New system - Silencing - Plan & Testing
[ Index ] - of Computer Noise Report
Overview - Trying to structure chaos :)
> 05.MAR 2001 : Started building a new silent PSU.
Table of content : Bookmarks
< Fill in info >
The "Doomsday" CPU Element(4)
Planned steps : Excecution & Followup
1. CPU fan related issues.
1.1 Reducing fan operation from 12V to 3V by installing noise-killer (fan-regulator).
Q : Will this cancel the monitoring abilities of the MB ?
1.2 Improve stability of MB under CPU socket base.
Serious vibrations resonating throughout the entire chassis. Done this before. Doing it at this early point may increase the life-time of fan. Using vibration-dampening & shock-absorbing material underneath the MB.
This was effective. Added a spots of shock absorbing rubber at stategic positions on the MB. Very well.
1.3 Improve stability of cooling rib on CPU.
Adding "shims" of some sort between rib and processor. This will eliminate some of the "wobbling" of the fan. As it is it's like balancing a helicopter at the top of a flag-pole. The "shims" should be of some non-conductive, and possibly heat-isolating material. ( I'm not sure whether it is a point to insulate all but the cpu-die from the rib by air, but assume that the creators of the chip and rib (boxed-type) knows what they're doing ).
From what I've seen it appears that this is called "spacers", and is supposed to "improve" heat conduction and ribb stability. I'm not shure about the value of the heat conductions stuff, but improving stability is a point particularly with the celeron FC-PGA pacaging. The custom made cooler exceeds the weight specs. for the ribb by miles, and something has to be done.
1.3.1 Made one copper-spacer.
Quite time consuming. Hope that it works. [Testing/Fitting/Picture]
1.3.2 Made alternative element support.
A heat tolerant shock absorbing rubber foam stuff where used instead. Easy to make. Very usefull for fitting the element on the die without getting the "twists".
2. Cooldown. Decrease Vcore & Vio voltage of CPU.
Since the CPU clock-speed is derived from the FSB speed, it is not possible to clock down the Celeron processor. The only way is up. For a P3 however it would be possible to clock down from 133 MHz and all the way down to 66 MHz. How large a factor the CPU speed is in heat generation I do not know, but migh look at this at a later stage. Possibly when my system becomes to slow, but as of now there is plenty of power.
The initial system set-up was as described in latter parts of section 1.1. Recap : CPU Fan regulated down to 3.77 Volts, apprx. 28 C CPU idle temperature.
2.1 Decreasing Vcore from 1.7 to 1.5 volts :
CPU idle temperautre falls to 26.5 C, also observed lower peak temperatures under load. System appears stable. (not tested).
* Specify the difference, what was the temperature before. Peak measurments.
2.2 Decreasing Vcore from 1.5 to 1.4 volts :
CPU idle temperautre drops to 25.5 C. Lowerd peak temperatures under load. 29.5 C @ 25% Load. Quick stability test (using CPUStability test 6.0) holds for 15 minutes without any problems. Peak temperature more than 10 C lower than at standard settings.
* Specify the difference. Idle/Load measurments.
What effect does SWC or ACPI have ? Dissable and check ... Pops up 5 C instantly ... The SWC makes a difference of about 19 C in CPU idle temperature. Huge difference.
2.3 Decreasing Vcore from 1.4 to 1.3 volts :
Can't recall this. Think temperature where sniffing at 24 C, but the impression is that system was unstable. Redo another time.
Update : System failed to boot ( Note ! Vio set to 3.4 volts in BIOS, but boots at 3.5 volts )
Q: What is the effect of decreasing the Vcore voltage on CPU temperature. ? [2.x] Any ?
A: Great effect in lowering both idle- and peak CPU temperatures.
2.4 Decreasing Vio from 3.5 V (default) to 3.3 V (stable at 3.4 V ) :
For some strange reason the default setting for the Vio voltage on the MB is set to 3.5 Volts. Changed this setting on a hunch. The effect seemed to be a slight drop in CPU tempeature when measured without SWC. (From 44.5 C to 43.0 C).
The +3.3 Volts low-limit in the Winbond HW Doc monitor program was also changed to reflect the new setting. Changed from +3.39 V to 3.20 V.
Got some problems with the high-point controller poping up despite beeing disabled in BIOS. Had to increase the Vio to 3.4 to make the problem go away. Hummm ... Strange stuff. Don't know the standard setting.
