[quicknick]

First of all I ask you to forgive me for being somewhat off-topic, but I need some feedback from you fellow Taichi owners. My problem with the Taichi + Ryzen 1700x (at stock freq.) is that the cpu single-core boost doesn't seem to work as it should. From what I read on other forums, running single-threaded benchmarks should always have one core boosted, even if that core changes dynamically as it seems the cpu (or windows?) assigns the thread not to one single core, but it often jumps from one core to another.
I tried Cinebench (single core), WinRar bench (single thread), Prime95 with only one worker, cpu-z, old single-threaded apps from the early 2000s (Super Pi, PiFast) - none of these are boosting one of the cores to 3.8 or 3.9GHz.
I know that individual cores do jump occasionally to 3.9, hwinfo64 shows that if i leave it running for a longer time (in the "maximum" column), but I think it just hits that freq for a very short time as I have't been able to see the current speed of a core at 3.9. Tried all the windows power plans (Ryzen Balanced included), all that changes is the idle behaviour (Ryzen Balanced and High Perf keeps all cores at 3.5, Balanced allows cores to downclock to 2.0 when idle). I don't remember ever seeing the nominal frequency of 3.4 - under load it always prefers 3.5. Also I don't remember ever seeing 3.8 in the "maximum" column in hwinfo. So it seems XFR is always active, which I'm not complaining about.
So, is the above behaviour normal on Taichi? If not, what should I look for? BIOS or windows settings? Maybe some special windows update for ryzen? Maybe a clean install, because this windows install is actually from a Gigabyte AB350 Gaming-3 that killed me for 3 weeks before getting the Taichi?

 

[quicknick]

Regarding the long POST time, this is my experience: Enabling ultra fast boot made no difference, still around 20 seconds from power on until the speaker beep. Disabling AM4 Advanced Boot Training cut the boot time significantly, I'd say it's more like 5 seconds now (ultra fast boot still enabled ofc).
1700X @ stock
32GB Corsair Vengeance LPX3200, 4DIMMs @2666

 

[quicknick]

I have 32GB (two dual-channel kits, four sticks of CMK16GX4M2B3200C16). Couldn't go higher than 2666 on bios 2.0, updated today to bios 2.4, there is marked improvement but still no 3200.

3066 seems to be stable on new bios, but testing is in progress. I noticed something weird however. If I try 3200 I get the famous boot loop, after which it starts at the default 2133. But after refusing to POST at 3200, I cannot seem to make it boot again at 3066 or even lower, so I have to reset the cmos. Thus, I believe the bios is thwarting my attempts to reach 3200, because after a unsuccessful try it simply disregards the changes that I make at the subsequent attempts (increasing SOC and DRAM voltages, playing with timings). Anyone experiencing this odd behaviour? Or am I doing something wrong here?

 

[quicknick]

Under "OC Tweaker", the setting you're looking for is called VDDCR_SOC Voltage.

 

[quicknick]

Quote:

Originally Posted by RaphyFR

Now even 2933 wont post

i m scared

That happens to me (and others) as well. When I try 3200 I get a boot loop, then it boots at 2133, after that it boot loops again regardless what speed I set. So I needed to clear cmos after each attempt. Couldn't find the "reset defaults" menu entry, thanks to those that pointed out F9

 

[quicknick]

Hiya folks! Kind of stranded here, so any help / advice is welcome!

System specs:

AsRock X370TaiChi
Ryzen X1700 (trying to upgrade to 5900X)
4x 8GB Corsair Vengeance 3200 (worked at 3066 for years)
GTX1070
Samsung 960Evo 500GB NVMe

Been using the TaiChi for more than 5 years with the 1700X, I stayed with BIOS version 2.40 as I was afraid newer versions might break something. The combo was flawless during that time, but now the itch for upgrading was getting stronger so I got a 5900X.

