I usually connect all the available power cables between MB and PSU. Did not realize that one of them is “optional”.
UPDATE:
I am very c lose to solving the issue.
Over last 48 hours I’ve been using my pc with 2 monitors plugged in by 1x DP and 1x HDMI.
Both cables were of a typical quality, the PC worked perfectly, no hesitating on the boot anytime. The boot was 100% repeatable.
I ordered 2 new UGreen DP cables, certified for 8K. (I checked them with a voltmeter for a potential 20th pin wiring) So I took the HDMI out and continued testing my PC with 2x DP. The failed boot happend a few boots later.
And here comes my question.
Would it be true for RTX cards, that the DP cables have to be plugged into the certain ports? Or may it be my GPU that is faulty and cannot boot with a certain DP plugged?
I start a new stopwatch now, I moved one of the monitors over to another DP socket and I will try to achieve the fail boot again. If it happens I will consider replacing my GPU.
I have only seen cases in which a port was permanently damaged in some way.
It will be very unusual for a GFX card to consider connected devices while booting. It should boot with no monitors at all.
These kind of problems (wiring causes no boot) make me think of electrical grounding and stuff like that… but it’s just a wild guess (I’m not an electrician) - make sure your electrical sockets (wall) are ok. If You have a multimeter, try to measure voltage between “neutral” and “ground” it should be (almost) 0V. If it is something like 20-40 Volts - there is your problem.
That would have no connection to particular setup, that would randomly cause the boot to fail.
Now I am sure, that it’s a matter of using 2 DP cables. When I use HDMI + DP (in any of the 3 sockets, whatever configuration) everything is rock solid. When I unplug the HDMI and replace it with another DP, the fail boot happens relatively quickly.
Today I put my hands on another GPU, that hopefully will shed some light on the situation.
UPDATE 2:
I’ve been running a lot of tests and different setups.
All possible configurations where one of the displays was plugged via HDMI worked without the slightest issue. The boot happens to intermittently fail only when using 2 DP cables.
This happens also with a different GPU (new RTX 4070 OC - not SUPER) that has a different power socket so at the same time we eliminate the possibility of faulty NVIDIA plug/power adapter.
Now I will come back to my GPU and try setting the MOBO PCI-E Link speed to Gen 4 and see if that changes anything. I will also try out the new motherboard tonight. That’s my last hope.
This is probably the last update and now I tag this problem as NOT SOLVED.
Nothing would help but using the HDMI + DP combo, which I most likely will have to stick to (which I hate, just for purity).
I asked ASUS for help and after 3 tries to get some help on the chat they assigned (I hope so) some high level tech support assistant to my issue.
EDIT:
Using another fresh motherboard (the same ProArt x670-e which is the 4th one I’m testing) with bios rev 2007 (it never ran the newest bios) everything seems to be ok. 2 DP Cables cause no boot issue so far (testing since yesterday 9pm - almost 20 h now).
I have a hunch that the latest bios irreversibly changes / damages something in the motherboard, so even after rolling back to rev 2007 it will still be broken for good.
Now I’m affraid to do any bios update… on one hand I have a stable system, on the other hand the temps are higher and Im lacking some crucial optimisations…
Asus plz fix.
Where did you find DDR5-6000 ECC RAM?
The MB has no way of knowing how many monitors are connected, or which ports are used. The best it could do during POST is “ask” VGA card “are you okay?”.
It is in card’s power to say “yes” or “no”.
In the past a failure at this stage caused pc speaker to emit distinct sequence of beeps.
Nowadays, with MBs communicating via leds, it is somewhat more unreliable (my MB claimed VGA error even while displaying things on both monitors).
While I have not ever seen a case like yours, the fact that you went through several MBs and GFX cards makes me think that maybe the problem lies elsewhere… I am not certain where that will be, but I have seen many monitors with failed DP and working HDMI. All of them failed over time. Maybe one of your monitors is on its way to failed DP?
Still, how should this cause failed boot… I do not know.
Maybe ask at GPU subsection of this forum about your card booting with “broken” monitors?
I think I found a very obvious answer.
Every bios upon boot checks if the display is connected.
There is loads of threads about how to get it bypassed and people are buying hdmi dummies so the motherboards can boot with no display. Not every mobo has an option to force boot with no display (to the best of my knowledge the server mobos have it).
So my theory is that my Dell monitors may have some conflict with this rig and very intermittently will not get recognized over DP. If there is HDMI plugged the signal is straightforward like 0 or 1, however DP works as a hotplug and gets initiated every time upon connecting. I even encountered some boots that took longer because whilst booting my monitors were in a process of going to sleep. They had to wake up and then the system would boot to be able to post.
Bingo!
Side note: old bioses (before eufi) had a setting to ignore keyboard and/or VGA errors during boot.
We truly live in the worst timeline
I ordered a hdmi dummy for testing but I hope that this issue will be addressed with the new bios revision. I just let the asus tech support know.
Obviously Never worked on NT4…
Technically it is NEMIX DDR5-5600 CL40 ECC RAM. But on ASUS motherboards, since the RAM doesn’t have any XMP or EXPO profiles in the SPD chips on the sticks, ASUS implements their own overclocking algorithm called AEMP (Asus Enhanced Memory Profile) that is selectable in the memory settings and that pushes the 5600 memory kit to 6000 CL32 timings.
I thought at first the 6000 timing was in the SPD chip but learned after investigating that it is simply an ASUS overclocking setting on the ProArt X670E motherboard. I don’t know whether it is available on the other Asus X670E board like my Crosshair Hero boards since I haven’t tried the ECC memory in those boards yet.
Same BIOS versions on both boards so I am assuming it would have the same setting available if ECC is toggled on the Hero boards.
Well, Asus is weird enough about blocking headless boot it might be a consistent behavior across their boards. None of the other mobos I’ve built in the past three generations have shared Asus’s hangups about this.
Haven’t hit the exact problems here with an Asus board but I have had Asus refuse to boot with DisplayPort to something else adapters. So whatever logic’s used to decide if a GPU’s hooked up the way Asus believes it needs to be is flaky.
Interesting. I have that same board, I need to look into this.
Oh well, I hope that the HDMI dummy dongle will solve my issue for good. This would not be this much of a pain in the ass whilst booting my pc in person, but when I consider remote booting, this would be painful to see no immediate feedback and having it failed…
For the $160 the Nemix kit cost, I thought it a cheap experiment to prove that the Asus X670E boards have good working ECC. Happy that is the truth and even happier to have the same exact timings as the non-ECC RAM I was using prior.
I am going to expoeriment with that AEMP setting some as my attempts at manually setting the RAM settings didn’t lead to system stability.
So which board do you have? Wasn’t sure with your reply. The ProArt or the Crosshair Hero?
Both are on BIOS 2604 from late November.
You need to set the ECC toggle to Enabled in the BIOS. The setting is in the AMD_CBS section under the UMC section if I remember correctly. Or just use the F9 search function in the BIOS with ecc keyword and it should go right to the setting.
I have both edac-util and rasdaemon installed to verify the functioning of ecc.
I have changed the refresh interval from the AEMP default setting from the default ~11K to 50000 to get a bit more bandwidth and lower latency in Intel Memory Latency Checker. Moved it from 78ns and 71795MB/sec to 72 ns and 81036 MB/sec. That is the only timing value I changed from the default AEMP timings. Haven’t played with anything else. Good enough.
I have the X670E ProArt. Is the Intel Memory Latency Checker a windows only thing? I don’t know how to check memory latency in Linux unless rasdaemon can do that as well.