Modifications of the motherboard AsRock 4CoreDual-VSTA, 4coreDual-SATA2…
This article is relevant for the following motherboards:
4coreDual-VSTA
4coreDual-SATA2
775Dual-880Pro
775Dual-VSTA
The article was born with technical consultations and support from comrade R-998.
This article assumes that the reader has a soldering iron and understands the basics of electrical engineering.
The article is written as a note on a napkin to myself.
Photos in this article are of the motherboard 4CoreDual-VSTA (unless otherwise stated) Revisions differ slightly in the location and layout of SMD components.
Real computer experimenters have no time to write articles, shoot videos, teach someone something. That is why information on the topic has to be collected grain by grain, from different sources, overcoming inactive links, studying factory documentation, and, of course, poking around the stand.
Prelude
We are interested in the E5800 processor. Why E5800? Because the board limitation is the FSB bus 300-350 MHz. The E5800 processor is cheap, works normally on FSB200, and is easily overclocked on the bus (up to 4200-4500-5000 MHz, multiplier 16). It all depends on the copy you have in your hands, but 4200 MHz takes almost all the stones. We have access to manipulations with the CPU multiplier from 6 to 16: FSB262*16=4200;FSB280*15=4200; FSB300*14=4200 …
The best results on this board are shown by Extreme X9650/QX9770 processors (large cache, free multiplier), but their market price (to put it mildly) is so-so. These stones with an unlocked multiplier from the factory, which gives flexible manipulations in overclocking. Even multiply the 100 bus as many times as you want, within the frequency limit of the processor itself. Naturally, the 100-200 bus will not suit anyone)), because in addition to the processor frequency, we will be very interested in the performance of the RAM. And all overclocking on the stand AsRock 4CoreDual-…. exclusively on the bus. I hope I explained the idea clearly.
Out of the box motherboard AsRock 4CoreDual-…. does not understand the E5800 processor. And if you stick the E5800 into the motherboard, it will supply 1.0V to the cores at a very low x6 multiplier (which is blocked in the BIOS): the core frequency is 1200 MHz at FSB200.
So out of the box the motherboard can't NOTHING. When buying this motherboard for testing AGP/PCI-E cards or for overclocking, you need to immediately understand that without a soldering iron there is nothing to catch from this board. The board was developed as a universal office solution during the transition from AGP to PCI-E, from DDR to DDR2. The idea was to provide maximum flexibility to enterprise administrators: go to the closet, whatever you find there, put it in the motherboard and it will work FOR OFFICE TASKS.
The motherboard was originally designed NOT for overclocking, and NOT as a stand for video cards. This is why modifications to the board are required to make the motherboard be at least somewhat of a working tool in the hands of a computer enthusiast.
This article does not replace the study of datasheets and other materials on the topic, but summarizes all the accumulated information as briefly as possible (maybe I did not succeed). For a deeper/detailed understanding of the operation of individual board units, search independently in the vastness of the Datasheet and study. The article is written on the principle of “do as I do” and everything will work out. But you need to understand that the Author does not bear any responsibility for your rash actions due to not reading the article to the end, or when understanding the article differently than the Author intended, errors in the article (which the Author could have made unintentionally). All modifications and other actions that you repeat according to this article are at your own risk!
Flashing the motherboard to a modified BIOS
It is advisable to boot into DOS and work in DOS from an HDD (or SSD). Floppy disks are not reliable!
For ease of use in DOS, it is advisable to rename the long name to a shorter one, for example vsta.rom
AFUDOS program for firmware in DOS
I managed to flash it correctly with this version 236U.
Other versions AFUDOS sewed mother crooked: inscription CMOS Checksum Bad ne disappeared, reset to factory settings did not help, ClearCMOS did not help. It was impossible to save the settings in the BIOS!
After loading into DOS, you need to go to the directory where the program and BIOS are located. Going to folders is done with the cd command. For example, going to the vsta folder:
C:\cd vsta
If the program and BIOS are in the root of the C:\ drive, then we immediately write the command with the keys. Command and keys:
afu236u /ibiosName.rom/pbnc/n
If the bios file name vsta.romthen the command with keys will be as follows:
afu236u /ivsta.rom /pbnc /n
Example what the whole line will look like:
C:\vsta\afu236u /ivsta.rom /pbnc /n
After the firmware is completed, the following message will appear: Please restart your computer.
Turn off the computer. Turn off the power supply. Use the ClearCmos jumper to reset to factory settings.