In preparation for the CPU swap I upgraded the BIOS to 7.10 using the bridge versions in-between. Starting with 7.00 I got a warning about bitlocker/ftpm and boot drive becoming inaccessible, but I'm not using bitlocker and I don't know what ftpm does, and frankly not being able to use the same Windows installation after a BIOS or CPU upgrade seems... inappropriate.

Anyway, after upgrading to 7.10 I let the system boot into Windows, all was still ok. However, first start after swapping the CPU I got a message that I'll abridge, but it basically said:

New CPU installed, fTPM/PSP NV corrupted or structure changed.
Press Y to reset fTPM
Press N to keep previous fTPM record and continue system boot.

Pressed N but nothing happened. (maybe code A9 appeared on the screen, bottom-right. I didn't look at the onboard led display). Waited a few minutes and pressed reset.

Same message appeared so i pressed Y this time. Board restarted a few times then got stuck in a comatose state where:

• it doesn't respond to reset button
• it doesn't respond to power button (short, long, extra-long press, no matter. I waited for some minutes before trying to reset or power down)
• all fans spin at full blast
• it shows no code on the onboard POST display

Seeing that nothing is about to change, I flicked the switch on the PSU.

Found some info that using only one stick of RAM increases your chances of success, so I removed 3 out of 4 8GB sticks that I'm using, and at some point (after many times getting stuck as described) Windows said it tried to start but couldn't, tried a repair but during repair the system would reboot and become stuck again! AAAaaarggghhh!!!

Reset the CMOS a few times, even re-flashed the 7.10 BIOS, no avail. Some restarts, codes flashing on the led diagnostic display, then stuck in coma.

Put back the 1700X, Windows said that it's undoing some changes (that were made during the reset/repair attempts, I suppose) and luckily it seems nothing is lost. I made a backup of critical stuff before starting the BIOS updating, but I'd very much like to keep the current Windows install.

At this point I'm out of ideas, I don't know if the systems gets stuck that way because it doesn't like my Windows install (might test that tomorrow, by trying a new Windows install on some spare SSD), or the board is not really compatible with 5000-series. Wouldn't surprise me too much, as my previous board (AsRock 880G Extreme-3) claimed to be AM3+ compatible with the latest BIOS, but trying to upgrade to a FX-series CPU resulted in strange behaviour / unusable system, so back to my trusty Phenom I was...

 

[quicknick]

Thanks for the replies. I know I should've been more clear - I'm using Windows 10 and I'd have liked to keep my install, but it seems this is a futile attempt.

Today was a case of heaven&hell for me, because at some point I managed to boot my existing install.
For that, with the 1700X installed I did the following:

• "Clear TPM" via tpm.msc from Windows
• disable TPM in UEFI, save settings. Windows still booting fine.

Then I shut down to swap the CPUs, Windows boots no problem from first attempt. Then I successfully restarted a few times - to fit all the sticks of RAM, to bring them to rated speed (hit 3200 from first try, which was impossible for 5 and a half years with the 1700X + 2.40 bios). Did some benchmarking, played around a bit, but after a couple of hours or so I needed to shut down the system, to add the second fan to the Noctua cooler, tidy things up, close the case and move it back beside the desk where it belongs.

Big surprise, instead of shutting down the system entered the locked state where fans howl and your only choice is to pull the power!

Further attempts to start failed, at some point there was a "Starting automatic repair" message but after a few seconds it locked again. So I'm back with the 1700X and really don't know how to solve this. I'll try a fresh install on a spare SSD and report back.

 

[quicknick]

This looks more and more to be un-solvable.

Swapped back to the 5900X, cleared the CMOS, removed my NVMe boot drive and connected a SATA SSD that I had lying around. Started Windows 10 installation from (freshly-created) USB stick, cleared all the partitions that were present on the SSD and during the install the system restarted a few times. At one of the restarts everything locked with the already well-known symptoms. Forced the restart a few times by power cycling the PSU, a few times the "Attempting automatic repair" appeared but in a few seconds it would lock again.

