Overclock.net › Forums › Volt Mods › General Volt Mods › Motherboard Volt Mods › Asrock 525 undervolting (and other Atom Platforms)
New Posts  All Forums:Forum Nav:

Asrock 525 undervolting (and other Atom Platforms)

post #1 of 4
Thread Starter 
Hello, this is my first thread here.
I bought a 525 Board and was not really satisfied with the overall power consumption:
I got 2x1GB RAM, and one Western Digital Eco HDD and one PCI Soundcard (RME DIGI 96). The Mainboard is powered by a pico PSU, which transforms 12V to 3,3 and 5V. The whole system needed 32-36W out of the plug.
Thats only 10W less what my old P III do consume, and 10W less than a Core 2 Duo in an Dell Optiplex 745.
The Problem is, that the MB gives several overclocking and overvolting features but no undervolting ability!
So i searched for the voltage Regulator Manual in the WWW

The Power Regulation of the 525:

There are two APW 7120 onboard. One is for the CPU 1,1V Voltage, the other is for the 1,8V DDR Voltage.
The 1,5V and 1,05V is made with a linear regulator made of a quad opamp and a couple of fets. Those linear regulators are fed from the DDR Voltage. I can't garantue this, but it seems logical for me.
Lowering the DDR voltage will also lower the power destroyed in those regulators.
The Datasheet of the APW 7120 is aviable here:
APW7120 datasheet pdf datenblatt - Anpec Electronics Coropration - 5V to 12V Supply Voltage, 8-PIN, Synchronous Buck PWM Controller ::: ALLDATASHEET :::

Modifications:

As you can see in the Datashet, the voltage is defined by a resistor feetback network. I did not manage to understand the whole schematics on the MB, but i noticed that it's really easy to modify the feedback.
If you increase the feedback with a smaller resistor R1 (see schematics pdf) you get a lower voltage.
So i soldered a Poti from the VT1 Pin to Pin 6 of the AWP 7120.Adjusting it to around 50 Ohm yields a V-core of 0,85V-0,87V. And the CPU runs stable at 1,8GHZ! So i changed the pot to a 57Ohm Resistor.
The Bad thing is, that the overvolting function is less intense.
The second thing is, that i did some modifications of the DDR voltage. I use my old overclocking RAM (ADATA Extreme Edition), which is able to run at 1GHZ. Here I did the same thing than to the cpu voltage: put a pot between VT2 and Pin 6. This time a resistance around 1,5k gives 1,5V DDR Voltage. I did not have a 2K Resistor here, so i let the RAM run at 600Mhz and it's rock stable.

Ok, is it worth using the soldier gun? Yes it is!
The whole PC needs 22W !
As you can read in the Atom Datasheet, the VID goes from 0,8V to 1,2V. I think Asrock did not use the VID and let every CPU run at 1,1V. So there is a good chance to get a fine undervolting result.

I don't whant to write a "exact" diy manual, because i don't whant to risk that someone who is not familiar with hardware modding tries this.

post #2 of 4
Impressive work.
FreeNAS
(9 items)
 
Beast
(14 items)
 
Alura :)
(11 items)
 
CPUMotherboardGraphicsRAM
Intel Xeon E3-1245 v5 SuperMicro X11SSM-F-O C236 Integrated 2 x 16GB Kingston ValueRAM ECC KVR21E15D8/16 
Hard DriveCoolingOSPower
5 x 8TB Seagate IronWolf Pro in ZF2 Array CoolerMaster Hyper 212 Evo FreeNAS 11 EVGA SuperNOVA G2 550W Gold 
Case
Fractal Design Node 804 
CPUMotherboardGraphicsRAM
i7 4790k Asus Z97 Pro WIFI-AC Sapphire Nitro Radeon R9 Fury Tri-X OC+  G.Skill 4 x 8GB F3-2133C9D-16GXH 
Hard DriveOptical DriveCoolingOS
2 x Intel 730 480GB SSD in RAID 0 Asus DRW-24B1ST/BLK/B/AS Noctua NH-D15 Winblows 10 Pro x64 
MonitorKeyboardPowerCase
Pixio PX277 Logitech G510 Seasonic SS-660XP2 Fractal Design R4 Blackout 
MouseMouse Pad
Logitech G500 Xtrac Ripper 
CPUMotherboardGraphicsRAM
i5 2500k Asus P8Z68-V Pro MSI R7970 Lightning BE 2 x 4GB G.Skill Ripjaws 2133mhz 
Hard DriveOptical DriveCoolingOS
Mushkin Enhanced Chronos 240GB and WD Black 1TB Asus DRW-24B1ST Cooler Master Hyper 212 EVO Windows 7 Ultimate x64 
MonitorPowerCase
Hannspree 24 inch Thermaltake Tough Power TP-750P 750W Gold Cooler Master HAF 932 
  hide details  
Reply
FreeNAS
(9 items)
 
