Ryzen Owners Zen Garden

[tabascosauz]

Hey, one thing that I noticed since leaving B550 for X570..

My keyboard doesn't randomly disconnect on boots anymore..

Lol..

More platform hijinks?

I only have one X570 so I can't tell but keyboard disconnecting at login mostly was a Gigabyte B550 thing on my end

The keyboard not responding for first few seconds might be more a windows thing, have it on my laptop too.

 

[freeagent]

More platform hijinks?

I think this is all it was

 

[lexluthermiester]

Hey, one thing that I noticed since leaving B550 for X570..

My keyboard doesn't randomly disconnect on boots anymore..

Lol..

That's a good thing, right? LOL!

 

[Mussels]

More platform hijinks?

I only have one X570 so I can't tell but keyboard disconnecting at login mostly was a Gigabyte B550 thing on my end

The keyboard not responding for first few seconds might be more a windows thing, have it on my laptop too.

Usually that's dependant on what ports it's connected to, remember that four are from the CPU and others from the chipset.
Software can do it too, razer and corsair have done that to me with a clear handover/usb reset happening when all their RGB takes over - the xHCI handoff setting in the BIOS controls this, iirc.




Onto my rage filled desk flipping question: What the actual f*ck is this?

My motherboards description of PBO enabled vs PBO enabled, and "motherboard limits" from the BIOS

These images are *NOT* backwards.
PBO Disabled ----------------------- PBO on motherboard limits

This my ITX system with a 240mm EK AIO - and it's never hinted at thermal issues with PBO enabled or limited, but apparently "AMD's lower limits" means "more daddy more"

Do these board makers just make up whatever the feck they want?
No wonder people have nightmares getting their systems working, when something as simple as DISABLING PBO MAKES IT BOOST HARDER.

Auto means disabled, disabled means boost it, enabled means type a number?



These are my temps with a manual limit
It's a little different

 

[Klemc]

I learned one, and only one thing in my life about motherboard settings/OC, i don't do at all... i disable what i don't want (even some unused USB ports), set the fan by qfan, end of work.

 

[R-T-B]

So AGESA 1.0.0.7c seems to have landed GIGABYTE side. I tried the latest beta on my Aorus Elite X670, and vs 1.0.0.7b, it seems to have shaved a tiny bit off of the boot training time. Which rocks because I have to disable mem context restore due to a bug (can't disable mem powerdown otherwise).

Baby steps I guess.

 

[Mussels]

As much as i love my aorus b550 ITX board, the BIOS bugs are really pissing me off with it.
laggy interface if CSM is disabled (required for rebar)
doesnt work on some displays (it displays in 16:10 only)
PBO doesnt work as expected (Disable = boost harder)
Seeing comments like froggos about even more bugs in newer boards is annoying, I don't trust asus or gigabyte to get things right at this stage and primarily they've been the two most reliable board makers until now.

 

[tabascosauz]

As much as i love my aorus b550 ITX board, the BIOS bugs are really pissing me off with it.
laggy interface if CSM is disabled (required for rebar)
doesnt work on some displays (it displays in 16:10 only)
PBO doesnt work as expected (Disable = boost harder)
Seeing comments like froggos about even more bugs in newer boards is annoying, I don't trust asus or gigabyte to get things right at this stage and primarily they've been the two most reliable board makers until now.

Want a copy of my F13l beta BIOS? I presume you don't use your ITX board for your 5800X3D so 1203c wouldn't be an issue.

 

[Mussels]

It's fine as it is, since my custom PBO settings run cool

It's just annoying that they've known about these issues for years and have them on many many products and don't give a damn.

Google has a lot of hits from this forum thread over the years
GIGABYTE Latest Beta BIOS - TweakTown Forums

Updates for my Giga x370 board fixed a lot of issues like RAM compatibility and added CPU Support for the 5000 series, but brought these bugs back to it, as well. They're copy pasting known bugs to other chipsets and it's ridonculous. I've personally confirmed they exist on x570, B550 and x370, they've been around for 2+ years and... nothing. No attempts to fix them.

and people complaining about it still being around despite many many updates on their boards

Support tells you to clear the CMOS to fix it, which of course disables ReBar and doesnt actually fix anything - they dont support it except at BIOS defaults which explains why so many settings are outright broken

 

[Space Lynx]

what is power reporting deviation? is it ok that my numbers show red? my pc is fully stable and games run get, but this is only red number i see on hwinfo...