Cleared the CMOS again then tried installing Windows 11 from USB. Again removed all partitions, and this time it was more clear-cut: the system hung right after the first restart (when Windows announces the reboot in 10 seconds and you can bring it forward by pressing OK). Cycled power a few times, no "Auto repair" or any other message, just hangs after about 2 seconds of that Windows-thingy spinning.

Put back the 1700X and NVMe drive and everything automagically works, no matter the UEFI settings. Defaults or not, TPM or not. My old Windows 10 install from the spring of 2017, that I carried over from a previous board (had a Gigabyte B350 Gaming 3 before switching to the TaiChi).

At this point I'm almost sure that this board just wasted two days of my life, and unless some miracle BIOS shows up over the weekend I'll be on the lookout for a new board. If anyone has suggestions about a cheap but ok-ish B550 board please shout. Any brand but Asrock.

 

[quicknick]

RAM model is CMK16GX4M2B3200C16. That's a 2x8GB kit, and I have two identical kits installed for a total of 32GB.

How I ended with this config is another long story, but basically my first Ryzen board was the Gigabyte Gaming 3 and initially I had one such kit (16GB) but it wouldn't run anywhere near its rated speed so I returned the kit and bought another one instead (HyperX Dominator, maybe?) that was on the QVL list for that board. And guess what, it ran worse than the Corsair!

So again I returned the kit (guys at the store weren't happy to see me ) and instead bought two kits of the Corsair. "If I can't have fast ram, at least I'll have much ram".

The Windows 10 install attempt was with bios defaults but RAM bumped to XMP profile (3200MHz, 1.35V).

Windows 11 attempt was with everything on default ([email protected]).

What do you suggest, try another brand/model of ram? I guess I can return it if it doesn't solve the problems, the guys at the store surely can't remember me after more than 5 years )

I'll want to stay at the 32GB mark, should I go for 2 or 4 sticks to achieve that?

 

[quicknick]

Thank you for your time and patience.

Again swapped the CPUs, resetting the CMOS at the same time.
Removed ALL the RAM just to see how it behaves, if it will enter the "frozen with turbo fans" state. It didn't, just politely beeping from the speaker.
I also removed the NVMe drive and reconnected the SATA SSD because I'd like to leave my old install out of the equation.

After this I installed a single stick of RAM and followed your instructions in detail regarding the voltages.

Started anew with a W10 install, passed the bit where you set region and language, then left the room for some minutes. Found the system tits-up when I returned. I guess it reached the point of first restart and that was it. "Works as intended".

Forced restart, wanted to check if the voltage settings held up, pressed F2 at the right moment but it locked before entering the UEFI setup. I think that's a first.

Again power off from PSU, I disconnected the SSD and BluRay drives, was able to enter UEFI, voltages and LLC were as you advised, RAM running at 3200.

I booted a USB stick with memtest86 v10.0, it's now at 2nd pass without errors, and I have a hunch that this will be true even with all 4 DIMMs populated.

At moments I remember how I thought this would be an easy upgrade

 

[quicknick]

Midnight update (well, actually 1 hour past midnight for me):

Memtest-ing a single stick of RAM in slot A1 passed with flying colours.
Doing the same with all four slots filled - I kinda gave up after 3h40m, but no errors during that time and it completed 3 out of 4 passes.

Restarted the system and let it boot from the SATA drive that contained the last failed attempt to install W10. Setup continued and in a few minutes I was up and running in a fresh install, for the first time after this adventurous CPU upgrade.

At this point, wishful thinking got the better of me and I actually thought that the problem is solved and the culprit was the BD-RW drive. Crazy, but not crazier than a issue I encountered ~20 years ago where an Athlon XP system consistently froze because of a faulty mouse

First restart attempt brought me back to reality, though. Instead of restarting it just hung in there with fans roaring. But unlike yesterday, where I was unable to boot again into my old W10 install after the borked restart, today it managed to start the new install each time, and I even got to 4 working restarts in a row before one would fail again. Yay, record!

