Ryzen Owners Zen Garden

[Mussels]

You know, up until a recent AGESA update, sleep was completely broken on Zen 4. You'd put your computer to sleep and then seconds later it would wake back up. I wonder if that's somehow connected.

Sleep has a lot of software that breaks it that people never realise - discord for example is one of them, as it tells windows its a media player and prevents sleep from working

Browsers can do it too, so things like steam or Bnet being open on screen not only break sleep, but since they all use hardware acceleration they raise your GPU idle clocks too. Its a real PITA to make sure i quit discord before going AFK, tbh.

A single disabled device in task manager (or a yellow triangle) also breaks sleep, which screwed with some testing i was trying to resolve at one point with a virtual display driver for my VR headset

 

[tabascosauz]

But if the actual SoC voltage was higher with the older BIOS, then why is my idle power consumption and temperature 5 W and 5 °C higher now? Something doesn't add up.

Quick question: What is CPU_VDDIO? I've noticed that it's 1.35 V, same as my DRAM voltage. Is that normal?

I'm pretty sure VDDIO is that new memory controller voltage. Some boards call that CPU_VDDIO or VDDIO/MC iirc. Don't know much about it, only that it usually follows VSOC but slightly lower. Reminds me a bit of VCCIO the way the default values are. If VDDIO is what I think it is, you can try dropping it in increments to see if it changes power draw.

Otherwise, MEM_VDDIO is what Ryzen Master calls VDIMM. Pleasantly, MSI seems to still be calling VDIMM DRAM Voltage, unlike Gigabyte that invented some new gobbledygook for no reason

 

[Space Lynx]

Sleep has a lot of software that breaks it that people never realise - discord for example is one of them, as it tells windows its a media player and prevents sleep from working

Browsers can do it too, so things like steam or Bnet being open on screen not only break sleep, but since they all use hardware acceleration they raise your GPU idle clocks too. Its a real PITA to make sure i quit discord before going AFK, tbh.

A single disabled device in task manager (or a yellow triangle) also breaks sleep, which screwed with some testing i was trying to resolve at one point with a virtual display driver for my VR headset

I always disable sleep on every build I do, I either do a pure shut down or just let it run, I usually shut down every night before bed. I always felt like sleep messed things up.

That being said Arch Linux Steam Deck is insanely good at sleep mode. It's really refreshing hitting the power button and it powers down 0.5 seconds later, and hit it again and I am in the exact same spot in the game I was at before, no loading, nothing, all fluid and instant. Steam Deck and Arch Linux SteamOS are absolute game changers and I love it

 

[the54thvoid]

Watching all of this, I have a question:

If I buy an AM5 board, and use RAM rated at 6000(OC), how do i avoid the burn out? Or is it better to run ram at base speed until the BIOS firmware is certain to not fry the chip.

This all assuming the CPU is not overclocked.

 

[Space Lynx]

Watching all of this, I have a question:

If I buy an AM5 board, and use RAM rated at 6000(OC), how do i avoid the burn out? Or is it better to run ram at base speed until the BIOS firmware is certain to not fry the chip.

This all assuming the CPU is not overclocked.

a 6000 EXPO kit doesn't burn out the cpu, its only people who tried to manually oc their ram beyond rated speeds (turning on EXPO is fine) that experienced the burn out (this is my understanding anyway), also to my knowledge all motherboards already have the latest BIOS update fix for it, I know because I checked in last few days. as I am considering 7800x3d still myself.

 

[tabascosauz]

Watching all of this, I have a question:

If I buy an AM5 board, and use RAM rated at 6000(OC), how do i avoid the burn out? Or is it better to run ram at base speed until the BIOS firmware is certain to not fry the chip.

This all assuming the CPU is not overclocked.

From AMD's mouth 1.3V SOC is safe, and you should be able to do plenty of UCLK with 1.25V or less. Unless you get exceedingly unlucky, but you're not me

 

[3x0]

a 6000 EXPO kit doesn't burn out the cpu, its only people who tried to manually oc their ram beyond rated speeds (turning on EXPO is fine) that experienced the burn out (this is my understanding anyway), also to my knowledge all motherboards already have the latest BIOS update fix for it, I know because I checked in last few days. as I am considering 7800x3d still myself.

