Yet another O2 PSU fan replacement

SGI hardware problems, solutions, tips, hacks, etc.
Forum rules
Any posts concerning pirated software or offering to buy/sell/trade commercial software are subject to removal.
User avatar
dexter1
Moderator
Moderator
Posts: 2791
Joined: Thu Feb 20, 2003 6:57 am
Location: Zoetermeer, The Netherlands

Yet another O2 PSU fan replacement

Unread postby dexter1 » Sat Nov 07, 2015 5:25 pm

So i've been lurking on the forum for a modern equivalent to the Panaflo FBA08A12U in my O2 PSU. It's a racket and i cannot stand it 24/7
I have an R12K@270Mhz, 256GB ram, AV module and one ATLAS V 18GB disk

There have been many threads about this specific problem/solution, but often the proposed models don't exist anymore. So idid a bit of research.

So the Panaflo specs are: 80x80x25 mm, 12V, 3.2W, 38.2dB, 3450RPM, 79.8 m^3/hr, 46.9 CFM

I've found this Noctua in a previous thread:
Noctua NF-R8 has 17dB, 1800 RPM, 53 m^3/hr, 31.2 CFM

Another Noctua which looks to be even better:
Noctua NF-A8 FLX 16.1 dB, 2000RPM, 50.4 m^3/hr

The good ol'e Papst models are still available:
Papst 8412 NG has 32dB, 3100RPM, 69 m^3/hr, 40.6 CFM
Papst 8412 NGME has 26 dB, 2600RPM, 58 m^3/hr, 34.1 CFM
You want the G models with sleeve bearing, more silent. M means medium speed, E means rounded flange

My gut feeling for my O2 would be installing the 8412 NGME dropping 10 dB level while maintaining 3/4 of airflow compared to the Panaflo. Since i will build software again, this machine needs to have substantial cooling capabilities. I'm just not sure about the Noctua's. They seem to provide a lot of throughput, but can they sustain the performance in a cramped O2 PSU casing?

I have a second O2 system if the new fan fails to perform: a R5K "Cube" O2 would be perfect for a fan which doesn't live up to cooling the R12K

Any other recommendations? if not i'll go with the NGME
:Crimson: :PI: :Indigo: :O2: :Indy: :Indigo2: :Indigo2IMP:

User avatar
hamei
Posts: 10535
Joined: Tue Feb 24, 2004 4:10 pm
Location: over the rainbow

Re: Yet another O2 PSU fan replacement

Unread postby hamei » Sat Nov 07, 2015 6:14 pm

dexter1 wrote: Since i will build software again, this machine needs to have substantial cooling capabilities. I'm just not sure about the Noctua's. They seem to provice a lot of throughput, but can they sustain the performance in a cramped O2 PSU casing?

I ran the R8 for two years in a non-air-conditioned room, summer temps up to 37* C, without any problems. A hard drive finally failed, I s'pose that could have been heat-related, but it was old.

The only caveat I have is, that was an r5-350. The 5's run cooler, I believe ?

The Noctua was very quiet. I couldn't hear the computer over the noise of the cockroaches scuttling around the garret.
hey friendly ! come outta there ! you're a cheap lousy dirty stinkin' mug and I'm glad what I done !

User avatar
Trippynet
Donor
Donor
Posts: 900
Joined: Thu Aug 15, 2013 6:22 am
Location: Aberdeen, Scotland, UK

Re: Yet another O2 PSU fan replacement

Unread postby Trippynet » Sun Nov 08, 2015 3:41 am

I too have run the Noctura NF-R8 in my O2. It ran almost silently, and I didn't have any problems at all with it. However, to follow on from Hamei's comments, I too had a lesser CPU. R5000 200MHz in my case.