Disconnected the SATA drive and hooked my NVMe stick to the never-used-before slot (the one that isn't covered by the CPU cooler and I think is also slower... PCIe 2.0 maybe?) and guess what, I'm writing this from my old W10 install while in the background I have 24 'workers' of Prime95, none of which complains. HWiNFO reports 4.6GHz on all cores (isn't that much higher than it's supposed to be?) and a Tdie of around 75 deg.C so far. Also CPU-Z reports core #0 hovering at 4575 to 4600MHz.

All of this won't fool me again, and I'm aware that another 'failure to restart' is just around the corner. I think that's the best description of the problem. At each restart (and I think also at shutdowns) the motherboard spools the fans to the max for 1-2 seconds, and I guess what's happening now with the new CPU is that the process just hangs there.
Or maybe I'm just too tired and this doesn't make any sense and can't really explain what happened the other day where my old install wouldn't boot anymore after working just fine for about two hours.

Anyway, tomorrow I might be able to borrow some RAM from my dad, see if that changes anything.

 

[quicknick]

This means I've been very lucky so far, apart from the 5900X adventure

I ran all 4 sticks at 3066 since June 2017 (when BIOS 2.40 was released) until last Thursday (when the 5900X arrived), and that was their maximum. Attempting 3200 resulted in a boot loop, after which 3066 wasn't achievable again without a CMOS reset. I believe I posted about the matter in this thread, back in the day.

Now, after upgrading to BIOS 7.10, I can run all 4 of them at 3200 with either CPU, so that's a plus.

Starting to believe that the problem I'm encountering with the 5900X isn't RAM-related, since it strikes even when I'm using a single DIMM at default speed (2133)

 

[quicknick]

@sane? - I might try and go past 3200 with my RAM, but first I have to find out what's keeping my system from (re)starting successfully each time.

Last night I threw everything at my computer and it ran like a champ. Even managed five restarts before I finally shut it down.
Today it started ok, I did my stuff, and before shutting it down I tried a restart. It hung again, and wouldn't start again if I power-cycled.
I was greeted with "Preparing automatic repair", then "Diagnosing your PC", then offered the choice of restarting or going advanced. Not much I could do on advanced, so I restarted only for it to hang again, and so on.

Entered BIOS to check @Dekaohtoura's settings and they matched except for the PCIe speeds which were set to Auto instead of Gen3. Also upped the RAM voltage a bit, from 1.350 to 1.375 volts, saved, again Automatic repair (because of the previous failed start I assume) but after the restart here I am again in my old Win installation. I'm not imagining the problem is solved, it just randomly decided to work.

While we're at it, I noticed that the CSM is enabled. Is that the right setting for it?
Also, I believe W10 can be installed and used on non-UEFI systems, would there be a way to force such a "legacy" installation on the Taichi? If so, would there be any drawbacks in terms of performance, stability, security?

@foxreinhold - if you don't run into weird issues like it happened to me, the performance is definitely there. Here's my GeekBench score:
X370 Taichi - Geekbench Browser
And here's a friend of mine with the same CPU but in a prebuilt Alienware system:
Alienware Alienware Aurora Ryzen Edition R14 - Geekbench Browser

As you can see, I'm almost 20% faster, despite slower and smaller RAM.

 

[quicknick]

Risked a restart to put in the values suggested by @Dekaohtoura, saved them but no boot after that. Not because of the values, but because something's very off with my system.

New development: it's happening from within Windows, too. First time when I started Hwinfo64, second time when I tried to play some DotA.

Thank you everyone for the help, but I ran out of patience. Ordering a new motherboard.

 

[quicknick]

This crossed my mind, but I've nowhere to test it. New motherboard should provide some answers

 

[quicknick]

I'm back with an update. Sorry for the long post, I'm trying to bring all the relevant details.