2.5 Decreasing Vcore from 1.4 to 1.35 :
System appears to be stable. Temperatures lower. Typical idle temperatures @ 21.9 ambient : CPU 25.5 C, MB 27.5 C. Quick test of Load temperatures @ same ambient : CPU 41.5 C & MB 28.0 C. Measurements done 25 min. after dissabeling SWC. Slightly lowered temp.
3. Creation of custom cooling rib system for PC-FGA type CPU.
3.1 New ideas. The CuCo Tube.
Idea temporarily parked. Reality testing didn't workout as well as wished.
3.2 The "Doomsday" cooling element.
Work in progress : Copper-spacer complete. Will try to test the first custom cooler element to get an indication of the required dimentions of the new element. Skipped this, made the element on feeling. But test the first element anyway so that the difference in effectiveness can be measured.
3.2.1 Testing of the first fanless cooling original element, Q4 Element(2) :
Follow up in seperate file, or bookmark within this document. [ Q4 presentation ]
3.3 Limited overclocking testing.
See how far up system is stable with minmal Vcore and Vio settings. Do this with minimal settings first. Do the extreemes later. Followup in seperate file, or bookmark.
4. PSU issues.
After the burn-in test of psu(2), move it into the current system. Temperature testing with or without vent.fan. under different loads.
Gather sample data using mb-monitor of the operation of the current psu(3), and compare
this with the results for psu(2).
The maximum values ever recorded are :
4.1 The PSU(2) where inserted into the cabinett and tested.
There where no major differences in monitored voltages between PSU(2) and PSU(3).
5. Cabinet Issues :
Airflow / ventilation.( Increased and altered cabinet perforation. Ventilation fan's ?
- Cabinett modified sucessfully. Extremely time consuming. Significance noticed in the testing of PSU(2) with system(1). With system(2) the effect is noticed by a more than 3 C drop in cabinett temperature. And that is the difference without cab.fan, versus with.
6. Fallback systems. Overheating security.
(Fan's & regulation of ... tuning of noise-killers)
oo - The ABit MB has a super safety feature in BIOS. This enables the system to automatically shut down at a critical temperature. The minimum setting are critical though, but functions like a MAIN fuse in case of disaster. Another draw-back is that this auto-shutdown feature only takes into account the CPU temperature. I'm more concerned about my HD information being fried, than some expendable processor. Now this problem might be solved with newer IBM drives, that have built in overtemperature protection. Seems like they've started to use the SMART information present in their drives :).
6.1 Making the stuff for psu(4). Built around the exsisting fan-regulatur of psu(2) :
Mess mess mess. Trouble trouble. But the idea seem to work, although there is a minor clash with reality. Takes 10 times longer to do that imagined.
Monitor noise :
This problem solved itself as the 17" monitor(1) failed. And I got an 15" with a level that is lower than the machine noise. This problem also seems to disappear with time, as LCD monitors starts to become affordable.
Minimizing HD activity :
The only source of noise left in the system is that which is generated by the drives. The HD Noisebox dampens this to a decent level. After numerous modeficatons of the prototype, the hopes of getting a 0dB drive using this approach has been abandoned. More drastic measures are needed, and I don't really have the time or the energy to create one at the present.
To minimize the HD activity the windows vcache settings are optimized using the postcardware utillity "cacheman 3.70" by Thomas Reimann [ http://www.outertech.com ]
To futher minimize HD access, some experimentation with a ram drive was also tried. Using the RamDrive98 shareware tool [ http://www.jlajoie.com/ramdsk98 ] to create a 64+ MB drive. On this drive the windows swap file for the virtual memory settings were placed. This had little or no effect on windows related disk access. Placing mp3 files on this drive worked great however.
OO - What is needed is a great windows disk cache utillity, optimized to minimize windows disk activity. I also note that most of the programs that I use manage to keep their fingers off the drive most of the time, but most of the MS stuff have to sniff at some "#¤"#¤ .dll or what do I know at every screen scroll. MS-Windows prooves the saying : Less is More. I just totally give up trying to eliminate win disk accress, it will take years off my life and tonns of pills to get by. I'll spend the energy on creating a customized software-independent zero-dB drive instead. I'll expect it will be much more rewarding than spending eternity re-booting windows :)
The HD powerdown setting where set at 1 min in BIOS. This is a bit too agressive when working with programs that is disk intensive, but fine with other programs. f.eks. fruity-loops.
