the tests for 5 minutes (it looks quite stable after 5 minutes). My chassis is very poorly ventilated because the front and rear fans are rotatating at 330-350 rpm all the time. The ambient temperature is 23 Celsius.

CPU: i5-4430 (3.0 GHz under full load)
Max core temperature in idle state: 33 Celsius
Rotation speed of the CPU fan (NF-F12) in idle state: 430 rpm
Max core temperature in full load state: 58-59 Celsius
Rotation speed of the CPU fan (NF-F12) in full load state: 1110-1120 rpm

You can try to extrapolate this result. Temperature rise is approximately proportional to (linear with) CPU frequency. Add some more for hyperthreading.

nagi
Posts: 125
Joined: Tue Aug 31, 2010 6:48 pm
Location: Outside the box

Re: Troublesome mini-ITX

Post by nagi » Thu Jan 07, 2016 10:05 pm

tl;dr: partway there. Thermal runaway solved, hunting down coil whine.

...I think this is the most time I spent with building and troubleshooting any system, and I'm not new to it by any stretch.

A bit of an update:

0. Core temp vs CPU (aka CPUTIN) temp: after much googling, it seems they are wildly different. The first is a maximum value calculated by measuring input current of a core, the second is a measured value of the cpu socket. The maximum value of 72°C set by Intel (Tcasemax) refers to the CPUTIN temp. My initial problem was that speedfan does not see the CPUTIN, that it seems running CPUID HW Monitor will get it stuck until reboot, and that Core temperatures seem to be higher (5-15°C, a bit hard to measure with the HW Monitor glitch) than CPUTIN. At least the BIOS Silent mode won't raise fans to the max till the core temps reach ~85°C, and thermal throttling does not set in even even at 80-85°C core temps.

1. Found out what caused my thermal runaways of 90+°C: The AsRock board seems to employ a bit of a turbo hack. After about 30-60-120 seconds (seems to be varying) of prime95, frequency will jump from the official multicore turbo frequency of 4200Mhz to the official single-core only turbo of 4400Mhz. (Officially, AFAIK this could only be employed if one core is stressed, and the rest are idling.) This also pushes the HD Monitor reported power consumption from the ~80W range to 90-100W, also increasing the temperatures by ~5-20°C. Using this thread I set my long/short power limits to 85W, and expermientally found the current limit of 104 to work well.

2. Got the Noctua NH-U9B SE2 (single tower with dual 92mm fans) heatsink, put it to tests just on my table, without the case a couple days ago, no problems. I'm impressed with the package and what went into it. I might even ditch Scythe as my main supplier for CPU heatsinks. What I did not check was that the fans are 3-pin ones. Whoopsie.

3. Previously I said the NH-U12S might fit. I'm revising this: if the fan clips are anthing like that of the NH-U9B SE2, I doubt it would. It might fit without the clips, but as stated before, I'll be transporting this more frequently than a desktop case, so that is not an option.

4. It seems the "Silent Mode" fan setting in the bios gets the noise level quite a bit lower (~1100-1200rpm or ~70-75%) on prime95, provided you don't run into the turbo-hack. If you do run into the 4400Mhz, it brings the fans up to max, even outside the case.

5. Maxing out the Noctua fans manually on prime95 will not bring the temperatures down. I still get 68-70°C amongst the cores. Note that these are _core_ temperatures, as stated in #1.

6. I also got the NF-S12B redux, and tried it (without the case, just putting it approximately into place). It seems that on air cooling performance is really independent on whether I have the back 92mm on the heatsink, the S12B, or both. Inside the case is a bit more complicated, see below. What is interesting is that the fan seems to hum at certain PWM percentages: it has no sound at ~40%, and over 60%, but at 25-30%, 55-60% and ~75%, there is definitely a low pitched, very apparent hum. Unfortunately the Silen BIOS setting idles the fan at 30%.

6.b: Aaaand my old Noctua 120mm started sputtering after repeated loads, so I went and grabbed an older Scythe SY1225SL12LM-P that came with my Mugen 2.revB. It seems to work so far.

7. edh: it appears the m.2 slot is accessible even with an unmodified backplate. Although unfortunately AFAIK the only PCIe m.2 cards that offer a huge speed boost are too big for this motherboard (it supports only 2230 and 2242 sizes). There are some SSDs available in this size locally, but they are generally slower than a regular 2.5", and a bit more expensive.

8. I enacted some cable management to get rid of cables in the way of the airflow, attaching velcro cable ties to the top of the case, and putting in an unsleeved PSU cable extender that allowed me to have a much tighter turn radius.