 

[tabascosauz]

what is power reporting deviation? is it ok that my numbers show red? my pc is fully stable and games run get, but this is only red number i see on hwinfo...

If 78% is just at idle then that's fine. If 78% is at full load then you have a bit of a problem. But for PRD to be accurate I think LLC has to be stock and no voltage offsets iirc.

tldr the further you are under 100% then the more the system underreports power/current = you are running more current and power than the CPU is reporting under Package Power. 95-105% is a fine margin, I only had one board that was deplorably bad (B450I Aorus = 74%), it did run hot and wouldn't be surprised if it contributed to my 3700X eventually degrading into a dumpster fire.

These days most AM4 boards should also have telemetry controls somewhere in BIOS, if you would like to play with those to re-adjust your deviation back towards normal 100%. Not sure where it is or what it's called in MSI BIOS.

IMPORTANT - Explaining the AMD Ryzen "Power Reporting Deviation" -metric in HWiNFO | HWiNFO Forum

 

[Space Lynx]

thanks Tabasco, I'm not too worried about it then, as my cpu never goes above 59 Celsius even in the most demanding of games. and my mobo temps always hover in the 40s

 

[tabascosauz]

thanks Tabasco, I'm not too worried about it then, as my cpu never goes above 59 Celsius even in the most demanding of games. and my mobo temps always hover in the 40s

If you just want to know, just watch the number while you run a loop of R23, will tell you roughly where your board stands

At all other times, the PRD metric is meaningless

 

[Space Lynx]

If you just want to know, just watch the number while you run a loop of R23, will tell you roughly where your board stands

At all other times, the PRD metric is meaningless

what should the percent be at when r23 runs?

 

[tabascosauz]

what should the percent be at when r23 runs?

as close to 100% as possible, anywhere between 95-105% is fine

 

[Mussels]

what is power reporting deviation? is it ok that my numbers show red? my pc is fully stable and games run get, but this is only red number i see on hwinfo...

View attachment 309057

Only usable at full load.

In theory at 100% load like with R23, you'd be at a perfect 100%.
If you're undervolted, it'll go up too.


It's so that you can judge a stock settings board for accuracy - many (MSI and asus especially) would lie and throw ~10% more power at a CPU to help them seem like better overclockers, but it would also make CPU's overheat badly
(Very similar to the SoC voltage crap AM5 had)

 

[Space Lynx]

as close to 100% as possible, anywhere between 95-105% is fine

i get 105 and 106%, 5 minutes in and sometikes it aill go to 107.3% before dropping down to 105%

seems to be staying at 105 and 106% aboit ewual amount of time so far.


so i guess i am good?gpod enough anyway . temps haven't broke 68 Celsius

Only usable at full load.

In theory at 100% load like with R23, you'd be at a perfect 100%.
If you're undervolted, it'll go up too.


It's so that you can judge a stock settings board for accuracy - many (MSI and asus especially) would lie and throw ~10% more power at a CPU to help them seem like better overclockers, but it would also make CPU's overheat badly
(Very similar to the SoC voltage crap AM5 had)

yeah i run co at -25 all core, so that probably is why is it is reading 105 to 107% on r23 and not 100%?

 

[tabascosauz]

i get 105 and 106%, 5 minutes in and sometikes it aill go to 107.3% before dropping down to 105%

seems to be staying at 105 and 106% aboit ewual amount of time so far.


so i guess i am good?gpod enough anyway . temps haven't broke 68 Celsius

yeah i run co at -25 all core, so that probably is why is it is reading 105 to 107% on r23 and not 100%?

I think the point was not to run a Vcore offset while you're measuring. But that was very initial info from 2020, so who knows if that even affects anything. CO isn't direct undervolting but changing the V-F curve. My Impact is around the 99% mark and doesn't change from CO. I'm pretty sure it's been the same even on 5900X and 5700G, both of which had pretty different looking CO settings.

A bit above 100% just goes in the opposite direction, potentially ever so slightly less performance/temps/power, but it's nothing to worry about.