That's wrong, enabling EXPO 6000 rated RAM can juice the VSOC over 1.35v. Plus, the VSOC voltage is only part of the problem, but that requres more testing from 3rd parties since they haven't been upfront about all the issues.
A lot of the vendors updated the BIOS so it should be safe, as you've noted

 

[AusWolf]

Watching all of this, I have a question:

If I buy an AM5 board, and use RAM rated at 6000(OC), how do i avoid the burn out? Or is it better to run ram at base speed until the BIOS firmware is certain to not fry the chip.

This all assuming the CPU is not overclocked.

As I understand, update your BIOS to the latest, and/or never let your SoC voltage go above 1.3 V.

Also, watch your power consumption when PBO is enabled, as some boards seem to disable OCP with PBO for some reason.

I'm pretty sure VDDIO is that new memory controller voltage. Some boards call that CPU_VDDIO or VDDIO/MC iirc. Don't know much about it, only that it usually follows VSOC but slightly lower. Reminds me a bit of VCCIO the way the default values are. If VDDIO is what I think it is, you can try dropping it in increments to see if it changes power draw.

Otherwise, MEM_VDDIO is what Ryzen Master calls VDIMM. Pleasantly, MSI seems to still be calling VDIMM DRAM Voltage, unlike Gigabyte that invented some new gobbledygook for no reason

Ah I see. So you don't think 1.35 VDDIO is dangerous? I was wondering if it was connected to the SoC somehow, or just a fancy way of saying "DRAM voltage".

 

[Space Lynx]

That's wrong, enabling EXPO 6000 rated RAM can juice the VSOC over 1.35v. Plus, the VSOC voltage is only part of the problem, but that requres more testing from 3rd parties since they haven't been upfront about all the issues.
A lot of the vendors updated the BIOS so it should be safe, as you've noted

well I think I will skip this generation after all and just keep my current setup... a shame indeed...

 

[tabascosauz]

well I think I will skip this generation after all and just keep my current setup... a shame indeed...

5800X3D......whatever the price is, still gets a lot more value for your money than moving to AM5

Silver linings you know, I was on the verge of giving in to fomo when 7800X3D dropped, but AMD reassured me by blowing up some CPUs

Jokes aside, all the AM5 controversy doesn't change my opinion on any of the Raphael CPUs; I think the most important part is just getting to the bottom of the problem and having a certain explanation of what's actually causing it. Unfortunately no one has gotten to that point yet, it should restore a lot of confidence.

 

[Space Lynx]

5800X3D......whatever the price is, still gets a lot more value for your money than moving to AM5

Silver linings you know, I was on the verge of giving in to fomo when 7800X3D dropped, but AMD reassured me by blowing up some CPUs

Jokes aside, all the AM5 controversy doesn't change my opinion on any of the Raphael CPUs; I think the most important part is just getting to the bottom of the problem and having a certain explanation of what's actually causing it. Unfortunately no one has gotten to that point yet, it should restore a lot of confidence.

well my irl friend still wants my 5600 for 100, so only a 30 loss there. so i will grab the 5800x3d next time it hits 299 sale. he said he doesn't mind waiting. so yeah I think that is the best path for me. looks like my 6800 xt won't sell, so a 5800x3d will pair nicely with that gpu anyway. should be a solid rig for years to come.

 

[Mussels]

I'm pretty sure VDDIO is that new memory controller voltage. Some boards call that CPU_VDDIO or VDDIO/MC iirc. Don't know much about it, only that it usually follows VSOC but slightly lower. Reminds me a bit of VCCIO the way the default values are. If VDDIO is what I think it is, you can try dropping it in increments to see if it changes power draw.