Vcore mode
The essence of the modification: to gain the ability to increase the supply voltage on the processor cores, since the motherboard does not have on-board tools for regulating the CPU voltage.
Usually, during voltmods, the FB (Feed Back) of the PWM controller is interfered with. But on the board 4CoreDual-…. a kind of PWM controller L6714D. This controller has a separate input “OFFSET”, which is intended for adjusting the output voltage. At the factory, this input is connected to the ground via a zero resistor (i.e. deactivated). When connected, instead of a zero resistor, the adjustable resistance is introduced, the so-called positive offset.
We look for pin 32, visually find the zero resistor. We remove the zero resistor and put a 50 kOhm trimmer instead. The trimmer should be preset to the middle position, i.e. to 25 kOhm.
This is what I got:
I took the ground a little higher, on an empty soldering under SMD. I fixed it, as usual, with thermal snot.
We measure the Vcore processor voltage either on the tantalums on the back side of the socket or on the terminals of any capacitor on the CPU power supply, see the photo below.
We install a cold and patient processor like Celeron 420. We wind up 1.5V on it. And only then we install E5800 and adjust it precisely. The best examples overclock >4.5 GHz at a voltage of 1.55-1.65V. But processors are different, your particular example can be both hot and difficult to overclock. Perhaps, you will have to go through N-th number of processors to select a good one (the best of what is available). It is not recommended to increase the voltage above 1.67V under load. Voltage above 1.7V can lead to crystal degradation, it will work, but will no longer overclock, or will heat up like an iron.
ReCap (replacement of capacitors), filtering of PWM power supply of the processor
The purpose of this modification is to reduce pulsations on the processor's output power channels, thereby increasing stability during overclocking.
The manufacturer has prudently left empty mounting units).
All these through-hole and SMD mounting units are the output lines of the four-phase PWM. The rule for selecting capacitors is as follows: the lower the operating voltage of the capacitor, the lower the ESR (equivalent series resistance). The capacity of the jars is limited only by the size of the mounting sites and the dimensions of the capacitor (very high capacitors will not allow you to install many cooling systems). The output voltage of the processor power lines will never be higher than 2V. Therefore, I chose 2.5V 820μF jars. Removed from a donor Abit ip35 pro board. Before soldering, be sure to check the condition of the capacitors on an ESR meter. Used capacitors from motherboards and cards that were heavily overclocked or used for a long time can greatly degrade in parameters.
The task of SMD capacitors is to filter PWM needles. In my understanding, the higher their capacity, the more time for charging/discharging. And we, in theory, need high operating speed. Therefore, you should not get carried away with the capacity of SMD capacitors.
Typical solution is 1-10uF 6V. But I think that 0.1uF 6V is also possible.
What I ended up with:
I didn't stop there, and removed the factory blue capacitors. Replaced them with polymer ones. It's not necessary to do this, but if you want to, no one will stop you, right?
On the back side of the board, under the socket, there are empty mounting places for SMD tantalum capacitors! Let's not miss this opportunity, I installed tantalums 47uF 16V Vishay.
Again, there is no need to strive to install tantalums of maximum capacity, our task is to filter the PWM. And we already have more than enough capacity.
After all these procedures, I managed to reduce the ripple in the load from 400 mV to 100 mV.
Droop mode
The purpose of this modification is to reduce the difference between the processor supply voltage at idle and at 100% load. On the motherboard without intervention, this delta is 50 mV. With the help of modification, it can be reduced to 20-30 mV.
According to the datasheet, the optional DROOP function is activated by shorting the DROOP and FB pins together. According to the factory, FB and DROOP are together (track under the PWM chip). But the control is implemented as for an office board, i.e. very soft. A delta of 0.05 V is very unpleasant, especially at maximum overclocking with the maximum possible supply voltage of the processor.
If you lift pin #21 “DROOP” and ground it, you can get very rigid control. The voltage delta disappears completely, but all power supply units will heat up fiercely. This option is only suitable when cooling the CPU (and VRM, respectively) at sub-zero temperatures. With air cooling, the power supply will definitely fly out.
Therefore, our path is management, so to speak, of medium rigidity. Let's use the author's development.
Author's mod gustep12.
In order to implement the mod, you need to prepare smd resistors 1/2 of the original rating in advance; ceramic capacitors x2 relative to the original.