 

[Space Lynx]

Thanks, and sorry for typos I am on my phone, which I am notoriously bad at typing on.

 

[Mussels]

yeah i run co at -25 all core, so that probably is why is it is reading 105 to 107% on r23 and not 100%?

Yep, and shows that -25 is roughly a 5% undervolt


Curve undervolt is VID vs voltage - it's lying to the CPU about what it's getting in a way.
CPU has hardcoded info inside "If voltage = X, Clock speed is Y"

But they can send that signal on one pin, and the power on others - so it thinks it's getting 1.10v while it's really getting 1.05v, or 1.15v depending on the goal.

Covering up poor VRM's or wanting to make a board seem like a good overclocker? give it higher than it thinks it has.
Undervolting to get higher clocks within a power limit, while risking stability? send the standard VID signal, but a lower voltage


When things like PBO are measured and used to stay within power limits, if you cheat and tell the CPU you're using 5% less than you are, you get a 5% performance lead over the honest competition
You also get CPUs overheating and angry customers, but they blame the CPU or the cooler, not the good guy motherboard just doin his job
(This is what asus did on AM5, where sensors would see 0.05v less on the soc rail than probes measured - they had a hardware offset putting more power in than the sensors/bios could see)

 

[cigano710]

Hey mussels.

Yep, and shows that -25 is roughly a 5% undervolt


Curve undervolt is VID vs voltage - it's lying to the CPU about what it's getting in a way.
CPU has hardcoded info inside "If voltage = X, Clock speed is Y"

But they can send that signal on one pin, and the power on others - so it thinks it's getting 1.10v while it's really getting 1.05v, or 1.15v depending on the goal.

Covering up poor VRM's or wanting to make a board seem like a good overclocker? give it higher than it thinks it has.
Undervolting to get higher clocks within a power limit, while risking stability? send the standard VID signal, but a lower voltage


When things like PBO are measured and used to stay within power limits, if you cheat and tell the CPU you're using 5% less than you are, you get a 5% performance lead over the honest competition
You also get CPUs overheating and angry customers, but they blame the CPU or the cooler, not the good guy motherboard just doin his job
(This is what asus did on AM5, where sensors would see 0.05v less on the soc rail than probes measured - they had a hardware offset putting more power in than the sensors/bios could see)

i think i found you on youtube comments about the USB Glitches
I'm lost here. Soyo (MB Manufacturer) told me to download driver updates, which i did, and told me it would be a fix. it doesnt work :/

 

[R-T-B]

Seeing comments like froggos about even more bugs in newer boards is annoying, I don't trust asus or gigabyte to get things right at this stage and primarily they've been the two most reliable board makers until now.

The bug I was referencing goes straight back to AGESA, ie every brand has it and you can't escape it.

Fun, yep.

 

[freeagent]

So far my DH has been pretty good, no USB shenanigan's. It is pretty aggressive with its boosting. It will push hard right to 90c and keep it there lol. With PBO2 I can feed 246/166/190 and it plows though Linpack at about 4550MHz. About 90MHz faster than my last B550.. not bad. I do like it better than my Strix XE.. would've been better if it was AM5

After Christmas..

 

[Space Lynx]

AMD Updates Ryzen Chipset Drivers with Greater 3D V-cache Optimization?

AMD released updated Ryzen Chipset software, version 5.08.02.027. In addition to bug fixes for its various components, the drivers introduce application support for a "New program support and some new drivers added." The package now includes three new components—AMD SFH 1.1 driver, AMD...

www.techpowerup.com

anyone want to test if these x3d chipset drivers improve performance? i call 4 fps gain across the board!

 

[tpa-pr]

Quick question about Ryzen Master: I've set up my Curve Optimizer to -25 all cores as I did in my BIOS. The problem with my X670E AORUS Master is every time I update the BIOS it resets all of my settings (and i've tried to restore them from a profile backup with no success) so I thought that if I set up CO in Ryzen Master I can just:
1) Update my BIOS.
2) Boot into Windows.
3) Re-apply my CO settings with a single click.

Am I right in understanding that Ryzen Master will allow for the above? I already have a list of BIOS settings to revert post-reboot and this would mean a few less clicks and taps.