You could try the A8 and see how it runs? Most of the Noctura fans come with a resistor you can add in to lower the fan speed, so if you find it's still noisy, you can try a slower speed to see if it helps (keeping an eye on air temperature out the back of course!).
Systems in use:
:Indigo2IMP: - Nitrogen: R10000 195MHz CPU, 384MB RAM, SolidIMPACT Graphics, 36GB 15k HDD & 300GB 10k HDD, 100Mb/s NIC, New/quiet fans, IRIX 6.5.22
:Fuel: - Lithium: R14000 600MHz CPU, 4GB RAM, V10 Graphics, 72GB 15k HDD & 300GB 10k HDD, 1Gb/s NIC, New/quiet fans, IRIX 6.5.30
Other system in storage: :O2: R5000 200MHz, 224MB RAM, 72GB 15k HDD, PSU fan mod, IRIX 6.5.30

User avatar
Krokodil
Donor
Donor
Posts: 502
Joined: Fri Apr 17, 2015 2:32 pm
Location: The House of Particular Individuals

Re: Yet another O2 PSU fan replacement

Unread postby Krokodil » Sun Nov 08, 2015 11:34 am

Also have Noctura NF-R8 in my O2 with R5000 200MHZ. The panaflo fan was just too noisy. Now it's so quiet I can hear the hard drives chirping from time to time.
:Octane2: - :O2: - :Octane: - :Indigo2IMP:
Alphaserver DS10 - 466MHZ
Sun Ultra 5 - 440MHZ.

User avatar
dexter1
Moderator
Moderator
Posts: 2791
Joined: Thu Feb 20, 2003 6:57 am
Location: Zoetermeer, The Netherlands

Re: Yet another O2 PSU fan replacement

Unread postby dexter1 » Sun Nov 08, 2015 1:22 pm

Thanks for the responses, i'm holding off from purchase awaiting experience reports from people with R12K's
:Crimson: :PI: :Indigo: :O2: :Indy: :Indigo2: :Indigo2IMP:

User avatar
dexter1
Moderator
Moderator
Posts: 2791
Joined: Thu Feb 20, 2003 6:57 am
Location: Zoetermeer, The Netherlands

Re: Yet another O2 PSU fan replacement

Unread postby dexter1 » Mon Nov 16, 2015 5:18 am

Since after one week no fan replacement have been recommended for R12K O2's, i therefore have ordered the Papst 8412 NGME. Should arrive tomorrow and then i have to reserve a couple of hours of getting the fan replaced with all the glue and stick attached properly.
:Crimson: :PI: :Indigo: :O2: :Indy: :Indigo2: :Indigo2IMP:

Jack Luminous
Donor
Donor
Posts: 212
Joined: Mon Sep 26, 2011 12:59 am
Location: PARIS

Re: Yet another O2 PSU fan replacement

Unread postby Jack Luminous » Mon Nov 16, 2015 5:45 am

Maybe I have higher tolerance to noise than others but I find my bone stock O2 (R12K400) not too noisy. I doesn't bother me at all and I don't feel the need to do anything to it. My Octane2, on the other hand, is like a jet-plane taking off. I might have to do something to it to calm it down. :lol:
:Octane2: :Octane: :Octane: :O2:

rvic
Posts: 56
Joined: Sun Feb 08, 2009 6:55 am
Location: Romania
Contact:

Re: Yet another O2 PSU fan replacement

Unread postby rvic » Mon Nov 16, 2015 3:11 pm

Hi, i have an O2 with r12k 400mhz cpu ... and here are the facts:
The original fan Panaflo FBA08a12u is a very noisy beast but: it moves a lot o air and i found that an O2 with R10k processor need it.
The air flow is 46.9 cfm at speed 3450rpm with a stock noise of 38.2db ... it's adjustable from 7.0 to 13.8 Vdc ...
The Sony Power supply does a 3 mode Voltage adjust depending on temperature ... but in my tests with my r12k proc. even during winter ( and used outside in like 20 Celsius) doesn't want to switch to a lower speed .... When starts it cycles through the 3 states low rpm to max and stays there ...
Now all that came from a defective pcb design though the main engineer states that some thing needed to be changed because in tests it shows that on ambiental 40 grades celsius the system will fail ... and we're talking about r5k procs. here is the page ...

http://www.erimez.com/misc/O2/analyst/O ... index.html

As you have probably see his conclusions are :