Tuesday the new motherboard arrived, it's an Asus Prime X570-P (at first I ordered the MSI MEG B550 Tomahawk but it would arrive early December so I canceled that and went for something that was in stock). Installed it, connected everything, equipped all four DIMMs, no POST. No beep, nothing. At least fans don't go stormy and long-pressing the power button is able to shutdown the machine. I tried many times, and a few attempts gave me some beeps. Sometimes long-short-short, other times just the first (long) one. But not the normal POST beep, which is lower in tone and more pleasant, as I later found. Nothing was displayed on screen during these attempts.

Started disconnecting things, and it seemed not to like one of my front case fans, as it POST-ed with it disconnected, only to go back to no POST when I reconnected it. But this was just another chaotic and nonsense behaviour, as further tries resulted in POST with the fan connected as well as no-POST with it disconnected...

After maybe one hour of frustration, although haven't identified a cause for all this, I got it to consistently POST so I refined some BIOS settings, enabled the DOCP or whatever Asus calls the XMP, 3200MHz seemed to work just fine, closed the case and moved it on the floor, where it usually stays. Again, no POST. Put it back on the bench in horisontal position, all ok. Back on the floor, tower position, not ok.

Ok, damn you, back on the bench but in tower position. Wouldn't POST. After a number of power-offs (via power button long press) the following screen appeared:

http://imgur.com/a/jbmDu5X

Which seemed to suggest that it doesn't like the memory @3200...
I really should have written down all that happened, because I cannot remember if after this I let it at the default 2133 or brought it to 3200 one more time , but anyway soon it was back to non-POST-ing. Again, it seemed that disconnecting three things - SATA cable towards the BD-RW and the USB 2.0 and 3.0 cables that go to the front panel would solve the problem, because it would work after that. I used it like that for the rest of the evening without further problems (at this stage I'm sure the memory speed was 2133, although most certainly automatically set by the BIOS when it presented me the "POSTed in safe mode" screen, which I saw 2 or 3 times until then. Restarted Windows a lot just to check and it all went smoothly.

Wednesday I had a busy day, but still managed to be around the computer for some time. Starting, restarting, shutting down, everything ok. Moved it on the floor and it still worked
Got the RAM to 3200 and it passed a full memtest:

http://imgur.com/a/u3yX77k

When I got back in the evening I used the computer a lot, benching, gaming, everything perfect.

Today I also used the PC a lot, then towards the evening moved the PC back to the bench and used my multimeter to thoroughly check the front USB 2.0/3.0 wiring/connectors and found nothing wrong. No shorts, no high-resistance paths between individual pins. So I connected them back to the motherboard. Can you guess what followed? No POST! Two times in a row I got the long-short-short beeps, then nothing. Restarted maybe 20 times and the "POSTed in safe mode" wouldn't appear any more. I noticed the fans on the video card behaving strangely:

Thought maybe it went bad somehow, but there was no beep even when I removed the video card!

I finally did get it to "POST in safe mode" after I started removing DIMMs and I was down to the last one... Quickly installed its pair, set them back to 3200 because that's the speed they're supposed to work, and here I am typing this at close to 4 a.m.

I'll make a last attempt to solve this by getting some new RAM, which will have to play nice with the 5900X on the TaiChi board. Yes, the Asus will go back to the store because of awful fan control and annoying active cooling on the chipset. Also, if the "simple upgrade" (yeah) I dreamed of turns out to be comprised of CPU+MoBo+RAM (instead of just the CPU), maybe it would've been better to jump directly to AM5 and maybe spare me this pain...

So please advice me what RAM kit to get ([email protected] I think, but brand? and 2 or 4 modules?). Thanks!

 

[quicknick]

Make a simple test: go to the bios, select the slowest non XMP speed available in the DDR profile and test if all the system is booting well.

