PC Processors

[aquatica]

Specs:

Motherboard: Gigabyte GA-AB350-Gaming 3, UEFI F9d (currently, working)

CPU: AMD Ryzen 7 1700

RAM: G.Skill Flare X For AMD 2400MHz (QVL'ed), 2x8GB

GFX: Asus DUAL-GTX1060-3GB O.C

Operating System installed: Irrelevant for UEFI issue, but currently Windows 10 Pro 64bit

Drivers installed: Irrelevant, cannot boot to Windows most of the time with erroneous setup

Power Supply: Corsair 700GS, operating just quite nicely

As this issue seems to be with AGESA, I don't really know which forum would be the correct one... As it has a lot to do with the CPU, I decided to put it up here. MOD's please move this to elsewhere, if I'm posting in the wrong forum!

Simply: when I upgrade my motherboard's UEFI to anything newer, I start to get "No VGA"-error on start. It goes away exactly when I reflash UEFI with older version.

Sometimes I am able to boot with newer UEFI, but mostly not. Clearing CMOS usually helps, but if I change practically anything in the UEFI I get "NO VGA" light on the motherboard.

It might be enough to make the system completely inoperable by simply changing the boot order in UEFI, so the issue has to lie much deeper than just the UEFI itself.

Affected version of UEFI: Gigabyte's F10, F20, F21, F22b

Difference between the many versions of the UEFI:

As far as I know, AGESA. F9d apparently uses the older AGESA 1.0.0.6a, when F10 comes with 1.0.7.2a.

F20 and F21 then comes with 1.0.0.0a (I was told that this is complete rewrite of the AGESA?).

F22b comes with AGESA 1.0.0.0a + SMU 30.67.0 for Raven Bridge.

If I have any of the affected versions flashed, I *might* get the PC to POST. Most likely not. If I do and if I do enter UEFI, I probably fail changing *any* settings. But if I want to be very sure to get a fail, all I need is to tell UEFI to use my XMP Profile for the RAM. Guaranteed to fail after "Save & reset".

Now, even in the (very rare) case I do get to boot to Windows 10, the system is very unstable. Just like with broken graphics card or bad RAM. However, I know that my RAM is not bad (this is already once sent back to G.Skill and replaced, after which I did run extensive memtest86+ on the sticks... The RMA'd set failed at around 12hours 45minutes, so I ran the test for couple of full days to be as certain as possible, that my new RAM was OK.).

Same goes for the rest of the system: I have as much stability with older AGESA as I have ever had! Prime95? Bah, bring it on!

Running Prime95 with 3D Mark simultaneously? No problems! And so on, and so forth. So the hardware *should* be on full working order.

Also worth noting is, that the system is 100% stable with UEFI F9d, which is actually a pretty old BETA - afaik it's from around September 2017 - couldn't find actual release date.

So, do any of you great people have any ideas at all? This issue is not unheard of - in fact, there is one on this very forum on the Graphics card side (link to follow) with similar issue, but on Radeon graphics. Also few people online have had similar issues with AB350 chipset - so it looks to be inherently somehow connected to this - at least to some civilian like I am.

Gigabyte (motherboard manufacturer) has been informed. I am yet to receive a reply...

Asus (graphics card manufacturer) has been informed. I am yet to receive a reply from them as well...

Links to other similar cases I have come across so far:

https://www.youtube.com/watch?v=VAzNBtCiFws

https://community.amd.com/thread/221516

https://forum.giga-byte.co.uk/index.php?topic=18331.0

https://www.reddit.com/r/GAAB350/comments/6hqgdi/ab350_gaming_3_and_ryzen_5_1600_powering_on_but/

Many of these are of the same motherboard (GA-AB350-Gaming 3), but not all. So this is highly curious and I hope perhaps AMD could have *any* ideas - or anyone else here!

Thank you for your help in advance

PS.

I love my Ryzen, despite this issue!

[aquatica]

Apparently getting any input anywhere is difficult...

However, Asus replied already (very quick for them too, which is nice!) and they even tested the same GFX in their setup. No issues - they came in the same conclusion as I, that this is some weird off issue that happens only on certain graphics cards on certain setups without too much sense.

Gigabyte has not yet replied and we are 6 days in. AMD's support has not yet replied to my direct support request, that's maybe 3 days in (so in the weekends). I'm hoping I'll get a response from them during this week.

[aquatica · Accepted solution]

After prolonged period of time talking to Gigabyte, ASUS, AMD and my retailer I finally believe I have found a solution to this weird issue. Few more things I did to test was add a secondary GFX (ages old GeForce 6200LE PCIe) to notice everything works perfectly.

Then a user posted on Gigabyte forums that he had seen the same case as I and the reason was Gigabytes poor software - in this case the UEFI code for understanding graphics processing on UEFI (he called it UEFI GOP) of nVidia 10-series cards; this had been proven on GTX 1070 and GTX 1080 and "assumed on all" - and it seems to be so, as also GTX 1060's are affected.

This is supposedly fixed in newer UEFI code (while introducing an ANOTHER similar issue, if you disable HPET timer).

Oh, the way to fix this issue? There are few ways:

If you want Secure boot & all to work, you *will* have to have a secondary graphics card that the UEFI recognizes (as I did) on another PCIe slot, even though it would be disabled in the operating system (as I had).

If you don't need Secure boot & all to work, you only need to enable "CSM" in Gigabyte's UEFI setup. That's it. As it will then use Legacy mode to poll devices, it recognizes graphics adapter...

Third option is to get an another graphics card, preferably one that doesn't say nVidia anywhere near it.

So this is it, thank you for AMD's support who replied to me and even forwarded this case to the Ryzen team. Hope this helps someone one day!