I wanted to do exactly the same as gustep12but finding 1.5nF=1500pF ceramic capacitors was not an easy task (neither in stores nor on donor boards at my disposal). Therefore, I installed 1000pF SMD capacitors.
nF – nF (nanofarad)
pF – pF (picofarad)
The resistors were found exactly as required. Instead of 2.4 kOhm I put 1 kOhm; instead of 33 kOhm I put 15.5 kOhm; instead of 2 kOhm I put 1 kOhm.
As a result, I got a delta of 0.03V. At the same time, the processor power management system became more rigid. In my case, under load, I began to hear the whistle of the chokes.
Vtt mode
The essence of the mod is to be able to twist the voltage on the north bridge power supply so that the board can more easily go along the bus. This mod is not mandatory, since your motherboard may easily go along the bus without raising the voltage. There is a risk of burning the bridge, so if you do this mod, then do it very carefully. The bridge power supply on this board is common with one of the CPU power supplies. Therefore, there is a general rule: do not twist the Vtt voltage higher than the CPU core power supply (Vtt mode <= Vcore mode).
SMD layout on the 4CoreDual-VSTA motherboard:
SMD layout on 4CoreDual-SATA2 motherboard:
We look for a resistor with the schematic designation R2469 (marking SMD 18C, nominal value 15 kOhm), remove it and put a trimmer instead. We need a trimmer for 20 kOhm. We pre-wind 15 kOhm on the trimmer. I put one leg of the trimmer (which should be pulled to the ground) into the empty mounting unit C2536. It is not necessary to do this, the ground can be pulled from where it is convenient. But I thought that my version is a good idea; and the trimmer itself is well fixed. We solder the second leg to the output of R2469, which is closer to the operational amplifier LM324 (under the schematic designation U3304).
What did we actually do with our interventions? On LM324, we are interested in pins 9 and 10. This is the same comparator inside LM324, which is responsible for stabilizing the voltage on the northerly. We soldered the trimmer to pin 9. This made it possible to correct the comparator by mistake (introduce a voltage offset).
The reference voltage on the comparator (on this board) is 1.25V. Let's calculate the standard adjustment: Vo=1.25*(1+R1/R2)=1.25*(1+1/15)=1.33V. The factory bridge power supply is 1.33V.
By changing the resistance of the trimmer from 15 kOhm downwards, we adjust the output voltage. I would not recommend setting it higher than 1.45 V. Let's calculate the resistance of the trimmer for a voltage of 1.45 V:
R2=1/(1.45/1.25-1)=6.25 kOhm.
Vo=1.25*(1+R1/R2)=1.25*(1+1/6.25)=1.45V.
I do not recommend going below 6.25 kOhm on the trimmer.
The output from Vo (Out 3) goes to control the transistor in linear mode. The transistor is located next to the bridge. The LM321+Transistor bundle forms a linear voltage converter.
We look for a control point near the bridge (on the front side of the motherboard) under the schematic mark VT2. It is there that we take measurements and turn the trimmer.
After the mod, do not forget about cooling the Severnik. Ideally, install a copper cooler with a propeller. But if not, then at least so that the bridge is blown either by a separate fan or a flow from the processor cooling (as an option).
BSEL mode
Chipset straps are the internal timings of the chipset. Tighter straps mean lower timings. Which equals: heavy bus overclocking.
The motherboard chipset straps are too hard when starting on the 200th bus. The essence of the mod is to make the starting bus 266 (while the straps will be softer). To do this, one of the clocker legs must be pulled to the ground. The minimum starting FSB threshold will become 266, and the board will overclock wonderfully.
The clocker is on board our RTL866-890 motherboard. We are interested in the 5th leg of the clocker. This leg needs to be pulled to the ground. But it is inconvenient to solder to the leg itself. So I did it like this:
I took the ground from the empty mounting unit of the SMD component. I pulled the ground to the fifth leg of the clocker through the SMD resistor pin, because it is more convenient.
That's it! Now the minimum motherboard starting bus is 266!
Epilogue
There are still a number of mods. This article describes the most basic ones to get a very good result.
The most advanced ones may find this insufficient, and they go further: for example, they completely abandon the power supplies on board the motherboard and install replacement ones :))
If you came to the understanding that such brutal mods are necessary and started doing this, then you should write the articles, and not read mine).
That's all! Everyone who read the article from beginning to end is a great guy. Everyone who scrolled through is just a great guy. Those who read to the end and successfully repeated the mods – write about it in the comments.
Good luck to all! Bye everyone!
P.S.: I have a YouTube channelmods are flashing there. Including for this board.