The Virtual memory settings where disabled. Unceritain results. Have caused a unstable system in the past.
Conclusion : Optimized vcache settings. More memory. Aggressive HDD powerdown setting.
A bit late, and highly inaccurate ... and limited to the scope of this document.
Idle state. Load state. m.m.
Report : Testing PSU(2/4) in System(2)
After the testing on system(1) was sucessful, it was time for testing it with system(2). Note that the CPU fan is still on, as the custom cooler for the celeron is not ready yet. This will be done later. Also note that the PSU grid and blower-fan (just in case) has not been mounted in the testing below.
Minimal Testing :
Started out with a minimal setup to see if it would hold at all. No CD-RW/Floppy/Soundcard etc., only Video-card(2) and HD. Open cabinett. ACPI/SWC on. CPU fan on.
The system booted up fine. And when the HD powered down, the room was filled with a strange; uncomfortable silence.
The PSU & system voltages was withing the range of PSU(3). Monitored using the Winbond HW Doctor program. The +12 V and +5 V voltages was slightly higher than before; around 0.1 Volts higher.
Tested for 3 hrs 45 min.(stable temp reached before this)
Conclusion : stable at 45 C @ 19.7 ambient (delta 0.0025 C/min) on idle with open cab.
Power doodling : see findings in table below (Power consumption)
Full scale testing :
After the minimal test was passed, all the normal stuff was plugged in. The PCI card was replaced with an AGP card. Soundcard added. CD-RW added. Cab.cover on. ACPI on. CPU fan on.
Testing started off from standby and a PSU temp of 27 C @ 20 C ambient.
Tested for + 8 hrs (stable temp reached before this)
[ Table of measurments to be added ]
Other related measurements :
Conclusion : stable at 56.5 C @ 22.1 C ambient on idle with closed cab. ? Other temps at this state
Update #1 (08/02/01) : Measurments of PSU temperature another day with open cab.
Disabled the ACPI feature & swapped the AGP with the PCI card. Windows troble. Booting Booting. But anyway. After this heavy use psu temperature peaked at around 50 C @ 22.8 C ambient. That is with all of the equipment of the Full scale testing. Normalized this is about 2 C above the minimal testing result. This result is also more directly comparable with the full testing, and indicates that the difference between cab. cover or not is around 7 C. So there is still room for improvement.
Other related measurements :
Update #2 (09/02/01) : Overnight testing of PSU temperature with closed cabinet.
The system was left idle overnight to find the stable idle temperature. Video and HD power down.
This gave the following measurements :
Load Testing :
Diverse testing of system under load. See it is stable under extreme conditions. Disable SWC and use CPU Stabtest. This will be with done with fan on CPU. Do other testing later (different Vcore later, when custom cooler element (2) is ready).
Testing with SWC disabled successful. Readings when connected to the universe.
Testing with CPU Stab. Test V 6.0. CPU warming feature (maximized power-consumption).
No significant difference from the testing with SWC disabled. A 1.5 C increase in CPU temperature, and a 0.7 C increase in psu temp after 1 hour of testing. Seems like the function of the test program is to twist that last drop of power out of the system.Stable at any system load with this setup.
A snap shot of the open cab tempearature measures :
These measurments shows that cab. ventialtion is very good. Small differences with or without cab. cover.
Final Testing :
With psu shield and backup blower-fan/regulator on.
TEST 1 : Open Cab. Shield on. Load testing : Peak psu core temperature measurment
Prelude : The SWC have been disabled for a small hour ... starting measurments from there.
22:32 54.7 C @ 21.9 C
22:45 56.4 C @ 22.0 C Ohhh ... delta 0.15 C/min ... ETA ?
23:03 58.2 C @ 21.6 C delta 0.10 C/min ... Guess 61.5 or 63 C in 1 to 2 hours.
23:17 58.8 C @ 21.8 C delta 0.043 C/min
23:32 59.5 C @ 21.9 C delta 0.046 C/min conv. 20 min pr. C
23:44 59.8 C @ 22.0 C
23:49 60.0 C @ 22.1 C delta 0.4 C/min
00:45 61.4 C @ 22.3 C delta 0.025 C/min v 40 min. pr C
01:30 62.0 C @ 22.4 C delta 0.013 C/min v 75 min pr C
02:25 62.1 C @ 22.4 C delta 0.0018 C/min v 550 min pr. C
02:37 62.2 C @ 22.4 C
02:42 62.3 C @ 22.2 C
02:45 62.4 C @ 22.2 C ??? Wobbeleing a bit.