Beast
(14 items)
 
Alura :)
(11 items)
 
CPUMotherboardGraphicsRAM
Intel Xeon E3-1245 v5 SuperMicro X11SSM-F-O C236 Integrated 2 x 16GB Kingston ValueRAM ECC KVR21E15D8/16 
Hard DriveCoolingOSPower
5 x 8TB Seagate IronWolf Pro in ZF2 Array CoolerMaster Hyper 212 Evo FreeNAS 11 EVGA SuperNOVA G2 550W Gold 
Case
Fractal Design Node 804 
CPUMotherboardGraphicsRAM
i7 4790k Asus Z97 Pro WIFI-AC Sapphire Nitro Radeon R9 Fury Tri-X OC+  G.Skill 4 x 8GB F3-2133C9D-16GXH 
Hard DriveOptical DriveCoolingOS
2 x Intel 730 480GB SSD in RAID 0 Asus DRW-24B1ST/BLK/B/AS Noctua NH-D15 Winblows 10 Pro x64 
MonitorKeyboardPowerCase
Pixio PX277 Logitech G510 Seasonic SS-660XP2 Fractal Design R4 Blackout 
MouseMouse Pad
Logitech G500 Xtrac Ripper 
CPUMotherboardGraphicsRAM
i5 2500k Asus P8Z68-V Pro MSI R7970 Lightning BE 2 x 4GB G.Skill Ripjaws 2133mhz 
Hard DriveOptical DriveCoolingOS
Mushkin Enhanced Chronos 240GB and WD Black 1TB Asus DRW-24B1ST Cooler Master Hyper 212 EVO Windows 7 Ultimate x64 
MonitorPowerCase
Hannspree 24 inch Thermaltake Tough Power TP-750P 750W Gold Cooler Master HAF 932 
  hide details  
Reply
post #3 of 4
Thread Starter 
I mesured the DDR RAM voltage again:
It was only 1,38V.
And the 1,5V is nearly the same. So my idea, thats a linear regulator is right.
I tried lowering the FSB to 167Mhz (1,5Ghz CPU clock) and lowering the DDR Voltage to. At 1,25V is the minimum. This is caused by the chipset, not the RAM.
I tried to bypass the linear Regulator (to get 22mV more Chipset voltage) but then the board did not start anymore.

The second thing is, that i lowered the Vcore to 0,83V no problem, even at 1,8GHZ.

The overall Improvement is little, its 1W for much more work. Lowering the 1,05 V Line is work intensive and there is not a big amount of power to save (overall consumption is about 2W).

At 100% Cpu load with prime, i got 23W now. This is an improvement!

Now i checked the efficiency of the cheap 12V Supply. The Machine needs around 1,5A at 12V. Its around 18W. So the Supply cost's 3W of power. It's 86% and there is no way to improve.
Edited by Revoxi - 7/26/11 at 11:32am
post #4 of 4
adding larger capacitors may increase the voltage some... if the circuit is clipping voltage faster than the little cap's can supply, you stand to gain another .02-.2v just from adding larger caps. You can monitor input and output to see which is drooping more, and cap there first.

*nice work on the mods too
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Motherboard Volt Mods
Overclock.net › Forums › Volt Mods › General Volt Mods › Motherboard Volt Mods › Asrock 525 undervolting (and other Atom Platforms)