Overclock.net banner
661 - 680 of 744 Posts

·
Registered
Joined
·
209 Posts
More or less the same silicon however it is a similar deal as with e.g. CPU or GPU variants: The same silicon however, the configuration made with fuses (which are immutable) differ.
So I'm guessing we need to somehow "bootstrap" the firmware to always think it is "operating" on the right parameters. Maybe either by firmware mod or something like a console modchip (with the BIOS chip).
 

·
Banned
Joined
·
1,185 Posts
More or less the same silicon however it is a similar deal as with e.g. CPU or GPU variants: The same silicon however, the configuration made with fuses (which are immutable) differ.

Practically, the only real option would probably be using a very early AGESA version based bios, one of the first that support the chip.
This is what i was planning to do with my X370-F.

I noticed that the size of the initial X470-F bios, are the same as the X370-F ones, 16Mb.
So i could try to flash them, but they run an older version of the AGESA, than the one currently available for the X370-F.

I ordered and received a couple of WSON Winbond 32Meg 25Q256JWEQ, i thought to program it with a X470-F bios, just a bit older than the one i currently run on the X370-F.
Lift the VCC pin of the current 16Meg chip and solder on top the 32Meg bios, with a little switch for the VCC rail to be able to choose the chip, we will see how it goes.
 

·
Registered
Joined
·
209 Posts
This is what i was planning to do with my X370-F.

I noticed that the size of the initial X470-F bios, are the same as the X370-F ones, 16Mb.
So i could try to flash them, but they run an older version of the AGESA, than the one currently available for the X370-F.

I ordered and received a couple of WSON Winbond 32Meg 25Q256JWEQ, i thought to program it with a X470-F bios, just a bit older than the one i currently run on the X370-F.
Lift the VCC pin of the current 16Meg chip and solder on top the 32Meg bios, with a little switch for the VCC rail to be able to choose the chip, we will see how it goes.
I think you could do the same if you do something with the CS (Chip Select)/Slave Select pin.
 

·
Banned
Joined
·
1,185 Posts
I think you could do the same if you do something with the CS (Chip Select)/Slave Select pin.
I have downloaded some laptops schematics, implementing the CS chips select.
I think one need an additional chip to be able to implement the chips select line.

But looking at the chip select schematics circuit, both SPI roms are wired in parallel.
So i suppose that the SPI signal should not degrade too much, if i just pigtail the new rom on top of the older one.
And feed the VCC rail to the rom i would like to power.

It avoids me unsoldering the old one, leaving the motherboard in its original state, if i wish to sell it.
 

·
Registered
Joined
·
209 Posts
I have downloaded some laptops schematics, implementing the CS chips select.
I think one need an additional chip to be able to implement the chips select line.

But looking at the chip select schematics circuit, both SPI roms are wired in parallel.
So i suppose that the SPI signal should not degrade too much, if i just pigtail the new rom on top of the older one.
And feed the VCC rail to the rom i would like to power.

It avoids me unsoldering the old one, leaving the motherboard in its original state, if i wish to sell it.
Right.

How did the testing went?
 

·
Banned
Joined
·
1,185 Posts
Right.

How did the testing went?
I will start the testing this weekend.

I was in holidays and i spent some time cleaning and rebuilding my custom AIO watercooling.
I also finished lapping both cpu and AIO block, but i lost all the pictures i took of the assembly progress.

Unfortunately, the bios chips arrived a couple of day later, when i everything was already assembled.
My friend wanted to play BFV, got it for 1.5euro, so i put the bios and soldering aside, to be able to play with him.

The rig, H100 + ID 240vga + Orcus pump.
2522292


The bios chips, Winbond 25Q256JWEQ
Sensitive content, not recommended for those under 18
Show Content
2522293



I think this weekend i will start by flashing the initials revisions of the X470 bios, that were also 16Mb.
And i will prepare the new 32Mb chips with the same bios i'm using currently, but the 32Meg X470 version.
 

·
Banned
Joined
·
1,185 Posts
So, little follow up.

I started by building a new bios module, enclosing the chip into slow curing epoxy resin.
I used the SATA cable wiring as electrical connections, single core copper wire, to connect the bios chip.
Nb, the chips is a WinBond 25Q256JWEQ 1.8v 256Mb, WSON 8x6mm package.
Sorry for the potato pictures quality.
2523100

Dropped it into a crudely shaped mould and poured epoxy on top.
2523102

Here the result one day later, i planned to reuse the built-in motherboard programming header, hence i started to also disassemble an USB connector.
More on that later.
2523104

I then proceed to check if i could read this abomination, but unfortunately, the chip remained silent.
I tried all the Chinese versions of the CH341A software, tried with compiled windows10 version of Flashrom, nothing.
I grabbed an old laptop motherboard from my pile of junk/spare parts, the bios chip came up, recognized right away.
2523107

At this point, i was worried that the seller provided fake chips, even if these feel quite OG WinBond.
I threw the second chip onto the CH341A adapter PCB, crudely soldered it and proceeded by reading the chip.
This time, it got right away recognized, tho partially, the programmer sees the chip, but can't tell the brand, capacity, etc.
Note the burnt marks and the solder mask on the pcb, top-notch quality here, i barely brushed it with my iron.
2523114

Wondering why my moulded beauty didn't work, maybe the stress of the epoxy curing, alongside solid copper wires, damaged the thin connections.
Or i broke the soldering joints, when pouring the thing into the mould, dunno, but i was happy to have at least one working chip.
Still, i recovered the first chip and gave it a second chance, this time with less epoxy and stranded copper leads.
2523113

The second chance monstrosity, now reads fine, still not as strong i wanted, but good enough.
I have now two wired 32Meg bios chips, ready to be programmed.
2523116
 

·
Banned
Joined
·
1,185 Posts
I proceeded by erasing, blanking and writing the chip, with the various versions of the Chinese software.
Some couldn't read the chip properly, but most of them detected it, at least partially.
I thought i managed to program into, the Asus X470-F 5406 bios.
2523117

When i load the bios, i get this message.
I noticed that the Cap_to_Bin or UEFI Tool extracted image, are bigger than 32Mb.
And sure thing, the end of the blob, that must be written, is missing!
2523118

I will try again with Flashrom, because today, i read the chip again, but no avail, it is blank.
Dunno, maybe these aren't good, usually the software detected the old one pretty easily.
I flashed the own X370-F motherboard 1.8v SOIC bios, without issues.
And this is what i get, with an old version of Flashrom.
2523121

I will report how it goes, if this one isn't working, i will grab and try to compile the latest version, on my Debian laptop.
And, just for the sake of completeness, i tested the second chip, and sure enough, same flakiness.
 

·
Banned
Joined
·
1,185 Posts
So...

I kept working on the flash procedure for the WinBond 25Q256JWEQ 1.8v 256Mb, 32Meg chip.

I think i killed and/or fried one of the two chip, the one soldered to the CH341A adapter.
Initially, i thought it was working fine, but when i try to detect it more than once, the programmer get stuck on "IC not responding".
I reworked the joints, tried again, same thing, until it stopped being detected at all.
I soldered back to the adapter my working moulded beauty and gave the broken one a last chance, by soldering some leads onto.

As far as the software goes, i found some prebuild CH341A software packages, on the WinRaid Forum, uploaded by the user Lost_N_Bios.
Here is a screen of the software tools, one can find around, for the CH341A programmer.
2523995

It is true that one of many of these prebuild .rar packages, triggered Windows Defender, but the tools themselves are not.
I tried all of them, most of them detect the chip, can read/write, but only few are able to properly read the 32Meg address space.
Luckily, after a lot of time spent, i ended up finding a couple of them that seem to work with these new 256Mbit bios chips.
2524000

The last version of NeoProgrammer works the best, great tool, detect the chip right away, as belonging to the W25Q256JW 1.8v chip family, registers included.
Colibri seems to work as well, even if it does not recognize the chip, but can read the registers too.
The CH341A v1.1.1.32 too, detect the chip, read the registers and seems to support 256Mbit bios.
The registers description can be found in the datasheet attached below, here the S09 read-only register is flagged.
The CH341A does not support Quad SPI, luckily the flash can still operate in standard mode too.
2524016

I then proceeded by erasing, blanking and reading the chip with NeoProgrammer, before trying flashing, once again, the X470-F 5604 bios.
2524001

But unfortunately, the bios image size issues, arose again, the extracted UEFI .bin image is bigger than the chip size.
It is the first time i got this error, i don't know if the extraction procedure changed, but i remember that the extracted UEFI Tools .bin image, can be directly flashed into the chip.
2524004

I got the same error with all the 32Meg X470-F bios, very frustrated, i decided to flash the latest X570-F bios instead.
And sure enough, finally this time, the flash was successful, no errors given when loading the X570-F bios.
I read back the chip, compared it with the .bin hex file, and sure enough, the data is there at the right address.
Colibri seems to read the bios fine, but the CH431A v1.1.1.32 doesn't seem to work well, the data is all over the place.
2524003

To be sure, i disconnected the CH341A programmer and plugged it back into the motherboard rear usb port, instead of being plugged into the front usb panel.
I read the chip again, with NeoProgrammer and also with AsProgrammer, the latter doesn't recognize the chip, but can read, if the address range is provided beforehand.
2524010

OP SUCCESS!




TL:DDR
It seems that there is a way to read and write these new 256Mbit/32Mb 1.8v WinBond bios chips, using the CH341A programmer.
NeoProgrammer seems to be the way to go software, for the CH341A programmer.




Tho, can the bios gurus explain me, why i would flash the X570-F bios without issues, but i couldn't flash the X470-F.
And sure enough, the Cap_to_Bin X470-F image, is bigger than either the 33554432bytes of bios chip and the X570-F .bin image.
2524012

There are 2048 additional bytes added to the X470-F extracted .bin bios, ideas??!!!!
 

Attachments

·
Registered
Joined
·
1 Posts
With hex editor remove first 1000h for 32Mb, (800h for 16Mb roms). After that your 5406 BIOS file should start with 55 aa 7f e9 29 02 00 00 00 00 00 00 00 00 00 00 and its size should be 33554432 bytes, open it with FD44Editor and add your unique MAC, UUID and S/N in respective fields and then save the result and flash it.
You can download newer CH341A software here:
 

·
Banned
Joined
·
1,185 Posts
So...


With hex editor remove first 1000h for 32Mb, (800h for 16Mb roms). After that your 5406 BIOS file should start with 55 aa 7f e9 29 02 00 00 00 00 00 00 00 00 00 00 and its size should be 33554432 bytes, open it with FD44Editor and add your unique MAC, UUID and S/N in respective fields and then save the result and flash it.
You can download newer CH341A software here:
Thank you a lot for the tip, removing the first 1000h from the rom, with a Hex editor, did the trick.
Now, the .bin file fit the flash size, i was able to flash the X470-F 5604 bios successfully into the bios chip.

Before reading your message, i started walking through the bios hex file, before starting checking the modules structure with UEFI Tool.
When reading the hex file, there is some plain text describing the bios image composition, as you were suggesting.
Rectangle Screenshot Font Software Parallel

Strangely, both the X370 and X570 rom images start with 55 aa 7f e9 29 02 00 00 00 00 00 00 00 00 00 00, after being edited with the Cap_to_Bin tool.
When reading the X470-F hex file, i noticed that the CAP data are repeated twice instead, at the beginning of the file.
And sure thing, if i run twice Cap_to_Bin, on the same output .bin file, i end up with the right file size.
Font Material property Screenshot Parallel Number

At this point i have mostly completed the project, the last part involves connecting the newly flashed bios, to the motherboard.
I have planned to reuse the programming header, that is widely available on nowadays motherboards.
The programming header is wired in parallel with the bios chip, each manufacturer have its own header pinout.
Photograph Motor vehicle Hood Light Automotive tire

Remember the USB front panel connector, mentioned in my post above, i planned to reuse the individual dupont connectors pin, to hook up onto the programming header.
The USB and Audio female connectors can be disassembled easily, by lifting the small plastic tab, meanwhile pulling the lead, taking out a single female pin connector.
I soldered some leads and potted the second chip i was thinking to be faulty, it turned out working perfectly well, even after being abused.
The CH341A pcb adapter quality is awful, i noticed that a pad lifted when removing the bios chip from the pcb, and another one broke not so long after.
I proceeded by soldering the single USB dupont connectors to my third moulded beauty, both moulded chip are working, but this one is definitively stronger.
Circuit component Passive circuit component Gesture Electronic component Jewellery

Proof of concept, once plugged onto the CH341A adapter pcb, ready to be programmed.
Toy Gas Electric blue Plastic Household supply

Here we are, everything is ready to be mounted onto the motherboard, as removable 32Meg bios chip!!
I can disable the motherboard resident bios chip, by cutting the VCC pin, with a very sharp knife.
When working on a bios chip soldered onto a motherboard, always take the power cord off and the CMOS battery out, beforehand.
If cut carefully where the pin bend, the pin can be easily soldered back, reverting the motherboard bios, to its original state.
Once the on-board bios disabled, i can hook up the newly made removable bios chip, to the motherboard bios programming header.
In this way, the system should boot from the newly installed bios, instead of the old on-board soldered chip.




I will start testing starting with the X470-F 5604 bios, i'm currently running the 5406 on my X370-F and i will report back the results.

Nb.i, i can't add my MAC address and MB info, the save button stay greyed out.
I also took the time to save my current bios with Afuefi64, same thing, no GbE data or version saved, the fields are blank.

Nb.ii, everything shown in this academic tutorial, has been realized with basic cheap tools and parts, commonly available almost everywhere.

Nb.iii, @TMS now the latest CH341A 1.43v version is able to read the chip, the 1.1.1.32v was not.
The CH341A 2.2.1.0v comes from NeoProgrammer, i suppose.
 

·
Banned
Joined
·
1,185 Posts
So...


I spent a lot of time flashing and booting different bios, without many results, unfortunately.

Instead of cutting the VCC pin, i managed to unsolder and lift the pin from the pad.
The rig is still fine, i boot onto the on-board bios, by jumpstarting the VCC pad against the lifted VCC pin, not sure how much the latter will last.

Here are the bios i tried:
-X370-F 5606 (16Meg) => Nothing.
-X470-F 5007 => CPU led light up, memory led light up, get stuck with both leds on.
-X470-F5406 => CPU led light up, memory led light up, get stuck with both leds on.
-X470-F 5809 => Nothing.
-X570-F 2802 => Nothing.

Dunno if the board is actually capable of taking a 32Meg bios chip, i mean, the 16Meg and 32Meg chip are quite similar, but the package.

Dunno if plugging the new bios into the header, when one is already soldered onto the board, is an issue.
Knowing how dual bios circuits are implemented, having a silent chip, should not bring down the SPI communication.

When i got the CPU+RAM led light up, i tried various ram slots combinations, i cleared the CMOS, unplugged the gpu, but nothing allowed the cpu to boot further.

Also, dunno why the OG X370-F bios didn't booted the cpu, i erased the chip before flashing it.
The bios blob begin and end where it should, the rest of the chip is blank FF.

Ideas??!
 

·
Premium Member
Joined
·
2,830 Posts
Discussion Starter · #674 ·
I'm not sure why the additional hassle, following should work as is, unless the piggyback flash configuration is causing problems.

Download the newest, correct bios file for the board.
Open it with a hex editor, delete the first 2048 bytes (800h, that is the capsule signature), go to the end of the file and add 16777216 bytes (1000000h) which will bring the bios to the correct size for a 256Mb (32MB flash).
 

·
Banned
Joined
·
1,185 Posts
So...


I'm not sure why the additional hassle, following should work as is, unless the piggyback flash configuration is causing problems.

Download the newest, correct bios file for the board.
Open it with a hex editor, delete the first 2048 bytes (800h, that is the capsule signature), go to the end of the file and add 16777216 bytes (1000000h) which will bring the bios to the correct size for a 256Mb (32MB flash).
I tried twice the latest X370-F bios, once writting on top of a blanked chip, the next 1000000h after the UEFI image were FF.
The second try, adding 00 to fill the rom with 1000000h, after the UEFI image.
No avail, the cpu fan don't spin, just slightly barely move, no debugging leds, no boot.

It is quite time-consuming, reading, erasing, blanking and programming the chip take at least 10min, 4.30min for programming alone.
Then, i was not sure to get the best connection, unfortunately i realized a bit late, that the motherboard programming header, is slightly smaller than the USB header.
Plugging the modded bios chip and verifying that there are no shorts, also take quite some time.
Electronic instrument Audio equipment Gadget Computer hardware Electrical wiring

So i have decided to solder directly an extension header to the programming header, for easier chip reach and swap.
In retrospect, i should have simply modded a USB female to female header cable, removing each single dupont connectors and plugging them back in the right pin order.
I also added a small switch, that goes out the read case, to be able to power the on-board chip, when i want to boot my main bios, the chip leg survived till now!
Computer Computer cooling Personal computer Circuit component Computer network

Still no avail, the motherboard refuse to boot with the OG X370-F bios, it powers up only with the X470-F 4804/5007/5406(sometimes) bios, latter version or X570-F bios are not able to trigger up the board.
This is what i got, when it finally wakes up, with the X470-F 5007 bios, the cpu fans spin, the cpu led light up, then the ram led turn on, then nothing else.
Circuit component Light Blue Passive circuit component Electrical wiring

At this point, i tried to fiddle with both bios altogether, when powering both bios at startup, the motherboard would not boot, same as with the other roms.
I tried to remove the add-on bios pins one by one, checking when the on-board chip, would be able to boot the board.
And sure enough, the motherboard don't like this configuration.
The add-on bios 5007 can turn on the board with the VCC, GND, DO, DI, CLK, CS pins connected, the RESET, WP pin are not essential.
I thought that the motherboard would not been bothered by a silent chip, still it seems to have an effect on the chip detection.
When implementing a dual chip solution, the chip are wired in parallel, but both chips are powered up and pulled up or down through the chip select, hold/resets pins.

I'm still really puzzled, why some bios can trigger the motherboard to boot, and others didn't.
The newer AMD/ASUS 32Meg bios, are nothing more than two stacked 16Meg images, each one having different cpu support.
I'm wondering if when the motherboard boot, it boots the first 16Meg image, where the is no Matisse support, instead of the second one including Matisse support.
Hence, getting stuck with the CPU + RAM leds on, if we assume that in this case, the motherboard successfully load the bios and was able to successfully initialize the cpu.

I have now wired both 32Meg bios chips and can try 2 bios at the time.
Electrical wiring Electronic engineering Networking cables Cable Pliers

I suppose that the next step, would be to unsolder the on-board chip completely, but being my main test rig, i will follow this path when it will finally retire.
I found nobody that tried this mod, a lot of time spent, shorted some stuff in the process without killing everything, lots of soldering and finger blisters.
Fun project, the setup is ready for further testing, if some have ideas or want to provide modded bios to play with, why not, i'm opened to suggestions.
 

·
Premium Member
Joined
·
2,830 Posts
Discussion Starter · #676 ·
There has been at least two board that I know of, which have initially been using 16MB bios image despite having 32MB flash present on the system. In those cases there wasn't anything special with the bioses, they were simply FFh padded.

Because of that, I'd expect the original bios to work normally when flashed on a 32MB flash, given that the flash is properly installed.

Meanwhile nowadays, I don't expect ANY of the bioses for other chipset variants (400 or 500-series board) to allow the board even to post. Regardless if the board, outside of the chipset was 1:1 identical.
 

·
Banned
Joined
·
1,185 Posts
There has been at least two board that I know of, which have initially been using 16MB bios image despite having 32MB flash present on the system. In those cases there wasn't anything special with the bioses, they were simply FFh padded.

Because of that, I'd expect the original bios to work normally when flashed on a 32MB flash, given that the flash is properly installed.

Meanwhile nowadays, I don't expect ANY of the bioses for other chipset variants (400 or 500-series board) to allow the board even to post. Regardless if the board, outside of the chipset was 1:1 identical.
Yeah, was worth at least a try, just as an academic exercise and maybe to be able to get more cpu control options.
I have no interest into going AMD in the future, i will leave the ship as soon i upgrade.

I also tried to flash an UEFI image, where the bios images are switched, the one at the end being first, but no avail, no boot, not even the CPU+RAM leds.
I spent a bit more time fiddling with the SPI connection, the on-board bios boots, if all the pins of the add-on bios are connected BUT the clock signal pin.
I tried to twist the data line leads, but no avail, i was getting the same behaviour when i was close to the board anyway, so maybe something bother the SPI clock signal.
 

·
Banned
Joined
·
1,185 Posts
So...


Fiddling with the SPI clock signal pin, playing around with both bios chip, powered altogether.
I ended up corrupting the main on-board bios chip, the rig would not post further than the Strix logo.
Luckily, having wired the programming header beforehand, i hooked up the programmer to the laptop, and flashed the X370-F 5604 bios.
Again, big shout-out to TTAV134, the NeoProgrammer developer, great tool, detected the WinBond 25Q128 bios right away.
The 5406 is the latest bios for the X370-F, that doesn't kill memory support and overclocking, the following bios revisions, kill mem/fclk OC, in my case.

Since everything was hooked up, i was still curious about what would happen, if i would flash the half 32Meg bios, coming from the X470-F UEFI image.
Asus AMD X470-F bios are composed by 2x16Meg bios images, with different cpu support, the 1st beginning at offset 1000h to 1001000h, the second from 1001000h to 2001000h.
Rectangle Font Material property Parallel Number

I deleted the CAP Header, build 2 separate 16Meg bios images, and flashed them into the flash chip.
I started by flashing the 1st half of the X470-F 5406 UEFI image, big nothing, no boot, no led, no fan spin.
I then proceeded by flashing 2nd half of the X470-F 5406 UEFI image, got stuck, as usual, on CPU+RAM led on.
Which mean, at the end, that my piggyback moulded beauty, was at least partially working, it was able to load a bios and initialize the cpu.
Computer Personal computer Laptop Peripheral Netbook

I did the same for the X470-F 5809, just for the sake of completeness and as a last try before ending the project.
I straight flashed the 2nd half of the X470-F 5809 image, the 1st half doesn't have Matisse support, hence the motherboard staying completely silent.
To my surprise, the motherboard finally woke up from its slumber, booting with the ram led still turned on!
Rectangle Font Technology Display device Multimedia

It prompts the RGB firmware update, which is broken when using this bios, i also lost the motherboard RGB lighting.
It continued to post as usual, prompting me that no keyboard were attached, i checked it was, which mean i also lost some USB ports.
I checked carefully, i lost 2 USB 2.0 rear ports, 2 other USB 3.0 rear ports and the front USB that i'm currently using.
Font Wall Technology Signage Display device

It even managed to boot into windows, Micro$haft did its things, and installed new devices the 1st time i logged on.
Light Personal computer Computer Output device Lighting

I could reach windows, everything is working fine BUT the RGB, some USB ports, the memory support, the gpu stuck at 8X speed.
Unfortunately, it seems that the bios memory control is missing, loading anything lower than CAS18, would crash the board at boot.
Font Screenshot Software Technology Electronic device

In few words, as other X370 bios ports, these have an execrable memory stability, i cannot go easily past 3200Mhz, i can reach 3400Mhz with loose timings.
I haven't fiddled with the bios, i felt overwhelmed by all the new options, especially the one in the AMD overclocking menus.
Nevertheless, i finally got what i came for, the per CCX clock setting, that have been lacking on X370 for no reason, 3 years after Ryzen 3K launch!

Furthermore, i also tried to flash the following bios images:
-X470-F bios from the 4804 to the 5606 boots but get stuck on CPU+RAM led.
-X470-F bios from the 5809 to the 5861 boots fine to Windows BUT minus USB, minus RGB, minus memory support, minus pci-e 8X on the GPU .
-X570-F bios 1407 and 4021 boots but get stuck on RAM led.
-I didn't try to flash other ASUS bios, nor other brands bios.