"The simulation results show that the temperature of I/O ASIC, CRIME ASIC and CPU may be too high when the system is operated in maximum power dissipation and the ambient temperature is at 40 C. This high temperature may cause system failure or significantly reduce system reliability. The main reason for this high temperature is lacking of cooling air. Since most air is blocked by the PCI riser connector and the plastic model around the connector, the ideal flow path is hard to reach. To reduce this high temperature, following suggestions has been made:

move both PCI riser connector and the SIMMs to a vertical position so that the main air flow stream will not be blocked
use high heat conduction package, TBGA, CPGA, to replace high power disspating plastic packages
close part of the front panel openings to force more air flowing through CPU board"

However his suggestions were never taken in consideration and the O2 was left with a problematic design ... At that time they probably never anticipated that newer cpu-s will be fited into that case ... that was even worse ...

So to cut the chat ... because the original fan was too loud i replaced it with a Sunon at about 30 cfm ... and of course as expected the first problem that appeared was the ram ... it started giving ecc errors ... that was during summer in 28 celsius ambient ... now it stopped because it's winter and in house temperature is about 24 Celsius

If you look in my past posts i launched this ideea in the "o2 ram death plague" post where i said that i suspect that it's related to temperature ... but it was just an ideea ...

Now i'm sure of it ... in fact it's explainable cause the memory is paired like in a Raid0 matrix ... so therefor the temperature is high ...

The little fan in the r10k-r12k heatsinks is hilarious because it spreads the air in a wrong direction ...
In the end i recommend now (for me also :) ) a Thermaltake tt-8025tu smart fan with a max 75.70 cfm ( has a ntc thermistor input connector so it can be temperature controlled ) and it switches from 17db, 1300rpm, 20.55cfm at 20Celsius to a max 48db,4800rpm, 75.7cfm for 55 celsius ...
This is more suitable in my opinion (it can be tweaked easily to stabilize at around 25 db ) because in takes the right amount of air out ...

R12k cpu-s were too much for this case and pcb design ... but at that time (around 2001) was too pricey to make adjustments ... so this is what we're stuck with ...
Sorry for my log post and my bad english ... this was one of my research two years ago regarding my little O2 ... and i just found this opportunity to share it ...

User avatar
hamei
Posts: 10535
Joined: Tue Feb 24, 2004 4:10 pm
Location: over the rainbow

Re: Yet another O2 PSU fan replacement

Unread postby hamei » Mon Nov 16, 2015 6:41 pm

rvic wrote: in tests it shows that on ambiental 40 grades celsius the system will fail ...

Only a guy from Hong Kong would model at 40* C ambient :P Have you ever been there in summer ? It's like the third circle of Hell ...

Otherwise, very interesting ...
hey friendly ! come outta there ! you're a cheap lousy dirty stinkin' mug and I'm glad what I done !

User avatar
dexter1
Moderator
Moderator
Posts: 2791
Joined: Thu Feb 20, 2003 6:57 am
Location: Zoetermeer, The Netherlands

Re: Yet another O2 PSU fan replacement

Unread postby dexter1 » Tue Nov 17, 2015 3:22 am

Thank you Rvic for the insight in O2 air cooling. Unfortunately the Papst fan is on its way already, and will test it soon with the R12K.

Mitigating factors is that i have a 270MHz model and ambient temperature do not exceed 22 degrees in the summer in my home office. Agree with Hamei that 40 degree Celsius is way too high ambient temperature to model heat dissipation, especially in continental Europe.

With the added knowledge that i can block the front panel openings to allow more air over the PCB, i think this will be workable with my Papst. If things go bad i can get a higher specced Papst. I am glad that i did not go for the Noctua's since all their models available in the Netherlands move too little air.
:Crimson: :PI: :Indigo: :O2: :Indy: :Indigo2: :Indigo2IMP:

rvic
Posts: 56
Joined: Sun Feb 08, 2009 6:55 am
Location: Romania
Contact:

Re: Yet another O2 PSU fan replacement

Unread postby rvic » Tue Nov 17, 2015 1:58 pm

It'll be fine ... your fan choice is better than mine ... and your room temperature is good for electronics :)
Unfortunately i worked for a company that used excessively air cooling and i developed a sensibility to that ... and i no longer use air conditioning at home ...

But i'd like to clarify some facts :

Regarding the little joke from hamei ... :)

Well at some point i worked for a company that was a huge fujitsu seller and we had a training in w600 workstation (which is a standard in low noise) regarding thermal vs noise research done by fujistsu-siemens.
And the ambient temperature is considered the lowest temperature in that specific cage after stabilization point ( after thermal acceleration and rise of all overall temperatures).
So that is the temperature used to cool the the higher ones and also the one that determines the stabilization point.
I remember that the w600 case was made from aluminium alloy so it will help to a lower ambiental point ...

You can see from the link in my previous post that the purple-blue colour on the main board is 40 celsius and is considered the lowest and ambient temp ... we can speculate it's not equal to room temperature

The O2 with r10k-r12k cpu's has this small fan attached to the cpu heatsink and it forms an air cushion (wall) exactly on the lower margin of power supply box ... there are some round holes there .. then are the two big holes for the 2 Hdd (first hdd is occupied by the pci riser on our boards because of the bigger cpu heatsink) and a big hole for the video board ...
Now the problem is that the power supply box starts having this small holes (on the lower side) from half to back ... that means that this air wall is hitting metal and returns back and some part goes to sides (in the ram compartment) ...
This helps to thermal acceleration and to a higher stabilization point ... if the main fan (power supply) doesn't take a bigger volume of air out we'll likely have an ambiental temperature around 35 though room temperature is 24 ... so a 35 celsius is used to cool a 50 celsius (cpu heatsink) ... this is what is happening inside ...

More ... every time i took out the mainboard i couldn't touch the heatsink or ram modules (too hot) ... even when outside is 24 ...that means exactly high thermal acceleration ...

I never finished this research ... i was about to make some measurements to different components inside (when it's working) with some sensors ... and have a better picture of all temperatures inside ...

Another issue ... i lost already a ram dimm of 128Mb and always the one that fails is the upper one ... the one that is taking all temperatures from the lower ones ( i have 4 dimm of 128 from lowest to half ) ... and i did a reflow on the defective one but it wasn't a matter of desoldering but it failed a certain chip ... and because all are tolerant to electrical signals but temperature the conclusion is clear ...

This is the hot air path :
The temperature of heatsinks and chips under the cpu heatsink are rising up ... they warm up the first ram dimm which warms up the next dimm and so on until it reaches the last ram dimm ... that one has the higher temp (and most likely will fail) ...
Now all this path is covered by the the air wall build up by the small cpu fan ... and the main fan will evacuate only what this fan generates ... leaving the mentioned path mostly untouched ... it leads to high stabilization point ... and perhaps failure ...

My idea was to fit another fan on the upper ram area to force moving the air from there to power supply box ... or to change the main fan to a more powerful one so it has a chance to move the air on ram area)

If you noticed when you mount the power supply box ... the round holes (on the side) just overlaps perfectly ( that is a picture how serious the design was ... i'm still amazed by the quality of this computer ) ...

Of course because the case is made out of plastics it helps on thermal acceleration cause there is no possibility for an external thermal dissipation (by radiation) ... despite the fact that it has a metallic shell which is covered by this plastic design (it never gets a chance to cool down by radiation)

I noticed the material for the cpu heatsink is special (didn't encounter a similar alloy yet) so i think that was an object of research from their part ... also from other sgi models pictures i think the heatsink is unique (specific) to O2's ...

As long as air is moving across a component (even hot) it will not fail (because of overheat) ... i'm an electronist ... so i have some experience in that ... that is the reason plasma tv's without active cooling will fail eventualy (the y module mostly ) ... cause the hot air is not swapped fast enough ...

I'm sorry again for a long post ... when i'm gonna finish this little research i'll come with better numbers and solutions ...

Regarding your choice for the main fan ... i think it's good ... maybe you could post a hdd temperature from that little smart app ... it's somewhere here on nekochan ... or you can make an estimate putting your hand on exhaust (hot or middle hot) :)

User avatar
jan-jaap
Donor
Donor
Posts: 5071
Joined: Thu Jun 17, 2004 11:35 am
Location: Wijchen, The Netherlands
Contact:

Re: Yet another O2 PSU fan replacement

Unread postby jan-jaap » Tue Nov 17, 2015 3:30 pm

rvic wrote:I'm sorry again for a long post ... when i'm gonna finish this little research i'll come with better numbers and solutions ...

