|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
58.152.208.212
In Reply to: RE: cMP - the open source high-end Memory Player posted by cics on December 30, 2007 at 05:42:01
Please refer to the link at the bottom of this page for the background.The following were removed from an experimental GA H55M-D2H MB:
2 x DVI to HDMI Video Convertors
1 x Network chip
1 x Sound chip
2 x MOSFET for CPU and system fans (two legs were cut off, not removed)
1 x Optical ouput
1 X OP amp and cap for optical ouput and
a couple of regulators and a number of transisitors together with all resistors and caps around these chips.Current consumption on the 5V regulator measured 1.73A in CMP and 1.8A @96K. Removal of the above chips resulted in a decrease in current consumption of 0.5-0.6A.
Most important of all, removal of the two video chips brought about significant improvement in SQ, which sound somewhat like changing from ATX PS to Regulated PS. This is probably due to removal of two noisy video chips.
The experiment confirmed that although some chips remained unuse or are turned off in BIOS, they still draw current and can significantly impact on SQ.
Unfortunately, there is no easy way to remove these chips. A butane gas soldering iron (with hot air blow) or a hot air gun can be used for the desoldering job. Please note that the back of the video chips and network chip are soldered to the ground plane of the MB and their removal are much more difficult.The above modification is meant for some die-hard DIYER like me, and is not recommended for normal users. For fun and information only.
Edits: 06/05/10 06/05/10 06/05/10 06/05/10Follow Ups:
A further report : the IDE controller, a regulator and 2 caps were removed from the MB. I was very cautious in taking such action because the BIOS chips are next to them. After checking they have no connection to the BIOS chips, I took them off one by one, firstly the IDE chip, then the caps and lastly the regulator.This time I remove the IDE chip without using any soldering tool. I used a cutter to cut off the small pins around the chip. It was not soldered to the ground plane and the job was easy. The regulator was removed using two soldering irons.
The computer is working fine after such removal.
The IDE chip needs two power sources, one directly from the 3.3V and the other from the regulator which is also powered by the same 3.3V, the two caps are for the input and output of the regulator.I measure the 3.3V current which is 556ma. I don't know whether the reduction is solely due the above removal or as a result of all removals because I didn't measured the 3.3V current during my last attempt.
Take a closer look at what can be removed from the H55M D2H MB.
Edits: 06/06/10 06/06/10 06/06/10 06/06/10 06/06/10 06/06/10 06/06/10 06/06/10 06/06/10 06/06/10 06/06/10 06/06/10
Firstly well done on this finding. Reducing load on 5V line will yield most SQ improvement. I would guess cutting 12V fan supplies would yield subtle improvements at best. But those nasty video chips run on 5V...
The wild idea is to disable those chips (and other unused stuff) by finding the 5V pins on P24 connector. I started with switching from P24 to P20 removing 4 pins. This leaves us with fewer 5V, 12V & 3.3V lines to mobo. Is there merit in trying to selectively disable remaining 5V pins (5 including standby on P20)? Also, we could do the same on 3.3V pins (3 on P20) and the sole remaining 12V pin.
You unplugged the 4 pin connector on the end of the p24 connection so you in effect have only a p20 connection on your mobo from your main pc power supply. Do I understand correctly? If that is correct that is a very elegant way of disabling more stuff on the mobo (less current/less emi/less power draw).Also do I understand correctly that you tried this, had no problems booting up, and got a resultant sq improvement? If so that is an easy one to implement.
One further thought rather than pulling pins on the mobo p24 connector how about covering un-needed pins with a plastic sleeve or electrical tape to eliminate?
Edits: 06/06/10 06/06/10
If that is correct, that is a very elegant way of disabling more stuff on the moboAh, but it's not correct, at least for every mobo I've looked at.
The various leads for each voltage are joined together on the board. It follows that the only effect of disconnecting some of (say) the 12-volt lines is to increase the load on the others.
