Resolve on AMD = Frustration

So, first of all im not blaming anything on Resolve, i just dont know what to do anymore.

My old notebook (1660Ti mobile) died, so i decided to go for an awesome editing machine.

Looking at prices and overall performance i decided to go with a 7800X3D backed up by a 7900XTX and 32G Ram(6000Mhz expo profile) on a B650E - F Board....should be enough i thought, and if next gen is more impressive imma buy a 5090 or whatever is op at this time^^

So, after TONS of Problems with ASUS/AMD Software i finally got to record some clips with okay quality.
After i put my 3rd clip in my timeline my playback became very stuttery and dropped to ~20fps.

One more clip and resolve shut down without error message. Tried to open my project...Crash.
Can't even open that project anymore.

So i thought its on me because Resolve NEVER crashed before, and my PC for sure can handle it.

New project, same problem. But i ended up with 7 clips on my timeline by saving after every added clip which kind of worked (copy copy copy copy...) until it completly broke again.

So i was heading over here and read about latest amd driver can cause problems.

Im now on the recommended 23.4.1 instead of 23.5.1, still no change.

I'm recording with AMD Adrenaline Software. AV1 Encoding, mp4, 1080p, 60fps, Video bitrate set to 100Mb/s max and audio at 192kb/s.

I mean i would buy a 4090 if thats the problem...but my 1660ti had no problems at all... sooo i guess i'm doing something wrong here...

Hope u guys got some ideas...i can record clips of whats happening etc... if that helps

btw, pc aswell as windows is just a few days old, so still completely fresh and clean.

As you are using overclocked memory together with a Ryzen 7800X3D I would suspect the problem here. It has been in the news that to drive the memory overclocked it needs higher voltages and those in turn can kill the CPU without triggering its protection.

I just read this article at German computer magazine c‘t:
https://www.heise.de/hintergrund/Defekt ... 48734.html

And some details in English:
https://www.tomshardware.com/news/amd-i ... c-voltages

TLD‘R:
Switch off the XMP/Expo profile and make sure your memory clocks stays at 5200 or lower. Otherwise your CPU gets literally fried.

Install the latest BIOS update, which should solve the issue:
https://dlcdnets.asus.com/pub/ASUS/mb/B ... ING%20WIFI

There are issues with AM5 ASUS motherboards that pump too much VSOC voltage that cooks X3D cpus.

As suggested, you should try and update to the latest BIOS before doing anything else as they fixed the voltage issue. Then disable EXPO and see if that helps.

The recommended gpu driver is 23.5.1 as it contains the fix for the memory leak that was happening on 23.4.2. Use DDU and with the internet disabled, try and install a clean version.

Make sure you also have the latest chipset drivers, X3D cpus need it.

Disable Windows fast boot / startup from the power options. This can mess with drivers that need proper system restart.

If nothing works, you should try and post some Resolve capture logs so BM can ake a deeper look.

The VSOC thing could have damaged your CPU (or degraded the CPU memory controller) if you ran it for some time without knowing of this issue.

Basically ASUS done effed up. If thats the case, I would RMA that CPU, AMD should replace it.

P.S If you have any Asus software installed, especially RBG stuff, remove all of it. Those cause nothing but trouble.

Okayyy soooo know about that soc thing and flashed new bios before i even booted it up.

Did not know about the 5200MHz rule, imma try that thank u.

My drivers and everything are 100% upd to date (except for the gpu for now...imma fix that in a minute...) and i will install the drivers without internet...but why is that important? =)

Uhhhhm one more question, about that capture logs ? where to find those? because theres no error message when it crashes^^

and again, thanks for helping me

Seems like my reply didn't work...

Uhm i've done all the missing steps now.

I flashed the BIOS before even firing it up the first time, so that should be okay.
Lowered memoryspeed and updated drivers back to current version etc...

Same as before. But i have all stages of failure saved in projects.

Project A: 3 or 4 clips in timeline, crashes instant on opening the project.

Project B: One clip in timeline, sometimes 60fps playback, but sometimes just stuttering.