Otherwise, MEM_VDDIO is what Ryzen Master calls VDIMM. Pleasantly, MSI seems to still be calling VDIMM DRAM Voltage, unlike Gigabyte that invented some new gobbledygook for no reason

Somewhere betwen ryzen 1000 and 3000 they renamed a bunch of those voltages, and split one into two like that on AM4
It was in one of the agesa notes i read yesterday, but i have no clue which ones they were

certain explanation of what's actually causing it. Unfortunately no one has gotten to that point yet, it should restore a lot of confidence.

So far i've gotten a few solid facts from the high end techtubers

1. Some boards dump far too much power while booting, trying to get the system to boot (before the CPUs safeties kick in)
1.a. AM5 long post times at high RAM speeds might expose the CPU to that for long periods
2. Auto SoC can bug out after sleep states, going right to max values (1.8v) (May be board specific)
4. At least one sample was shown that as the CPU degraded from high SoC voltage exposure its resistance lowered, causing voltages to be even higher than expected
(Board safety features should have cut that off, but didnt)


It's like AMD assumed the board makers had OVP and OCP enabled, and the board makers assumed AMD had it covered - but in situations where the CPU wasnt fully initialised, the lack of board safeties led to CPU damage, which then led to gigabyte PSU level explosions.
How one high end premium board sent 400W to the CPU socket, when nothing on AM5 remotely approaches that wattage is a massive mystery - NOTHING on the board prevented it??

 

[A Computer Guy]

Somewhere betwen ryzen 1000 and 3000 they renamed a bunch of those voltages, and split one into two like that on AM4
It was in one of the agesa notes i read yesterday, but i have no clue which ones they were



So far i've gotten a few solid facts from the high end techtubers

1. Some boards dump far too much power while booting, trying to get the system to boot (before the CPUs safeties kick in)
1.a. AM5 long post times at high RAM speeds might expose the CPU to that for long periods
2. Auto SoC can bug out after sleep states, going right to max values (1.8v) (May be board specific)
4. At least one sample was shown that as the CPU degraded from high SoC voltage exposure its resistance lowered, causing voltages to be even higher than expected
(Board safety features should have cut that off, but didnt)


It's like AMD assumed the board makers had OVP and OCP enabled, and the board makers assumed AMD had it covered - but in situations where the CPU wasnt fully initialised, the lack of board safeties led to CPU damage, which then led to gigabyte PSU level explosions.
How one high end premium board sent 400W to the CPU socket, when nothing on AM5 remotely approaches that wattage is a massive mystery - NOTHING on the board prevented it??

I feel like this is a real nightmare for AMD consumers - perhaps much worse than the memory compatibly issues in early AM4.
Ryzen X3D brand becomes Braisin 7800 X3D (as in the root word Braising - to fry (cpu) lightly and then stew it slowly in an enclosed heat spreader)

I saw a recent video der8auer has a new AM5 direct die cooling kit he is working on that seems really effective. I wonder with these AM5 issues if it becomes possible to still fry an AM5 CPU from these issues despite much better cooling. (Sorry for the umteen post edits. I shouldn't post so late a night)

 

[tabascosauz]

Somewhere betwen ryzen 1000 and 3000 they renamed a bunch of those voltages, and split one into two like that on AM4
It was in one of the agesa notes i read yesterday, but i have no clue which ones they were

So far i've gotten a few solid facts from the high end techtubers

1. Some boards dump far too much power while booting, trying to get the system to boot (before the CPUs safeties kick in)
1.a. AM5 long post times at high RAM speeds might expose the CPU to that for long periods
2. Auto SoC can bug out after sleep states, going right to max values (1.8v) (May be board specific)
4. At least one sample was shown that as the CPU degraded from high SoC voltage exposure its resistance lowered, causing voltages to be even higher than expected
(Board safety features should have cut that off, but didnt)

It's like AMD assumed the board makers had OVP and OCP enabled, and the board makers assumed AMD had it covered - but in situations where the CPU wasnt fully initialised, the lack of board safeties led to CPU damage, which then led to gigabyte PSU level explosions.
How one high end premium board sent 400W to the CPU socket, when nothing on AM5 remotely approaches that wattage is a massive mystery - NOTHING on the board prevented it??