The aim of going from P20 to P24 as the ATX standard was to increase the overall power capacity: you can still use a PSU with a P20 plug to power a P24 socketed mobo without disabling this or that. I wouldn't generally recommend it but IME it's possible.
As "proof", our neighbour is happily running her computer on a six-year-old P20 PSU I've lent her while I sort out a warranty claim on her P24 OCZ unit. Apart from the noisy fan, last I heard, it's fine.
HTH
Dave
Edits: 06/06/10
Ryelands is right. All leads of the same voltage are internally connected inside the MB, same for the ATX PS. Pin 1,2,12,13 for 3,3V, Pin 4,6,21,22,23 for 5V and Pin 10,11 for 12V. The multiple pins or wires are designed to share large current. Hence, the wild idea won't work. Connecting or disconneting the last 4 Pins in P24 will not make any difference to current consumption at all.
Sadly, I think you both right. Only exception would be pin 9 (5V standby) which would have separate traces. I know some (most?) mobos use this for USB ports and perhaps in other areas. Would be interesting to try.
Anyway of having a custom H55 mini-ITX mobo built? Depending on cost, there may be enough sponsors to get this going.
A Work Around to desoldering is to cut off the power supply lines next to the chips. I can check out where the voltage is supplied to the chips. A magnifier will be required for such work as the traces cannot be seen by naked eyes.
I want to avoid loading Intel's graphics driver and am wondering if those video chips power down when the display is off. Can you test for this?
Under Device Manager, make sure "show hidden devices" is checked. You can disable these 2 under "Non-Plug and Play Drivers":
- 1394... - this is the firewire device and should be disabled when not using it.
- gdrv
Disabling 1394 device yielded improvements:-)
Set gdrv off in device manager but there was no change in current consumption. 1394 was not present. Tried enabled in BIOS caused BSOD....I made a mistake in my previous current measurement, there was acutally no current reduction on the 5V line. I was trying to do two tests at the same time - I put a battery as buffer in the wrong position causing erronous reading.
The following pix help understand the actual current consumption from the ATX P24.
GA H55M-UD2H defalut BIOS (9x133) (device on, no chip was removed)
-
-
-
-
-
GA H55M-UD2H BIOS optimised (9x100)in CMP
-
-
-
-
GA H55M-UD2H BIOS optimised CMP@96K
-
-
-
-
-
GA H55M-(D)2H BIOS optimised CMP@96K (2 video, 1 sound, 1 network, 1 IDE chip, 2 regulators were removed)
Now it is clear that removal of the chips will reduce 12V and 3.3V current consumption.Further measurement confirmed that the 2 DVI-HDMI convertors are actually driven by the 3.3V directly from the ATX P24(I thought it would be easy to cut off the supply by desoldering the small chokes which turn out to be capacitors), the sound chip via a regulator from 12V direct and the network chip also by 3.3V by not directly from the ATX P24.
Edits: 06/13/10 06/13/10 06/13/10 06/13/10 06/13/10 06/13/10 06/13/10 06/13/10 06/13/10 06/13/10 06/13/10 06/13/10 06/13/10 06/13/10 06/13/10 06/13/10 06/13/10 06/13/10 06/13/10 06/13/10
Thanks, great to see how optimisations reduce power across different voltage lines. To summarise:
- default = (12v@0.32A, 5V@3.88A, 3.3V@0.860A) = 26.1 watts
- GA H55M-UD2H BIOS optimised (9x100)in CMP = (12v@0.28A, 5V@2.027A, 3.3V@0.698A) = 15.8 watts
- same as 2 at 96k = (12v@0.28A, 5V@2.241A, 3.3V@0.714A) = 16.9 watts
- same as 2 at 96k (mobo modified) = (12v@0.18A, 5V@2.226A, 3.3V@0.430A) = 14.7 watts
Interesting to see 5V drop is low compared to 12V & 3.3V lines when doing mobo mods. From your extreme mobo modification post, 5V line drives: RAM, H55 chipset, QPI Link, CPU PLL & ITE/Fan/etc.. Gcore and video chips take power from 12V & 3.3V lines. 5V line is most critical supply. Could a simpler mod be done that just provides clean 5V line from linear PSU or battery?