9. There is a new nasty coil whine coming from the PC. It might be the result of using the PSU cable extender, so I'll try removing it, although I will have to disassemble the whole thing again. I'm too tired right now for that.

Not-so-Fun fact: I have disassembled the sides so many times that one of the screw holes is partially stripped now.

I assembled the machine again, and after many tries, went with a combo of one 92mm fan between the PSU and the heatsink, and the old Scythe as the exhaust fan. I've tried having the 92mm with ducting, and also replacing the front 92mm with my Nexus 120mm. As the 92mm Noctua has the same airflow and higher pressure, it is not that much surprising that the best combo seems to be having a 92mm at the front and the scythe at the back.

Pics:
m.2 slot + some visible cable ties at the top
Image

Through the window: not much cabling visible
Image

...unless you are viewing from the same level
Image

WIP shot, still with 3 fans and not final cable management (I tightened them and moved that USB3 double cable out of the way)
Image
Last edited by nagi on Wed Feb 03, 2016 8:26 am, edited 1 time in total.

ggumdol
Posts: 154
Joined: Mon Aug 05, 2013 1:01 pm
Location: Scandinavia

Re: Troublesome mini-ITX

Post by ggumdol » Fri Jan 08, 2016 4:06 pm

Hi nagi,

I'm glad to hear that it's working now anyway.

1. I also experienced the turbo hack of Asrock's Haswell motherboard (B85M Pro4). It didn't matter to me, and actually increased the overall CPU performance of i5-4430 which has the lowest clock among all i5 Haswell CPUs. It's good to hear that you have managed to get them back to the normal status.

2. Prime95 produces unrealistic workloads, perhaps not even close to what your CPU would experience during programming and compiling. Please rest assured that everything will be fine. I recently found that many SPCR reviews run only 2 threads in Prime95 tests (even blend ones?) on 4 core CPUs.

3. It might be high time to buy some cable ties to improve the status quo of your cables following the school of tentacle monsters, unless you are a genuine follower.

4. I think Asrock made a "grave" mistake of placing the CPU socket far from the PCI-E slot when it comes to their Haswell ITX motheboards. Now I managed to recall that I didn't go for Asrock ITX motheboards at the time partly because of the CPU socket position. I sincerely hope they rectify this mistake in the near future.

daniel_r35
Posts: 1
Joined: Thu Jun 18, 2020 5:23 pm

Re: Troublesome mini-ITX

Post by daniel_r35 » Thu Jun 18, 2020 6:02 pm

Hi Nagi. Sorry to revive an old thread. I created an account specifically to ask you this. I did PM you but it got stuck in outbox for some reason.

First off, THANK YOU for posting your build. You are the only one in my search that has used a Metis and a motherboard with a CPU located too far 'North', while trying to slap a decently wide cooler on it. I am planning a mini-itx build and I have a Gigabyte X570 Aorus pro wifi ITX board, and I want to use a Deepcool Assassin III cooler (140mm wide). I used pictures of your motherboard and mine on autocad Inventor and estimated you had roughly a 13mm overhang between your cooler and your motherboard. I estimated mine to be roughly 11mm. I am really concerned about whether I can fit this cooler 'width-wise' into the case (I'm well aware that it's 165mm tall, but I a concerned about the overhang).

My question is, you mentioned you would ideally like the CPU to be ~15mm further south of the motherboard. However, from your pictures, it looks like you got everything to fit (in the front to back configuration, not the 90 degree rotated one)! You managed to close all the panels. So I'm still a bit confused as to why you said it didn't fit :?

Thanks for taking the time.

Daniel.

nagi
Posts: 125
Joined: Tue Aug 31, 2010 6:48 pm
Location: Outside the box

Re: Troublesome mini-ITX

Post by nagi » Sun Jun 21, 2020 6:03 am

Hi.

NP. I would have liked the extra 15mm because roughly that's how much the Scythe Fuma was overextending into the bottom case panel. I'm not sure, but I might have tried to assemble the case without the bottom panel to try to figure out the total overextension. In the end, I got everything to fit by changing the CPU cooler. In the later post you can see it with the Noctua NH-U9B SE2, which has 92mm fans:
viewtopic.php?p=598747#p598747

If I remember correctly, I just couldn't close all the panels of the case with the 140mm wide fuma comfortably. The 15mm extra space would probably have helped in that big time. (Also, the Fuma does not support 90°rotation, otherwise I would have tried that.

Post Reply