Agreed, done that on the TaiChi and there were still a ton of problems with just a single DIMM at 2133MHz. Does this still apply now, with the new mainboard, if the system is perfectly OK with the memory at 3200MHz most of the time, only to act up at seemingly random times? I'm genuinely curious, is that the new reality with Ryzen (and, maybe, current CPUs from intel)? Because back in the good old days, things were clear cut. It either worked, or it didn't. Is the same RAM able to pass no matter how many rounds of memtest, hours of Prime95, then fail to POST next day?

Slowest non-XMP is 2133MHz, but when do I consider the problem solved? The system was flawless for almost two days with RAM at 3200, maybe at 2133 it will be also flawless for 5 days or one month before again failing to POST...

First off if you haven't already you absolutely need to update the bios to the latest one for that X570 motherboard before you do anything else. Make sure you load system defaults in bios BEFORE trying the update to try and make sure it won't crash. If that doesn't solve it for you then I'm serious: Disconnect the front panel power switch and see if that solves your problem. Sometimes there's a short in that and it might make the system soft turn off after turning on.

Updated today to version 4408, released this Tuesday - fan control is still awful, fans are ramping up every few seconds, tracking every temperature spike in any of the cores. There's an option to delay the ramp up/down but it doesn't do anything.

As for stability (successful POST at every attempt, better said), how can I assess that? I'm down at two DIMMs (16GB) at the moment, POST fail might never happen again, or things could get ugly again just after my RMA window closes.

During yesterday's prolonged no-POST episode I went full paranoid and disconnected everything from the board, except the power supply and the CPU and rear case fans and it still wouldn't POST when I used a jumper to short the PWRSW pins. Also multimeter-checked the Power and Reset switches and there's nothing wrong with them.

Didn't mention it in the first post, but during the initial troubles with the Asus board I also swapped the PSU. I wish I had another brand/model, but the one I had available is (almost) identical to the original one - SuperFlower Leadex Gold 650W, though they seem to be two different revisions as the replacement is around 1cm/half an inch longer.

I think the next test (tomorrow?) will see the TaiChi+5900X back in the system, but with another memory. For starters I'll use a 16GB kit from my father's PC, and if it changes something in terms of compatibility/chances to successfully restart, I'll go for a new kit of RAM.

 

[quicknick]

You seem to have some wild circumstances that are in no way normal for any of the systems you've tested and I'm not quite sure why. Are you sure there's no bent/broken-off pins on the CPU or something?

The 5900X looks perfectly fine, no bent or missing pins. Could it be that it's missing some bits internally

Kind of expected my dad not to have high-end RAM, but what I found still surprised me. A kit of HyperX Fury HX421C14FBK2/16. That's 2133MHz, folks! Okay, very bad, but at least it's different from what I have.

I had to postpone testing with these because I had to complete another job - assembling some cables for a Seasonic Prime PX-850 80+Platinum, a PSU that I got for next to nothing, but with all the cables missing. Job done, so all testing was done with this PSU - the third since starting my 5900X (mis)adventure. Post #1089 from the Repository Of Power Supply Pinouts on this forum helped me to a great extent

With the TaiChi back in the case, 5900X and Fury RAM, everything was sweet in BIOS, but when it tried to load Windows it hung with the already well-known symptoms. Fans at max speed, Power/Reset buttons acting like they are disconnected from the motherboard.

Tried everything I could think of. Reduced the memory speed to the minimum allowed by the motherboard, 1866 or so. Increased the RAM voltage to 1.35, although clearly not needed for such a low speed. Played with enabling/disabling TPM, CSM and other security stuff. Removed the NVMe and tried to install a fresh Windows on the spare SATA SSD - all well until the first restart. Not even one successful Windows desktop with the HyperX Fury.