Is the bugged out 1.8V SOC actually probed though, or just HWInfo logged? HWInfo has also been logging these 2.6V max values on Vcore and VSOC, which is clearly bogus since 2.0V should just about kill CPUs instantly, regardless of current - seems like software monitoring can't be trusted on this issue.

It's a shame, really, that all this high tech goes to waste due to incompetence and unfinished firmware. I guess that's the downside to having VRMs that are overkill by orders of magnitude - without OCP, CPU's gonna die long before a heatsinked 16 x 110A SPS even overheats.

I feel like this is a real nightmare for AMD consumers - perhaps much worse than the memory compatibly issues in early AM4.
Ryzen X3D brand becomes Braisin 7800 X3D (as in the root word Braising - to fry (cpu) lightly and then stew it slowly in an enclosed heat spreader)

I saw a recent video der8auer has a new AM5 direct die cooling kit he is working on that seems really effective. I wonder with these AM5 issues if it becomes possible to still fry an AM5 CPU from these issues despite much better cooling. (Sorry for the umteen post edits. I shouldn't post so late a night)

Braisin' lmao

The direct die stuff is cool on any CPU. Unfortunately, neither SOC degradation nor the catastrophic failures are something that cooling can mitigate. All the damage is in the die or the substrate.

 

[Athlonite]

Is the bugged out 1.8V SOC actually probed though, or just HWInfo logged?

Yeah Steve from GN had it probed and turns out that ASUS is the biggest of the problem mobo makers to be doing stupid stuff with their high end AM5 mobo's none of them had correctly working OCP/OVP for the CPU working allowing far to much current draw before cutting it off at which point it was to late the CPU was cooked when it reached 207 degrees C

 

[tabascosauz]

Yeah Steve from GN had it probed and turns out that ASUS is the biggest of the problem mobo makers to be doing stupid stuff with their high end AM5 mobo's none of them had correctly working OCP/OVP for the CPU working allowing far to much current draw before cutting it off at which point it was to late the CPU was cooked when it reached 207 degrees C

That's a different issue altogether; Asus is pushing high VSOC and many boards are further overvolting above software VSOC numbers when under load, but Mussels was referring to VSOC being bugged out and reading 1.8V after sleep (which I don't doubt for a second, given Ryzen's long history of buggy software monitoring). True application of 1.8V would kill CPU immediately - there's no time to read out a 1.8V reading anywhere in software.

Which, now that I think about it, also means my original question is incredibly stupid and answers itself, since you can't probe 1.8V out of a dead CPU any more than you can read 1.8V in HWInfo on a dead CPU

 

[Klemc]

that means the computer is not meant to be put in sleep state (in the BIOSses actual state) ?...

... the first time i had it more than 4 hours in sleep mode (overnight, while i slept) bluetooth totally disappeared when waken-up (yellow unknown in TaskManager) and i had to shut OFF PSU for more than 12>24 hours to see it back available, even updated from 1406 to 1409 between !!

 

[AusWolf]

that means the computer is not meant to be put in sleep state (in the BIOSses actual state) ?...

Not necessarily - the question, as I understand, is whether you're actually getting those insane VSoC numbers after sleep, or if it's just the motherboard/software reporting that's bugged.

 

[tpa-pr]

Checking in, Gigabyte dropped their F10d version BIOS for the X670E AORUS Master. And based on their (poorly translated) changelog i'd say this is the one that sets the SOC limits. It's still based on AGESA 1.0.0.6 mind you.

As always, no issues with the 7700x running pure stock. I keep watching the voltages in HWInfo but nothing ever changes or causes concern. I also don't use sleep, hibernate or Fast Startup. Perhaps my default settings have avoided the mess entirely?

(Also posting a PSA here because I noticed some panic on reddit and wasn't sure if this was common knowledge: if Gigabyte take down a beta BIOS it usually means they're just going to release a new version of it. There was a thread on reddit saying Gigabyte had taken all the latest betas down because they were "unsupported". Obviously I have no inside knowledge but it's a pattern Gigabyte have followed repeatedly that I noticed.)