I was wanting to see if video chips on mobo are affected when the actual LCD display monitor is off. Disabling "gdrv" will not switch off display - its an unneeded device driver in Windows. I installed Windows with default BIOS hence I got to see "1394...". In your case enabling it after installation and in cMP mode will cause BSOD.
As far as possible linear supply in place of PWM PS is concerned, the 5V line is no longer critical. The concerns are large current supplies for the QPI (@1.1V) and possibly the CPU(@0.7V). At such low voltage, normal 3 terminal regulator circuit cannot work. A work around is to use a negative voltage to generate low voltage.
Heat disspation is also a concern which is assocaited with the raw DC voltage. A 6V battery supply or 5V power source would be to high as the regulator would dissipate 13W and 12W of heat respectively - a waste of energy and case temp wiould rise. A Low dropout regulator such as LT108X together with a low raw DC voltage, say 3V(or 3.3V Lithium Battery packs), would appear optimial.
I will build two low voltage regulators to test the QPI and CPU.
Your current approach is superior. I'm looking for an easier way to get say 80% of benefits. Critical chain is:
RAM > QPI link > H55 chipset > PCI or USB
These components all share 5V line. What I'm thinking of is a 5V replacement supply at P24 connector. Possible?
Please elaborate more on your idea of a 5V replacement supply at P24 connector. Do you mean replacing the ATX 5V with a linear regulated 5V supply?
I'm think of an idea to use the P24 3.3V to feed a QPI linear regulator, presuming that an ATX PS is to be used. The 3.3V line has very light current loading. By doing so, the P24 5V will be greatly relieved and the interference between the QPI load and Ram etc would be greatly reduced.
Edits: 06/14/10
Do you mean replacing the ATX 5V with a linear regulated 5V supply?
Yes. This would power the main audio streaming/signaling chain. Other components including CPU have a secondary role.
Using 3.3V line for QPI is very interesting and would reduce 5V load significantly.
Hi Jack!
What is the current consumption of the CPU (socket P4) at the last configuration (GA H55M-(D)2H BIOS optimised CMP@96K (2 video, 1 sound, 1 network, 1 IDE chip, 2 regulators were removed - I don't see the pci audio card ??? ))?
Hi, No P4 12V measurement was taken last time, but it should be 0.46A - 0.18A = 0.28A.
For 0.46A ref. http://www.audioasylum.com/forums/pcaudio/messages/7/74498.html The P4 and P24 12V lines were combined for measurement.
This time P24 12V alone measured 0.18A.
Thanks Jack! Continue the experiments.It is very useful.
Regarding the sound card, I was using a USB DAC.
OK. Will do. By the way, I have identified suitable locations to terminate the power source to the chips. Simple soldering will be required but only one SMD device will need to be desoldered per chip in order to shut down the power. I will post more details soon.
Another option would be to load graphics driver to see if there's an option to disable multiple outputs. It seems mobo outputs to all interfaces as default.
ok Jack and Dave thanks. I guess I won't try it then.
That is outstanding Jack!! Way to go!!!!!
I am just wondering if there are any similar opportunities on the GA-31M-S2 family of mobo's.
Also this really confirms in spades cics theory of minimizing/eliminating power draw on the mobo.
Since the GA-31M-S2 family of mobos do not appear to have any DVI or HDMI output there is no video convertor on the MB. The VGA is powered internally by the Intel GMA 3100 inside the North Bridge. The sound and network chips can be removed but the SQ improvement is slight.
Thank you again Jack.
Great, Jack! Keep on trucking!
Post a Followup:
FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: