New ASUS X570 BIOS Update [02/05/2021] - Version 3405

I flashed Bios 2606 on my Asus X570-i motherboard last night. I set my ram according to ryzen calculator. So i have had no stability issues since. I ran passmark memtest86 got no ram errors. Everything has been working much better.
Version 2606 2020/08/17
ROG STRIX X570-I GAMING BIOS 2606
"Improve system performance and stability
Improve Fan control function
Improve DRAM stability
Update AM4 AGESA combo V2 PI 1.0.8.0
Improve system stability
Improve DRAM performance"
 
Updated to the 2607 for my board. No issues for 24 hours with manually setting all RAM timings and voltages (safe settings) from calculator.
 
No this is a standard practice after a bios flash you cannot reload old settings.
Oh, hmm, that's odd. Unless I'm forgetting BIOSes of the past, this could be the first platform I've used that it didn't preserve settings between upgrades.
 
Oh, hmm, that's odd. Unless I'm forgetting BIOSes of the past, this could be the first platform I've used that it didn't preserve settings between upgrades.

I mean standard practice on AM4, since this is a AMD motherboard thread.

Prior sockets allowed it.
 
Well if had no issues so far, altho if not really tryit overclocking my ram any further yet then xmp yet either not sure how i should start kinda wish i got 3800 mhz ram since effects minimum fps quite a lot
 
Well if had no issues so far, altho if not really tryit overclocking my ram any further yet then xmp yet either not sure how i should start kinda wish i got 3800 mhz ram since effects minimum fps quite a lot
Not many chips can handle fclk of 1900. I've had 2 3900x chips and one 3950x. None of them would do 1900 fclk although it seems others have had better luck. The lower end chips have a better chance of working with fclk 1900.
 
If had quite some luck so far with chips i got in the past pushing quite far, im not sure if cmos clear resets fclk tho
 
If had quite some luck so far with chips i got in the past pushing quite far, im not sure if cmos clear resets fclk tho
I would believe that clearing CMOS will definitely take fclk back to stock, which is definitely not 1900. On auto, 1800 is as high as fclk will go.
 
Not many chips can handle fclk of 1900. I've had 2 3900x chips and one 3950x. None of them would do 1900 fclk although it seems others have had better luck. The lower end chips have a better chance of working with fclk 1900.

Damn, that's some bad luck.
 
Damn, that's some bad luck.
I've never won the silicon lottery unless you count an i7-2700k that would run at 4.8. I had to buy 3 chips to get there. I also had a core i7-930 that I couldn't get stable at 4.0 at any voltage below 1.3v I think.

On the Ryzen, maybe it's my RAM or motherboard holding me back. Who knows. You would think any B Die would be good enough, right?
 
I've never won the silicon lottery unless you count an i7-2700k that would run at 4.8. I had to buy 3 chips to get there. I also had a core i7-930 that I couldn't get stable at 4.0 at any voltage below 1.3v I think.

On the Ryzen, maybe it's my RAM or motherboard holding me back. Who knows. You would think any B Die would be good enough, right?

Not all B-die is the same.

I've seen B-die 3200 CAS 16 sticks that are inferior to the 3200 CAS 14 B-die chips on a ryzen build.
 
Last edited:
Now that I let the watchers in this thread know what memory chips that I am using in my system, if you are suffering from random reboots with the 2407 BIOS, then it is very likely that your system wasn't 100% stable with earlier BIOS versions with the same RAM and the same speeds and timings: While such memory running on systems with earlier BIOS versions lock up the system or cause program crashes, the 2407 BIOS will force a shutdown and reboot without ever letting you know. You see, if you are one of the unlucky ones who suffer from random reboots with the 2407 BIOS even when your PC is idling, then the next thing to do is to reset the memory speed and timings to the default JEDEC (not D.O.C.P.) profile, and then download and run Thaiphoon on your PC. More than likely you have memory parts that simply cannot handle such high speeds on any AMD platform.

Geez.
Hi. I'm new here. I just updated the bios on my Asrock Taichi X570 to 3.40. Now I get those random reboots. I'm using Corsair Vengeance LPX, Part CMK32GX4M4B3200C16 . I purposely picked 3200 speed as I figured it would be the most stable. I have it on XMP. Shouldn't this memory be fine, or do I need to find different memory? Thanks.
 
Hi. I'm new here. I just updated the bios on my Asrock Taichi X570 to 3.40. Now I get those random reboots. I'm using Corsair Vengeance LPX, Part CMK32GX4M4B3200C16 . I purposely picked 3200 speed as I figured it would be the most stable. I have it on XMP. Shouldn't this memory be fine, or do I need to find different memory? Thanks.
It's not your RAM. It's because every single BIOS version for AMD motherboards botch the XMP settings badly. You see, the AMD BIOSes only change the major timings and DIMM voltage but leave everything else as if it were running only JEDEC settings. This causes a mismatch between the different settings. In addition, newer AMD BIOSes actually undervolt the VDDCR SOC to less than 1.0V, which is a major cause for these reboots.

In other words, every single AMD BIOS gets XMP wrong.
 
Yeah, going to hard pass. The last update fixed my Asus Prime X570 -P memory issues and memory that wasn't supported now works with update before this one. Corsair Vengeance Pro 32 (16x2) 3200Mhz. G.Skill 4x8GB 3000Mhz, and VColor 32GB (16x4) 3600Mhz all work at rated speeds now. Corsairs timings made it a faster set than the VColor. System hasn't been more rock solid than after than the last update. All core 12 core 12 threads is at 4400Mhz, 4300+mhz all core 12 core/24 threads. I am happy, why mess it up, right?
 
It's not your RAM. It's because every single BIOS version for AMD motherboards botch the XMP settings badly. You see, the AMD BIOSes only change the major timings and DIMM voltage but leave everything else as if it were running only JEDEC settings. This causes a mismatch between the different settings. In addition, newer AMD BIOSes actually undervolt the VDDCR SOC to less than 1.0V, which is a major cause for these reboots.

In other words, every single AMD BIOS gets XMP wrong.

Also to add that memory CMK32GX4M4B3200C16 is not on the QVL list for that motherboard.

https://www.asrock.com/mb/AMD/X570 Taichi/index.asp#MemoryMS
 
Not all B-die is the same.

i'm seen B-die 3200 CAS 16 sticks that are inferior to the 3200 CAS 14 B-die chips on a ryzen build.
Yes. One has to know this. There are actually two different B-die parts, where the CL16-advertised versions are the inferior native DDR4-2133 parts while the CL14 versions are the native DDR4-2400 parts. One can determine this by the maximum (fastest) JEDEC profile on a given B-die memory stick.
 
Yes. One has to know this. There are actually two different B-die parts, where the CL16-advertised versions are the inferior native DDR4-2133 parts while the CL14 versions are the native DDR4-2400 parts. One can determine this by the maximum (fastest) JEDEC profile on a given B-die memory stick.
Huh, I guess you learn something new everyday. I guess I have the inferior B-Dies then. F4-4000C19D-32GTZR I never would have thought DDR4 4000 19-19-19-39 was of the inferior variety, but the G.Skill website says SPD is 2133. I have run them at 3733, but it took like 1.45-1.47v to keep them stable. Can't boot at fclk 1900 (3800) speeds.
 
Not all B-die is the same.

i'm seen B-die 3200 CAS 16 sticks that are inferior to the 3200 CAS 14 B-die chips on a ryzen build.
I assume you actually mean
I've seen B-die 3600 CAS 16 sticks that are inferior to the 3200 CAS 14 B-die chips on a ryzen build.
 
So is my 32 GB G.Skill Ripjaws V DDR4-3200 CL16 kit. Relatively speaking, my particular kit is based on a bad-binned B-die. No wonder why it cost $110 instead of $200 back in December of last year.

My kit was at $200 when I built my rig in december

Yes. One has to know this. There are actually two different B-die parts, where the CL16-advertised versions are the inferior native DDR4-2133 parts while the CL14 versions are the native DDR4-2400 parts. One can determine this by the maximum (fastest) JEDEC profile on a given B-die memory stick.

I went back to check my kit and I am at the DDR4-2133 speed.

thaiburner.png
 
Last edited:
My kit was at $200 when I built my rig in decimeter.



I went back to check my kit and I am at the DDR4-2133 speed.

View attachment 274295
Thanks. It turned out that I bought a CL16 kit instead of a CL14 kit despite using the same chip part numbers for the die. In this case, you get what you pay for.

I also got somewhat bummed out by my current DDR4-3600 CL18 Hynix MJR memory kit: Its tRFC latency (in ns) is much higher than my bad-binned B-die kit. No wonder why it was no faster (app-performance wise) at 3600 speed than my old kit was at 3200 speed.
 
Last edited:
I also got somewhat bummed out by my current DDR4-3600 CL18 Hynix MJR memory kit: Its tRFC latency (in ns) is much higher than my bad-binned B-die kit. No wonder why it was no faster (app-performance wise) at 3600 speed than my old kit was at 3200 speed.
I tried the Ryzen DRAM calculator, using the "Import XMP" feature, and my system woudn't even POST at even DDR4-3466 speed with the calculated settings, and it crashed on loading Windows at DDR4-3200 speed. This kit only likes the pre-programmed XMP timings at DDR4-3600 speed, which is much less than optimal for the Zen2 platform.

Right now I'm testing it at only DDR4-2933 speed with the calculated timings.
 
I tried the Ryzen DRAM calculator, using the "Import XMP" feature, and my system woudn't even POST at even DDR4-3466 speed with the calculated settings, and it crashed on loading Windows at DDR4-3200 speed. This kit only likes the pre-programmed XMP timings at DDR4-3600 speed, which is much less than optimal for the Zen2 platform.

Right now I'm testing it at only DDR4-2933 speed with the calculated timings.

For Thaiphoon burner where it shows your Revision / Raw card how many layers do you have?
 
For Thaiphoon burner where it shows your Revision / Raw card how many layers do you have?
8 layers.

By the way, I used the calculator straight, using the Hynix MFR profile without importing my memory's XMP, and actually did get my system running stably at DDR4-3466 speed. The calculator does not currently support speeds above DDR4-3466 for Hynix MFR parts.
 
Last edited:
8 layers.

By the way, I used the calculator straight, using the Hynix MFR profile without importing my memory's XMP, and actually did get my system running stably at DDR4-3466 speed. The calculator does not currently support speeds above DDR4-3466 for Hynix MFR parts.

The imported settings don't work for my Hynix DJR based kit either (ony any version of the BIOS).
 
I updated to 2606. It is more stable than 2407 for me, but still ...
Because of it being unstable, I wanted to downgrade to 2203. 2203 is stable in my setup. But no. The flashtool does not accept the old version :(
 
I updated to 2606. It is more stable than 2407 for me, but still ...
Because of it being unstable, I wanted to downgrade to 2203. 2203 is stable in my setup. But no. The flashtool does not accept the old version :(

I posted a method on page 5 that you can follow to revert to an older bios.
 
I posted a method on page 5 that you can follow to revert to an older bios.
Thank you for the tip!
I dont feel comfortable with performing the procedure so im going to wait for a new version.
My computer is stable when there is some load on the system. It's on "idle" the crash occures.
One good thing with 2606 is that I get higer benchmark scores :)
 
Well, 2608 is out now. I'm thginking of giving it a go, but I am still hessitant, as my current BIOS 2407 is working great. I mean, if it aint broken, don't fix it :) Still, I'm a bit adventures, so I might have a go, then if I run into issues, I can easily get back to 2407, and hopefuly be great again. :)

So I updated the BIOS to 2608 and set everything as it were on the 2407 BIOS. As I write this, just booted from the update, and seems fine wiht the same settings. This bolds well for the future, I think. Not sure if it's me or just the imaginaton of a new BIOS update, but the boot sequence seemed faster. Not sure :)
 
Last edited:
Well, 2608 is out now. I'm thginking of giving it a go, but I am still hessitant, as my current BIOS 2407 is working great. I mean, if it aint broken, don't fix it :) Still, I'm a bit adventures, so I might have a go, then if I run into issues, I can easily get back to 2407, and hopefuly be great again. :)

So I updated the BIOS to 2608 and set everything as it were on the 2407 BIOS. As I write this, just booted from the update, and seems fine wiht the same settings. This bolds well for the future, I think. Not sure if it's me or just the imaginaton of a new BIOS update, but the boot sequence seemed faster. Not sure :)
Sheesh. And I thought I was just kidding. Now I have to change my question to "when will 2801 be out?" :ROFLMAO:
 
I think I know why my system isn't stable or refuses to even POST with my 64 GB kit using the Ryzen DRAM calculator settings at anything above DDR4-2666 speed:

The calculator fails to account for higher memory IC densities. It assumes that all memory ICs are only 8 Gb density, which typically have a TRFC of 350 ns. That means the calculator's recommended values are applicable for 8 GB (single-rank) or 16 GB (dual-rank) sticks only. 32 GB sticks use higher-density ICs, which have a TRFC of a much higher 550 ns. That throws everything off.

It also means that at even the RAM's default JEDEC timings at the native DDR4-2666 speed, 32 GB sticks will perform more like a 16 GB DDR4-2400 stick than a true DDR4-2666 stick.

And I am not blaming the author for failing to update the calculator profile for Hynix M die chips: That part is now EOPL. This goes for both MFR and MJR parts. The only current Hynix IC that's 16 Gbit density and in an x8 configuration is the CMR die, which is not widely available at this point. (The CMR should not be confused with the AMR die, which is of an x16 configuration and is most often used in el-cheapo, low-performance, 8 GB single-rank DIMMs rated by the DIMM manufacturer or vendor at only DDR4-2400 speed - those DIMMs with only four ICs per rank.)

And the Ryzen DRAM calculator does not currently include Samsung C-die because there simply aren't enough samples of DDR4 DIMMs based on those ICs available for 1usmus to test to provide a valid calculator profile.
 
Last edited:
This has been a very useful thread for me to troubleshoot my system. Registered just to post my experience. My configuration is TUF Gaming x570 Plus (no wifi), RYZEN 3900X and Crucial Ballistix DDR4 3600 2x16GB. Memtest works fine, going through 4 passes with no error.

Each upgrade from 1407 made things a little more unstable. With version 2607, I could get 3200 MHz to run stable. Anything above that would give random reboots. DOCP settings gave a very unstable setup. The instability showed up both under Windows 10 and Linux Mint 20. With Linux I could watch the boot messages on the console. Could only catch a glimpse of some types of "Hardware error." One that stood out was the infamous "...No IRQ handler for Vector ..." error.

In any case, I downgraded to 1407 and the problems seem to have gone. System seems stable and the boot error messages are gone all the while running with DOCP at 3600 MHz. Will try tighter timings later if this stays stable.

Note on downgrading: I had to format my USB drive to Fat32 in order for EZ Flash 3 to recognize it as a valid bios file for the downgrade. Upgrades worked fine without the Fat32 trick. I don't know why this is, but I am thankful somebody found this:
 
Last edited:
Note on downgrading: I had to format my USB drive to Fat32 in order for EZ Flash 3 to recognize it as a valid bios file for the downgrade. Upgrades worked fine without the Fat32 trick. II don't know why this is, but I am thankful somebody found this:


Nice, I was just gonna ask about this, how you pulled it off. Fat32.. so simple.
 
Yeah, I remember seeing this on asus.com when I was looking for the BiosRenamer tool after seeing people post on reddit that renaming the bios file would allow downgrades. A simple renaming in windows did not do it for me. Couldn't find that link at first, but here it is: https://www.asus.com/support/FAQ/1038568

They recommend FAT16 or FAT32. The interesting thing is that I did not have to rename. Just the FAT32 format of the drive did it. That is what the youtube dude is pointing out.
 
Back
Top