Never mind. That is some properly interesting research, even if I don't understand all of it.

robespierre
Posts: 1698
Joined: Mon Sep 12, 2011 2:28 pm
Location: Boston

Re: Yet another O2 PSU fan replacement

Unread postby robespierre » Tue Nov 17, 2015 3:59 pm

Cooling efficiency is expressed in units of W/K (watts per Kelvin, same as watts per degree C). It's the thermal analog of conductance, so you can also express the thermal resistance by K/W. The resistance of the die to the package, the package to the heatsink, and the heatsink to ambient are all added together to determine the system efficiency. Temperature at the die itself (the junction between bulk silicon and the gate) can be at maximum ~150°C depending on the fab process. This may not be the same temperature reported by software, since sensors are in different positions in relation to the junction and see different temperature effects. Each step outwards adds resistance, so the temperature is lower and lower. Ambient is idealized as an infinite sink, which is not completely realistic; but for use in an office or datacenter environment it's approximately true, the ambient temperature is controlled and the computer system doesn't affect it.
One of the issues with older systems (10 years and more) is that the thermal resistance of the system increases, primarily by degradation of the paste or other thermal interface material, but also because of dust and worn out fans. This makes the hardware less reliable and increases the fan speed (and fan noise) required to prevent overheating at high load. They didn't start out loud.
:PI: :O2: :Indigo2IMP: :Indigo2IMP:

User avatar
hamei
Posts: 10535
Joined: Tue Feb 24, 2004 4:10 pm
Location: over the rainbow

Re: Yet another O2 PSU fan replacement

Unread postby hamei » Tue Nov 17, 2015 5:08 pm

robespierre wrote:Cooling efficiency ...

This is probably no longer applicable (but then, neither are our computers) but my Electronicker Friend Fred from JPL (Voyager was his last project, Fred was in charge of the optics) said they had a basic rule-of-thumb : for every ten degrees C of increased operating temperature, electronics life was halved.

Seems like they got something right ...
hey friendly ! come outta there ! you're a cheap lousy dirty stinkin' mug and I'm glad what I done !

User avatar
dexter1
Moderator
Moderator
Posts: 2791
Joined: Thu Feb 20, 2003 6:57 am
Location: Zoetermeer, The Netherlands

Re: Yet another O2 PSU fan replacement

Unread postby dexter1 » Mon Dec 21, 2015 7:12 am

A rainy monday afternoon. Perfect for getting this project off my list. I just popped in the Papst NGME after disassembly of the old Panaflo. Opening the PSU was very easy and most components were easily accessible. What an excellent piece of engineering!
I bought a can of compressed air, since if the PSU was open, might as well clean it.

The sticky tape on the two plastic airflow guiders attached to the Panaflo was finished, so i got some Tesa double-sided tape and reapplied it.
Popped in the Papst, re-assembled and connected everything and it booted right away. Fan is spinning? Yup! Job done :)

Then i noticed the greenish hue on the monitor and remembered i harvested an old VGA cable to break off some pins so i could get proper sync-on-green signal for my DELL monitor. Worked like a charm!

After half an hour of gaming, xv and camera doodling, the temperature on the back of the PSU was 34 degrees celsius, measured with an old thermometer. Ambient temperature is 17 degrees celsius.

Now that the Panaflo woosh is gone, i noticed two things: a low buzzing humm and a high pitch squeak coming from the disk. The humm might very well be from the CPU fan, since it's an R12K. It's not so much the airflow but the vibration which is annoying. So i need to get that one replaced as well.

All in all, a simple and fun mod. I'm happy for now.
:Crimson: :PI: :Indigo: :O2: :Indy: :Indigo2: :Indigo2IMP:


Return to “SGI: Hardware”

Who is online

Users browsing this forum: Ahrefs [Bot], Majestic-12 [Bot] and 4 guests