Swapped the RAM for one of my Corsair kits, and actually the botched Windows install completed successfully. "Might be the RAM after all" - no, it just randomly happened. Just one lucky Windows boot, everything back to usual after I tried to restart. Again tried everything - down to one single DIMM in slot A2 as the manual recommends. Voltage up to 1.4, speed at default 2133. Used the secure erase SSD function from BIOS and tried a new fresh install - dead at first restart as expected. Moved video card to the other slot. Tried another video card, one so old that I'm ashamed to name it - UEFI didn't recognize it (5 beeps) then proceeded to load Windows and hung (could tell by the fans howling + buttons not working). Tried without any video card - same thing.

All the stoppages happened at the same point - one or two seconds after the "circle made of dots" thingie - the "please wait" sign of Windows 10. Can't really tell about the instances without video card, but I'm pretty sure it happened at the same point

I'm not immune to noob mistakes, but I'm pretty confident I haven't made any. There are no motherboard stand-offs where they shouldn't be, shorting something out. No half-inserted connectors, all golden fingers cleaned regularly with tissue+iso alcohol, dust blown from slots. With the 1700X back under the heatsink, everything just works. No matter the BIOS settings, CSM TPM *** or FFS, Windows just loads and I'm good to go...

Asus Prime X570P is already back in the box ready to be returned, question is - do I do the same for the CPU, or maybe give it one more try with a new kit of, let's say, G.Skill Trident Z Neo 32GB 3600MHz? The problems encountered on the Asus board - not starting until I removed 3 out of 4 DIMMs - suggest memory-related problems. Could the new CPU be so finicky about RAM that it's refusing to work with my Corsair kits (mostly on the TaiChi) and be outright offended when I try to pair it with the HyperX 2133MHz sticks?

 

[quicknick]

Was hoping to receive the new memory today (Trident Z Neo 32GB 3600MHz CL18) but the couriers are in no hurry. Can't say the same about me, because this week I must sort things out and decide what to keep and what to return from the stuff I bought.

During the last few days I read as much as I could from this thread - not all of it, but I went through all ~120 pages written since Vermeer launched, and to my surprise discovered that it wasn't smooth sailing for quite a few users who reported problems very similar to mine. Here's some snippets: (sorry for not using the quote function, but I had saved everything of interest in a text file).

tried 3333, 3267, 3200, 3133. Nothing works, it feels like mobo is not even attempting to work with those, everything just locks up.
The soft power and reset buttons stop working, not responding to long holds as well.
Set 3000 again and it booted like it was nothing. (@garych)

guys stop flashing 7.03 with Zen3 its dangerous.Its not stable with 4 dimms even with 2133 goes black screen,fans spins and reset/power buttons are not working.It need hard power off and clear CMOS to get training memory again. (@Senniha)

Did you made any changes to the PBO section ? i.e PPT/EDC/TDC values ? when mine is set to auto it won't boot. (@wizardwiz)

All good but windows boot stucks before starting, anyway I can see the round icon rotating, just for 20 second. Then at reboot ask me to recover for not good shut down.
System start correctly in safe mode! (@barrubba)

Last one, right after the release of 7.10:
1st major bug: Setting IF on auto, fcks up booting in windows, gets to recovery console, but everything is messed up. Setting it manually on 1:1 (1900, in my case), works as it should. (@Dekaohtoura)

I also learned a lot about various BIOS versions incorrectly setting the secondary ram timings and also about (God forbid!) crossflashing with the X470TC firmware. Anyway, since the new ram didn't arrive today and I'm gonna be quite busy for the rest of the week, I started tinkering with the system as if the new ram was there. And although I had a (mental) list of things to try, I went straight to the most radical item, crossflashing. And it actually works, so my board is now a X470 Taichi with P5.10 BIOS. Dunno if I'm the first to report it, but I confirm that the 5900X works with this BIOS (there was much talk about Vermeer not working with previous X470TC BIOSes when crossflashed to the X370).

Problems? Yeah, still, but nothing of the magnitude I experienced with 7.10. Twice I wasn't able to enter UEFI setup, system just froze with the Asrock logo and "press F2 to enter" still displayed. This seems to be solved for good after a CMOS reset and now I can enter setup at will.