As said previously, i'm not interested into upgrading to newer AMD cpu on old X370, i will go back to Intel, once i retire this test rig.
Tho, it truly deeply saddens me, that one have to go this far, just to be able to get the proper bios option support, that a 200e price tag overclocking Strix board deserve!
 

·
Registered
Joined
·
209 Posts
Hi, great work on testing a 32Meg BIOS chip upgrade for X370 motherboards!

In case you missed it, we do have a summarized list of BIOSes that crossflashes to Crosshair VI Hero (over at Crosshair IV Hero overclocking thread).


Link to the other thread:
Also, if you want Zen 3 support, I think 5821 should be the last BIOS to work. It has 1.1.8/9.0 which was not locked by AMD. Though 1.2.0.0++ should work with Zen1/+/2.
 

·
Banned
Joined
·
1,185 Posts
So...

Hi, great work on testing a 32Meg BIOS chip upgrade for X370 motherboards!

In case you missed it, we do have a summarized list of BIOSes that crossflashes to Crosshair VI Hero (over at Crosshair IV Hero overclocking thread).




Also, if you want Zen 3 support, I think 5821 should be the last BIOS to work. It has 1.1.8/9.0 which was not locked by AMD. Though 1.2.0.0++ should work with Zen1/+/2.
Thank you a lot, i lurked through the cross-flashing posts, but missed the spreadsheet.

Looking at the sheet, it seems that the ram control is a common issue, alongside getting 8X pci-e on the 1st gpu slot, as i did.
Maybe the pci-e slot tree was revised for the X470 boards, getting it closer to the cpu for pci-e Gen4 purposes.

I looked at the ram issues a bit closer, i thought it was related to memory layout profiles missing for the X370 boards, causing instability and the RAM led to light up.
I checked the ram voltage on the board, to my surprise, i found out it is capped to 1.2v, the VRM ram controller bios management is broken, when using the X470-F 1.2.0.0 bios.
Truth is, Asus replaced the Richtek single phase ram controller on the X470-F by an Upi controller, which could explain the lack of voltage control and the ram led lit up.

So i tried to flash the Prime X470 Pro bios, that ships with the same Richtek VRM ram controller, as usual, i tried to flash both the old and the newer AGESA versions.
The old AGESA get stuck at boot with the usual ram led lit up, the latter versions boots with the ram led lit, but lack ram voltage control.
So it has nothing to do with the VRM controller itself, the AGESA locks the VRM ram controller at a low voltage, effectively disabling memory tuning.
I wonder if Asus switched from a no boot on earlier AGESA, to a boot with locked ram on the latter ones, due to the number of users ending with a dead board after cross-flashing.
I did not try to switch my gpu in the second slot, to check if it would be recognized as 16X, unfortunately the 1st X370-F pci-e slot, becomes a 8X slot when flashing the X470 bios

The memory issue could be solved in two ways, either patching the bios with the old X370 MCU firmware, to restore the memory voltage control.
Either hard mod the VRM controller, to bypass the 1.2v lock, imposed on the memory by the AGESA firmware.
The single phase VRM controllers are pretty simple, the motherboard sets the memory voltage by changing the controller REFIN resistance.
On can either hook up a pot on the REFIN controller pin, or directly on the voltage control FB pin, or even play with the FB resistance network values, to restore the desired memory voltage.
Rectangle Slope Font Parallel Pattern


Anyway, i'm back to the OG X370-F lacklustre options bios for now, luckily i took pictures of my bios settings, so i could revert to my stable daily oc, without too much trouble.
Running 3800Mhz/C16/1T/1:1 on a R53600, my b-die are not binned well enough, to be able to pass anything bellow C18 with 1.2v.
I'm not sure that the old MCU firmware can be easily extracted and injected into the AGESA 1.2.0.0, pretty sure the keys or signatures would not match.
The VRM controller mod would be the easiest way to recover one of the most vital board functions, memory voltage tuning.
Unfortunately, i'm working on a water-cooled assembled rig, the probing and soldering job are not easy as on an open bench, but i will try to look further into the VRM controller voltmod.
 
661 - 680 of 744 Posts
Top