I’m having some difficulties, recently replace an X99 Sabertooth and 5930k with a B550 Vision-D and 5950x.
The two Titans worked correctly in the X99 system, individually they work in the B550, but the board reports No VGA when both are installed.
Bifurcation is on in 2x8 mode. Above 4G decoding is on. I’ve tried with and without CSM enabled. No SLI.
System produces the No VGA beep code (long-short-short-short) but continues to boot into Windows successfully, but obvious no display.
I’ve tried the various ROM options with FAST boot and with that disabled.
I do use a riser cable on the second card usually but I have tested with the card installed directly in the board with no change.
Any ideas? Something I should test that I haven’t yet?
Plan is to replace the Titans with a 6800XT or 6900XT once stock becomes available (in South Africa) and prices are slightly more stable, so if the issue cannot be resolved it’s no biggy, I’m just curious as to why is happening.
More observations: it does NOT appear to make it into Windows. The fan curve changing as it does when the software loads must have been coincidental.
System will boot up with display with an 8400GS installed in slot two with one Titan. It also boots with an NVIDIA GRID M40 in slot two with one still populated by a Titan.
I’m going to swap the two Titan’s around and run some tests on the second card to make sure it isn’t experiencing any issues. It was working correctly a week ago, but who knows.
Use Gigabyte’s USB flashback functionality to reflash the BIOS
if it doesn’t help, then
Take out the GPUs, the CPU (!) and the CMOS battery, wait a few minutes then reseat them, then repeat 1)
This issue sounds familiar to something I’ve been experiencing with a few ASRock X570 Taichi Razer Editions, I think it’s a bug with certain newer UEFI-supporting GPUs and AMD’s current AGESA, at first I thought it was only Radeon VII GPUs but then I also experienced it with a 3090. There is some non-volatile area of UEFI settings that seems to lock up and the motherboard is f’d afterwards (CMOS Resets without any effect).
One ASRock X570 Taichi Razer Edition had been completely bricked (Error code 00, even USB flashback did no longer work, switching GPUs or only using a single GPU did not help) but with an older 2015 Maxwell-based GTX 960 2 GB it booted up again normally, after reflashing the UEFI via UEFI GUI and then with USB flashback the Radeon VIIs and the 3090 worked again, too.
Yes, it took me quite a long time to recognize the pattern, yesterday was the first time I’ve experienced it with a 3090 and I thought that my hypothesis was completely off until I also checked if a 960 would maybe make a difference.
If the motherboard had “reached this stage”, really nothing helped (error code 00, even after swapping out every component and a CMOS Reset) except using the old 960.
I think you avoided further degradation by testing the older 8400GS GPU that resets some motherboard UEFI setting otherwise you would have also reached the error code 00 lock-up.
Yeah this seems to be a hardening problem,
which i’m not fully sure if it is actually only B550 / X570 chipset boards related,
in regards to pci-e 4.0.
But it definitely looks like there is a pattern there.
I’ve seen this happening with two Radeon VIIs so PCIe Gen3-only, so while I agree that quite a few things have issues with PCIe Gen4, I think it’s a red herring here.
(BUT there is some issue with the way AMD disabled PCIe Gen4 on Radeon VIIs in hardware, if you don’t specify PCIe Gen3 in the AMD PBS settings of a motherboard’s UEFI but leave it on “Auto”, the AMD driver reports the Radeon VII to be running at PCIe Gen4 which is (sadly) impossible as far as I know)
UEFI-only/CSM didn’t make a difference in my case, Windows (20H2 or 21H1) was installed multiple times on an erased SSD so the drive had been GPT-initialized.
Since no mentioning of it I guess the following didn’t help?
If not, did USB Flashback actually work (blinking LED on motherboard for a few minutes)? As mentioned I’ve experienced a motherboard state where USB flashback itself could no longer be triggered.
No worries, I myself should be testing much stuff on personal systems that have no obvious causes like defective components and sometimes am one step before throwing everything out and becoming a technology-abstinent ascetic.