The CPU temperature is a bit on the high side, often hitting 90 degrees C under Prime95. For now I'll assume this is because I completely run out of MX-4 (ordered a new tube, should arrive together with the RAM) so I had to use a very dubious and old "Halzniye HY510".

Also, adding the other 2 DIMMs and setting the frequency to 3200 (XMP) and entering the correct secondary timings resulted in BSOD-ing even at 1.4 volts, so for now I'm at 3000MHz @1.35V. And while we're at this, how can I find out the proper values for ALL of the secondary timings? The BIOS (XMP information) reveals little, and Thaiphoon burner even less, compared to the plethora of settings available under "advanced dram timings" or how it's called...

 

[quicknick]

Thank you, that's a lot of useful information!

Also a lot has happened to my system since my last post. New memory (G.Skill Trident Z Neo) arrived a day after that, I installed it and it was good to go @3600CL18 from the first attempt. To my surprise, I lost almost 2000 points in GeekBench 5 multi-core. Tried new memory @3200CL16 and all other timings from the old Corsair, didn't do much difference.

Turns out that it's well-known that Zen 3 performs worse with 2 dimms vs. 4, but somehow I was unaware. No problem, tried to recover some of the loss by playing with PBO limits, curves, etc. Managed to bring back around 1000 points, but still short of 14.5k. Anyway, after 1-2 days of trying various stuff I decided to go back to the original BIOS and turns out that it works perfectly with the G.Skill kit. Performance still hasn't improved and thermals are more or less the same (I made a big mistake earlier when I said the CPU is much hotter with the X470 BIOS. At the same time I switched to a new Prime95 version and this is much more taxing for the CPU. Testing with the older version resulted in temps well below 80 deg. C).

During this time I had many doubts and second thoughts, and for all the pain this upgrade inflicted to me I was very tempted to just put everything back in its box and get my money back and go AM5 at some point next year, so one night I decided that I'll stick to the 5900X only if it'll run on my old board with its original BIOS and together with my old RAM.

So off with the G.Skill, on with just a single pair of Corsair, system started, I ran a full round of TM5 with zero errors. Installed the second pair, after some time found the system in a BSOD. Decreased memory clock to 3066, and at this point it would pass TM5 "[email protected]" with flying colours. I also (re)started Windows for no less than 15 times, doing simple tasks at each restart (starting programs, copying files, some benching). At this time I had another of those "fake eureka" moments when I thought the problem is solved, and maybe it was caused by a bad flash (when using the built-in tool vs. using flashrom).

As you might imagine, next day when I turned on the system I was greeted with the well-known sound of full speed fans. But this time I was armed with the knowledge gathered from this thread and proceeded to test out some hypotheses. Infinity clock was on Auto, setting it to 1533 didn't change a thing. But playing with PBO limits immediately brought the system to life (I set some values very close to the defaults). And it works perfectly to this day, although I don't think I started my rig for 15 times yet, and it has shown that it's capable of at least that number even with improper settings

Asus motherboard and G.Skill ram are back in the store awaiting their next happy owner(s). Return window for the CPU ended few days ago, so I'm stuck with it for the near future. Crossflashing the X470 BIOS is always an option if the board starts acting up again (I'd prefer not to, but I might be forced to do so if the wild beast cannot be tamed with just playing with some random values that shouldn't result in this behaviour).

I wish I had a conclusion, but it's hard to draw one. I think the CMK16GX4M2B3200C16, or at least the version that I have, is somewhat incompatible with the Zen 3 IMC, and the problem is somehow exacerbated on the X370TC. Also I think that there's a bug in version 7.10 (that propagated from 7.0x versions) that makes the board go rogue when it encounters some settings it doesn't like. IMO, a beep, error message of any kind, automatic cmos reset would be far more helpful than just hanging with fans on "take off" and case buttons disabled...