Ryzen Owners Zen Garden

[Mussels]

Requires local access - so with an antivirus, we're fine.
Edit: and requires the process to be locked to a specific corem, with the target of the attack locked on that cores SMT thread

So uh, we're fine.

Oh no my RSA keys

This might be an issue for servers and mission critical systems, but 99% of people it's meaningless

 

[Chomiq]

Looks like the latest chipset drivers (4.06.10.651) somehow affected CO stability. It's the second time within a week that I ran into a WHEA-18 error with processor core fail when basically idling with my so far rock solid -15 all core:

Disabling CO switching to -10 for now, going to see how it behaves.

 

[dont whant to set it"']

Wow!, that stock auto vcore with new firmware. Dialed 40x , core performance boost to off, vcore set to 1.1V, SMT off and quickly save settings.
I will edge out previous settings back in the ballpark throughout the day.

 

[Chomiq]

Wow!, that stock auto vcore with new firmware. Dialed 40x , core performance boost to off, vcore set to 1.1V, SMT off and quickly save settings.
I will edge out previous settings back in the ballpark throughout the day.

Memory at 2133?

 

[dont whant to set it"']

Yes, I initiated the motherboard EZ-Flash( or what is called) and went out for a quick fag.
When I returned and observed the vcore value on , I just set the previous posted setting and let it run into Windows where I ran an Aida cache/mem bench at stock setting(well, all the other settings).
And yes , as it still runs the cpu-z stress test while writing this , my rig is using ~106Watts from wall.
Le: jedec SPD.

 

[INSTG8R]

Looks like the latest chipset drivers (4.06.10.651) somehow affected CO stability. It's the second time within a week that I ran into a WHEA-18 error with processor core fail when basically idling with my so far rock solid -15 all core:
View attachment 257859
Disabling CO switching to -10 for now, going to see how it behaves.

Okay well I just installed them(I was in Canada for June and most of July) I run my CO at -18? and have my PB0 running 4850 all core. My PC is on 24/7 so this might get exciting....

 

[Chomiq]

Okay well I just installed them(I was in Canada for June and most of July) I run my CO at -18? and have my PB0 running 4850 all core. My PC is on 24/7 so this might get exciting....

I rolled back to 4.03.something from march and returned to -15 on CO. Maybe newer drivers didn't like the old agesa 1.2.0.3 B (waiting for some non-beta bios with 1.2.0.7+).

 

[INSTG8R]

I rolled back to 4.03.something from march and returned to -15 on CO. Maybe newer drivers didn't like the old agesa 1.2.0.3 B (waiting for some non-beta bios with 1.2.0.7+).

Yeah I’m on the Beta BIOS as well but I’ve been using the same BIOS settings pretty much since I finished the build and found my ideal settings.
Will be interesting to see if my 100% stable settings that have never failed me on every BIOS released so far might actually require some tweaking…fingers crossed…

 

[P4-630]

Ryzen 7000 isn't ready yet.

Ryzen-7000-Verschiebung: AGESA-Version scheint noch nicht gereift zu sein - Hardwareluxx

Als Grund für die mögliche Verschiebung des Ryzen-7000-Launches könnte das noch nicht ausgereifte AGESA sein.

www.hardwareluxx.de

 

[outpt]

that's great can't get it out of the gate or is amd/intel going to play chicken at this point.

 

[A Computer Guy]

Well there's always that rumor about zen4 + old io chip for AM4

I would be interesting if AMD brought back Overdrive CPU's. Imagine plugging in 7000 series chip in a 6 yr old AM4 platform.

 

[Chomiq]

I rolled back to 4.03.something from march and returned to -15 on CO. Maybe newer drivers didn't like the old agesa 1.2.0.3 B (waiting for some non-beta bios with 1.2.0.7+).

Well it turned out to be CO instability. Opted for -10 on all but gold and silver cores which I set to -15. Updated to 1.2.0.7 and latest chipset drivers. It ain't that easy to troubleshoot mainly because it happens at super low load instead of gaming and if it did it happened maybe 1 time over the course of entire week.

 

[The King]

Ryzen 5 5600 5.8GHz!!!

R

 

[Mussels]

125MHz BCLK makes many things possible

and unstable

 

[RoyZ]

When you are setting a OC on ryzen master, can you set an higher clock on the "better cores" and leave the rest on a lower all core setting? Lets say, I have a R5 3600, and ryzen master shows me 2 out of my 6 cores as my "better cores". Can you set lets say 4.3 on the two better cores and 4.2 on the rest as a permanent OC? Is this a thing?

And if so, when your PC is doing single core task, will it always go for these "better" cores that you OCed a little higher?

 

[Mussels]

When you are setting a OC on ryzen master, can you set an higher clock on the "better cores" and leave the rest on a lower all core setting? Lets say, I have a R5 3600, and ryzen master shows me 2 out of my 6 cores as my "better cores". Can you set lets say 4.3 on the two better cores and 4.2 on the rest as a permanent OC? Is this a thing?

And if so, when your PC is doing single core task, will it always go for these "better" cores that you OCed a little higher?

Windows 10 and 11 are aware of the preferred cores and use them first, yes.

You cant set per core clocks manually easily, that's what PBO does for you - the closest you'll get is the curve optimiser where it auto adjusts the voltages for you to maximise things for your particular setup

 

[tabascosauz]

When you are setting a OC on ryzen master, can you set an higher clock on the "better cores" and leave the rest on a lower all core setting? Lets say, I have a R5 3600, and ryzen master shows me 2 out of my 6 cores as my "better cores". Can you set lets say 4.3 on the two better cores and 4.2 on the rest as a permanent OC? Is this a thing?

And if so, when your PC is doing single core task, will it always go for these "better" cores that you OCed a little higher?