02:48 62.3 C @ 22.2 C
02:49 62.4 C @ 22.2 C
> Concludes stable : 62 C @ 22.4 C ambient open cab.
TEST 2 : Open Cab. Shield on. Idle testing : Peak psu core temperature measurment.
Leave overnight and let it cool down. HD and Video will power down. Lower than typical measurment values.
> Result.1 : ..... 56.4 C @ 20.6 C. ( HD has been kept on by a disk tool).
> Result.2: ..... 55.7 C @ 20.9 C HD Off. Very moisty air. Thick fogg outside.
TEST 3 : Closed Cab. Shield on Idle testing.
23:04 59.3 C @ 21.6 C HD on (1 min PD setting) , Video on, 85/90 % idle on average.
00:04 60.5 C @ 22.0 C Average 70 % Idle. delta 0.02 C/min v 50 min pr. C
Temporery conclusion : stable on idle at around 60 C @ 22 C : Validate overnight.
> Result 1 : 58.2 C @ 21.5 C HD off (1 min PD), Video off, 100% idle.
Temperatures : MB (28) HD(28) CPU(26)
TEST 4 : Closed Cab. Shield on. Load testing.
Prelude : Kicks off from peak idle measurment. Test Limit 65 C
00:07 60.5 C @ 22.0 C SWC disabled.
00:12 60.9 C rough delta 0.08 C/min v 12.5 min. pr. C
00:15 61.1 C
00:30 62.1 C @ 22.1 C delta 0.066 C/min v 15 min/C
01:22 64.3 C @ 22.2 C delta 0.042 C/min v 23.6 min/C
01:26 64.4 C @ 22.3 C delta 0.025 C/min ... ETA 65.4 C
01:31 64.5 C @ delta 0.020 C/min
02:57 64.8 C @ 22.4 C delta 0.0034 C/min v 286 min/C.
03:01 64.9 C @ 22.5 C =)(/=)(/ HD has for some reason powerd down. Peak not reached.
03:27 65.2 C @ delta 0.012 C/min v 86 min/C
03:39 65.3 C @
03:51 65.3 C @
> Concludes : stable at 65 C @ 22.4 C ambient. Closed Cab. PSU shield on. :)
Comment on the anticipatet results : Expect an temperature increase of max 5 degrees C. Depending on how successful the modefication of shield perforation were. Lack measurments for comparing the Open Cab.Load state.
For the open-cab measurment : Seem to recall a open cab measurment of around 59 C,,,, what was this before ? There are no matching measurments. This can only be compared to an measurment of stable idle temperature measured with open cabinet. [ stable at 45 C @ 19.7 ambient (delta 0.0025 C/min) on idle with open cab.]
[ measures & results ]
Update of recorded maximal PSU tempearture :
In conjucntion with fanless testing of the standard CPU Element(3), a higher PSU core temperature was recorded. This was with different Vcore and Vio settings than above, and also with maximal cpu heat generation/powerconsumption.
Prelude : 2 Hours of maximal powerconsumption.
03:58 PSU 66.1 C @ 20.8 C ambient temperature
Measurment Zummary :
PSU stable around 67 C @ 20.8 Ambient tempearture. Maximal CPU power consumption at 1.4
Vcore and 3.5 Vio.
Arising from the testing of the PSU(2/4) in system(2) :
Measurments are indirect, and individual values are based on the total effect used by the system at different states.
All values are approximate, and done using a high quality digital multimeter. Average measurments of steady states.
Measured wall socket voltage : 235 V (used in power consumption calculations)
PSU Netfilter consumes 7-8 mA even when powerswitch is off.
Standby power consumption 27-30 mA
Soundcard(1) draws 10 mA
HD(1) draws 5 mA when spinning idle, and 25 mA when awakening from power-down.
The CD-RW(1) uses 5 mA when idle (no spinning)
The PCI Video-card(2) uses 7 mA
The AGP Video-card(3) uses 10 mA
Total power consumption of system(2) with all components added (HD,CD-RW, SoundCard) and the new AGP videocard is 130 mA under normal use. i.e. everythings on, but fairly idle.
* Add measurments for peak- and minmal load.
[ Index ]