Project C: 3 Clips but 3rd has "media offline" flashing when playback, but stable 60fps playback

Project D: 3 Clips, stuttery playback all the time.

Would a 4090 solve my problem? Seems like AMD doesnt know how make gpu software

Man, I think playing back 60fps H.264 recordings on anything would be rough -- even worse for a laptop. Have you seen this work before? Does it have to be 60fps? Does it work at 30fps?

If you're running into crashes, go to the Help Menu and select "Create Diagnostics Log on Desktop." Place that on a download site like Dropbox, and post the link here so that BMD Support can examine it.

[Rohit Gupta]

Project A: 3 or 4 clips in timeline, crashes instant on opening the project.

Let's focus on this. Can you please provide a log file after the crash? Can you please confirm which AMD driver are you using? We might need the clips you are using to reproduce the problem.

Presume you are using the Studio version.

Man, I think playing back 60fps H.264 recordings on anything would be rough -- even worse for a laptop.

Nah... even my i7 8700 with RTX3060 has *no* problems with smooth playback of H264 or H265 UHD material at 60FPS.

The latest bios updates solve any issues with overvolted memory controllers. The issues were serious for a tiny minority, the vast majority had none at all (my self included).

My 7950X3D hit's 113fps on the BM's CPU 8K RAW test.

Setting EXPO is no issue at all. If you want reassurance, download HWInfo64 and look at your VSOC voltage. As long as it's around 1.25-1.30ish, that's normal.

Make sure you have REBAR enabled in bios. Should be on by default, but you never know.

Make sure you installed the latest AMD chipset drivers, which include the X3D drivers.

https://www.amd.com/en/support/chipsets ... -am5/b650e

The latest bios updates solve any issues with overvolted memory controllers. The issues were serious for a tiny minority, the vast majority had none at all (my self included).

Setting EXPO is no issue at all. If you want reassurance, download HWInfo64 and look at your VSOC voltage. As long as it's around 1.25-1.30ish, that's normal.

It was a specific Asus thing more than other manufacturers. And yes, VSOC is not an EXPO only problem. It can mess with I/O (disk drives) as well and another bunch of stuff. It shouldn't be confused with RAM 1.35V voltage requirements which are fine.

How's your 7950X3D in Resolve? Would love to upgrade but I'm still irked about the 128/192GB ram speeds. I use AE a lot and the way it gobbles up RAM is astonishing so I'm not that keen on only getting 3600 or 4800mhz with 4 dual rank dimms.

My drivers and everything are 100% upd to date (except for the gpu for now...imma fix that in a minute...) and i will install the drivers without internet...but why is that important? =)

Its important because with AMD, Windows tries to force install a very dated driver via Windows Update.

On top of that, if you also try to install the official version WHILE Windows installs its own version in the background it could corrupt the OS files, especially if Factory Reset is selected.

Microsoft being Microsoft...

Allright got some news for y'all.

But first: sorry for double posting, i just realized EVERY post gets moderated.(thought its just the first one)

Also, i've read that u like the people to have their real name on the board, soooo HEY i'm mike ! =D

BTT:
AV1 Encoded files will just not work with latest AMD Driver. Some people got it fixed with DIFFERENT older versions. tried a bunch but gave up on it alltho one guy stated 23.1.5 did work for him.(tried 24.x.x and 25.x.x) sooo it seems to be on AMD for that one.

Tried to convert my AV1 files to h265 and h264 just for intererst (i knew about quality xD) but even those converted files did NOT work.

So i got some native h265 footage. and its fine xD

Im my case i just gonna wait for new drivers and let my av1 files sit for test purpose. Maybe THEY are corupted somehow because even the converted versions did not work (or the converter(handbrake) made a mess).

But still, i want to send u more details on the crashed projects, i'm now a crahs pro, i can make resolve crash in under 5 seconds soooo...what is it i need to send u? xD Where are those log files? because there is no error message at all. Resolve just vanished xD


One more tiny thingey: If i watch GPU utilization on taskmanager....even on rendering i barley have more than 7% load, is it displayed wrong or am i not even using the full potential?