 

[Nordic]

I am testing out a swap from a 5950x to a 5800x3d. Is it normal for the 5800x3d to not have PBO settings in the bios?

 

[3x0]

I am testing out a swap from a 5950x to a 5800x3d. Is it normal for the 5800x3d to not have PBO settings in the bios?

A lot of the motherboards have Curve Optimizer settings with AGESA 1207 and 1208. PPT/TDC/EDC settings only allow decreasing the values, not increasing. Make sure your BIOS is updated.

 

[AusWolf]

I am testing out a swap from a 5950x to a 5800x3d. Is it normal for the 5800x3d to not have PBO settings in the bios?

X3D chips don't officially support overclocking, so... I guess so.

 

[tabascosauz]

I am testing out a swap from a 5950x to a 5800x3d. Is it normal for the 5800x3d to not have PBO settings in the bios?

1207 doesn't change much from 1206 in functionality aside from reducing performance and fixing the fTPM bug. 1208 is where most vendors will un-hide their PBO menus again for 5800X3D. Yes, 5800X3D usually does not have access to any PBO settings pre-1208. No, you do not actually need anything from PBO other than curve optimizer.

 

[Mussels]

I feel like this is a real nightmare for AMD consumers - perhaps much worse than the memory compatibly issues in early AM4.
Ryzen X3D brand becomes Braisin 7800 X3D (as in the root word Braising - to fry (cpu) lightly and then stew it slowly in an enclosed heat spreader)

I saw a recent video der8auer has a new AM5 direct die cooling kit he is working on that seems really effective. I wonder with these AM5 issues if it becomes possible to still fry an AM5 CPU from these issues despite much better cooling. (Sorry for the umteen post edits. I shouldn't post so late a night)
View attachment 294537

sending 400W to the socket at boot would fry any CPU, AMD or intel.
The only remaining question is why the OCP and OVP didn't activate, and why this failure happend on multiple motherboard brands.

Was this an AGESA bug?
Is AMD's motherboard standards to blame?
Did the motherboard makers do this intentionally?
Do they share components from a company that lied about a components functionality?

Because most of the problems vary between the board makers, it does seem likely that AMD isn't the root cause - other than perhaps not being strict enough with board makers. Maybe they say X features must exist but can be disabled when expo or PBO is enabled, we may never know.

Is the bugged out 1.8V SOC actually probed though

probe by steve at GN, verified as 100% what happened

1.8v SoC and 400W to the CPU at boot showed SERIOUS problems with the board(s)

Checking in, Gigabyte dropped their F10d version BIOS for the X670E AORUS Master. And based on their (poorly translated) changelog i'd say this is the one that sets the SOC limits. It's still based on AGESA 1.0.0.6 mind you.

As always, no issues with the 7700x running pure stock. I keep watching the voltages in HWInfo but nothing ever changes or causes concern. I also don't use sleep, hibernate or Fast Startup. Perhaps my default settings have avoided the mess entirely?

(Also posting a PSA here because I noticed some panic on reddit and wasn't sure if this was common knowledge: if Gigabyte take down a beta BIOS it usually means they're just going to release a new version of it. There was a thread on reddit saying Gigabyte had taken all the latest betas down because they were "unsupported". Obviously I have no inside knowledge but it's a pattern Gigabyte have followed repeatedly that I noticed.)

Asus do this as well, with the final version being bumped up 1 (so beta 4401 becomes final 4402)

Not necessarily - the question, as I understand, is whether you're actually getting those insane VSoC numbers after sleep, or if it's just the motherboard/software reporting that's bugged.

hardware probed, software read 1.10v while probe said 1.8v

 

[Nordic]

1207 doesn't change much from 1206 in functionality aside from reducing performance and fixing the fTPM bug. 1208 is where most vendors will un-hide their PBO menus again for 5800X3D. Yes, 5800X3D usually does not have access to any PBO settings pre-1208. No, you do not actually need anything from PBO other than curve optimizer.

I am on the latest bios and don't have curve curve Optimizer either