On Ryzen 3000 you don't have access to Curve Optimizer, so you can't vary clocks by core. You can still set per-CCX OC if your CPU is up for it. But that's still in fixed groups of 4 cores.

The "better" cores relies partly on AMD's CPPC hierarchy (unique order programmed into every CPU) and partly on the Windows scheduler. Generally all Ryzen CPUs will rank all their cores, but only 2 cores are singled out for single thread tasks (highest ranked 2 cores in CPPC).

The core preference only follows CPPC/Windows order. It doesn't matter if you single out two better cores - unless those two cores also happen to be the top 2 cores in CPPC/Windows, they won't get used. You can disable CPPC Preferred Cores in BIOS, but in my experience it changes little - Windows still maintains its own core quality hierarchy.

On my 5900X, Cores 0 and 1 are top rated, but after Curve Optimizer Core 2 by far outstrips every other core in clock. Unfortunately, being rated #3, it will never be used for ST work regardless of how much better it is.

Perhaps at some point in the future we will be able to edit CPPC or Windows core rankings.

 

[Mussels]

Perhaps at some point in the future we will be able to edit CPPC or Windows core rankings.

I swear i've seen the ability to mess with that mentioned somewhere, it might have been linux or zen master



Chipset driver alters how CPCC works, so make sure that's installed? Ryzen 3000 needed the power plan on balanced/ryzen balanced to work perfectly, 5000 seems fine on whatever

 

[Chomiq]

Buldzoid on 7950x's power/perf scaling.

 

[tabascosauz]

Confirming my worst fears - this X3D's Fabric can't do 1900, as I thought there's no reason the board can't train 3800 Bdie when it's done it every day for a year

Which makes it in that respect the worst CPU I've had, even worse than the 3700X

Heralding in a new age of Ryzen I guess where AMD wants us to forget about Fabric. Whatever, I'm not returning this CPU just for that, it works great

time to go sell the 5900X

 

[QuietBob]

Confirming my worst fears - this X3D's Fabric can't do 1900, as I thought there's no reason the board can't train 3800 Bdie when it's done it every day for a year

Which makes it in that respect the worst CPU I've had, even worse than the 3700X

Heralding in a new age of Ryzen I guess where AMD wants us to forget about Fabric. Whatever, I'm not returning this CPU just for that, it works great

time to go sell the 5900X

View attachment 264025 View attachment 264027

What about single rank?

 

[tabascosauz]

What about single rank?

Only free set I have lying around are the 4000CL19 Rev.E - too much a hassle to swap RAM on my daily because of the dual-fan C14S. I haven't used the Ballistix in anything since I got the 5700G RMA done, too slow compared to B-die

The RAM works fine. I can set my usual 3800CL14 profile as long as I leave FCLK on auto. Fabric can't keep up.

 

[freeagent]

I may consider selling my 5900X too lol.. the IHS took a beating with probably a hundred mounts lol.. maybe not that much.. probably won’t be able to get much more than 400 for it, half of what I paid

X3D idles so nice at the desktop, like 60w with the lights off





The backside of the case is getting low on cable space.. good thing you can use a thumbscrew to keep the panel from popping off

It’s loud when it does

 

[RoyZ]

On Ryzen 3000 you don't have access to Curve Optimizer, so you can't vary clocks by core. You can still set per-CCX OC if your CPU is up for it. But that's still in fixed groups of 4 cores.

The "better" cores relies partly on AMD's CPPC hierarchy (unique order programmed into every CPU) and partly on the Windows scheduler. Generally all Ryzen CPUs will rank all their cores, but only 2 cores are singled out for single thread tasks (highest ranked 2 cores in CPPC).

The core preference only follows CPPC/Windows order. It doesn't matter if you single out two better cores - unless those two cores also happen to be the top 2 cores in CPPC/Windows, they won't get used. You can disable CPPC Preferred Cores in BIOS, but in my experience it changes little - Windows still maintains its own core quality hierarchy.

On my 5900X, Cores 0 and 1 are top rated, but after Curve Optimizer Core 2 by far outstrips every other core in clock. Unfortunately, being rated #3, it will never be used for ST work regardless of how much better it is.

Perhaps at some point in the future we will be able to edit CPPC or Windows core rankings.

I see. I'll just keep doing the "all cores" OCs then on my 3600.
I'm using an all core at 4.2 and 1.2 volt right now. I'm gonna try 4.3.

What happens if you go too far for your CPU? The PC shuts down? Can you perma damage it even if you are just messing with core clocks and voltage in ryzen master?
I'm new to this so I hesitate.

 

[tabascosauz]

I see. I'll just keep doing the "all cores" OCs then on my 3600.
I'm using an all core at 4.2 and 1.2 volt right now. I'm gonna try 4.3.

What happens if you go too far for your CPU? The PC shuts down? Can you perma damage it even if you are just messing with core clocks and voltage in ryzen master?
I'm new to this so I hesitate.

Ryzen Master does not guarantee safety, that program is perenially buggy as shit. Not even PBO is always, 100% safe long term, ie. the Zen 2 EDC trick.

If you're set on doing all-core, just try to stay below say 1.25V and keep thermals under control. Ryzen degrades primarily with current (ie. power), so don't be benching all-core 24/7. Higher temps and/or higher volts accelerate that process.

Can't really tell you if or when you'll see degradation. You'll get all sorts of answers depending on who yku ask.

You may find some benefits to running all-core (e.g. better frame consistency), but I wouldn't worry about pushing a 3600 further. Clocks don't commonly help Ryzen much outside of MT work, it's the arch improvements in each generation that do most of the heavy lifting.

The shutdowns should only really come when you're running unstable. Stock operation has thermal protections, but when you all-core OC there are no more protections