You could transcode into ProRes files or DNxHD - those should fly. Can be done in Resolve when you export as individual clips.

One more tiny thingey: If i watch GPU utilization on taskmanager....even on rendering i barley have more than 7% load, is it displayed wrong or am i not even using the full potential?

It depends on what you are doing, not everything is taxing the GPU.

Crash logs, what to post and how to get those can be found here:
viewtopic.php?f=21&t=90190#25

So i got some native h265 footage. and its fine xD

Sounds like your original files are borked somehow or VFR (variable frame rate) maybe?

One more tiny thingey: If i watch GPU utilization on taskmanager....even on rendering i barley have more than 7% load, is it displayed wrong or am i not even using the full potential?

The default tab in task manager is for 3D usage. This is for games and some stuff displayed on screen (which would be the Resolve GUI).

For codec usage you need to change to Video Codec 0 1.
For compute usage -> Compute 0 1 or 3 (could be a little bit different on 7900series).

The latest bios updates solve any issues with overvolted memory controllers. The issues were serious for a tiny minority, the vast majority had none at all (my self included).

Setting EXPO is no issue at all. If you want reassurance, download HWInfo64 and look at your VSOC voltage. As long as it's around 1.25-1.30ish, that's normal.

It was a specific Asus thing more than other manufacturers. And yes, VSOC is not an EXPO only problem. It can mess with I/O (disk drives) as well and another bunch of stuff. It shouldn't be confused with RAM 1.35V voltage requirements which are fine.

How's your 7950X3D in Resolve? Would love to upgrade but I'm still irked about the 128/192GB ram speeds. I use AE a lot and the way it gobbles up RAM is astonishing so I'm not that keen on only getting 3600 or 4800mhz with 4 dual rank dimms.

Not just ASUS, Gigabyte had a bug that would completely ignore your manual VSOC, run 1.35 plus and tell you otherwise. The problem was a combination of AMD not clearly giving guidance on voltages, and motherboard manufacturers pushing out boards with multiple layers of adjustments that didn't sanity check.

The some manufacturers overclocking layers weren't talking to the AMD stock layers correctly because of shoddy programming, the AMD layer didn't have sufficient voltage protections, and finally manufacturers didn't realise the ramifications as AMD hadn't told them clearly. Every bios was affected to some extent to "potentially" have a problem, but it was very edge case.

Anyway, I love the 7950X3D. It's faster than a 7950X in the things I like to use. I have 2 x 32GB at 6200, basically stock but for EXPO. Tweaking really doesn't get you much more other than occasional blue screens, so I'll just leave it there.

I have seen 4 x 48GB at 6400, so it is coming...

The problem was a combination of AMD not clearly giving guidance on voltages, and motherboard manufacturers pushing out boards with multiple layers of adjustments that didn't sanity check.

That was speculation on some youtuber's part (like Gamer's Nexus). I really doubt they stopped giving guidance all of the sudden after 30 years in the CPU business. After observing in real life how some of the tech factories work with validation and such (much much lower down the scale from AMD for example but still known brands), this just sounds like the youtubers are clueless clickbait machines.

Contrary on popular belief, writing BIOS is very very hard stuff and people who do this are extremely rare. Which makes it more believable someone down the line messed up. Probably applying the X settings (completely unlocked) to X3D part that need a cap because of the 3D cache. And the overzealousness on the part of Asus with its out of the box OC is pretty known. Everything to get that 2% over the competition so you can charge triple the prices...

the AMD layer didn't have sufficient voltage protections

It never had protections before, all AMD cpus were and are unlocked (well most of them). When 5800X3D needed a voltage cap because of the cache, one would think the AIBs would not pump so much voltage through an X3D part...

Anyway, I love the 7950X3D. It's faster than a 7950X in the things I like to use. I have 2 x 32GB at 6200, basically stock but for EXPO. Tweaking really doesn't get you much more other than occasional blue screens, so I'll just leave it there.

I have seen 4 x 48GB at 6400, so it is coming...

Great. I've seen 4x48 at 6000 too on Level1 tech forums but its an unicorn. I'm waiting for all the BIOS shenanigans to pass before I switch platforms. Waiting on Wendell to also do some tests with OEM cheap RAM and see how much of a difference 4800 vs 6000 is...Nobody seems to care to do it.

Well, the change log for AGESA 1.0.9.0 states:

PLAT-131921: [RPL]and [RPL-X3D] Limit AOD menu setting for SoC voltage input to 1.3V when not in LN2 mode
PLAT-132133: [RPL]and [RPL-X3D] Update AOD menu help text to reference LN2 mode for expanded voltage ranges
[PMFW-13058] Change SOC max Voltage limit in FW to 1.3V from 1.4V

Source

So it looks like AMD is to blame.

Well, the change log for AGESA 1.0.9.0 states:

PLAT-131921: [RPL]and [RPL-X3D] Limit AOD menu setting for SoC voltage input to 1.3V when not in LN2 mode
PLAT-132133: [RPL]and [RPL-X3D] Update AOD menu help text to reference LN2 mode for expanded voltage ranges
[PMFW-13058] Change SOC max Voltage limit in FW to 1.3V from 1.4V

Source

So it looks like AMD is to blame.

What makes you think any of that points in that direction? All it says is they capped the voltage. Reminder that voltage capping was off before as to allow OC (especially LN2).

All this says to me is mobo manufacturers (asus and apparently gigabyte) went crazy because AMD allowed them to do so...

Well, the change log for AGESA 1.0.9.0 states:

PLAT-131921: [RPL]and [RPL-X3D] Limit AOD menu setting for SoC voltage input to 1.3V when not in LN2 mode
PLAT-132133: [RPL]and [RPL-X3D] Update AOD menu help text to reference LN2 mode for expanded voltage ranges
[PMFW-13058] Change SOC max Voltage limit in FW to 1.3V from 1.4V

Source

So it looks like AMD is to blame.

What makes you think any of that points in that direction? All it says is they capped the voltage. Reminder that voltage capping was off before as to allow OC (especially LN2).

All this says to me is mobo manufacturers (asus and apparently gigabyte) went crazy because AMD allowed them to do so...

Quote:
"AMD has declared that anything above 1.30V VSOC is unsafe and should be avoided, but it didn’t clarify at which point. On the VRMs of the mainboard, at the back of the socket, or the CPU’s internal circuit? I had some talks with Gigabyte’s people, and they told me that they try to keep the VSOC voltage below 1.3V according to the SVI3’s information, meaning at the CPU’s internals, where we cannot take readings through any physical means, but only through software. Any lower VSOC SVI3 voltage than 1.30V, for example, 1.25V, creates problems for RAM kits with the EXPO profile enabled."

https://hwbusters.com/news/amds-scalabl ... d-is-safe/

Why are you so defensive of AMD?

Why are you so defensive of AMD?

Because it doesn't have anything to do with AMD? Blame who needs to be blamed.

It was known since the 5800X3D that the vcache doesn't like more than 1.3V and throughout the history 1.2ish VSOC was the norm.

All of the sudden Asus pumps more than this into their mobos and its AMD's fault?

How come?

The only thing they can be blamed is that they allowed uncapped voltages in the first place. But that's like blaming car manufacturers because you drove your car into a tree and they allowed it.

This was and is a 100% Asus and/or Gigabyte problem.

You can see the Asus in action here (minute 11:05) (the only known youtuber that knows what he's talking about). Even Hwinfo64 gives a warning on Asus boards...(minute 11:05)

Even your quotes from your source don't support blaming AMD, quite the contrary.

AMD said more than 1.3V is bad, Asus and (maybe) Gigabyte ignored that...why do they need to mention how much more than 1.3V is beyond the scope of anything.

Related to the OPs problem:

Adrenalin 23.5.2

Known issue: Application crash or driver timeout may be observed during playback of AV1 video content using DaVinci Resolve™ Studio.

Should be fixed in the next driver or so, probably.