Ryzen Owners Zen Garden

[Xzibit]

not out for mine, whats new?

reddit has a lot of angry people
"Sadly it's another MSI scam. It's just good old plain 1.2.0.3a NOT b!

MSI: 'look at us pushing 'updates'!'"

there is fun arguing and weirdness here
(1) AGESA1.2.0.3b BIOS for X570/B550 boards : MSI_Gaming (reddit.com)

Maybe just bitter. If he/they even bothered checking MSI never released 3a for some of those boards but hey its reddit what do you expect.

Also that example of "Look at us pushing updates" hes comparing same agesa update within the same beta iteration, did he expect it to change. The bios naming gave that away

 

[Space Lynx]

zen 5950x back in stock on amd.com official store. just an fyi. msrp. 799

 

[Xzibit]

Just flashed mine and I guess that guy got a lemon because mine is showing 1.2.0.3b in CPU-Z & HWinfo as well

 

[Mussels]

Just flashed mine and I guess that guy got a lemon because mine is showing 1.2.0.3b in CPU-Z & HWinfo as well

conspiracy guy said you could open the files up, and it would show as a

I dunno wha the real deal is, just what came up first when i googled it

 

[Xzibit]

conspiracy guy said you could open the files up, and it would show as a

I dunno wha the real deal is, just what came up first when i googled it

He post his CPU-Z and HWinfo screens with 3a later on in the conversation for proof. So I checked mine and it says 3b.

 

[tabascosauz]

A better visualization of what exhaustively testing with Corecycler looks like. The new per-core HWInfo temp sensors are a great help.

Was testing Core 0 (always a problem child) and Core 10 (only remaining untested core) last night. As before, 68 minutes each per iteration - 5 iterations - All FFTs.



Core 0: 2/5 failed @ 128K, 10752K
Core 10: 5/5 failed @ 2240K, 128K, 720K, 720K, 12288K

Core 10 isn't a problem; I had just set -30 for shits and giggles and because the next two lower ranked cores (#8 and #9) both did -30 stable. I'll just step down to -20 and test again tonight. There's no general pattern or correlation between CPPC/power/clock and stable offset - each core needs to be tested independently.

Core 0 (being the problem child), however, demonstrates the sort of inconsistent instability I've been talking about. If you had only run 2 iterations of the script, even at 68min All FFTs, there's a chance that it may have come back "passed" both times, giving you the impression that the current offset is stable.


I kinda wish AMD would just up and give us per-core OC. CO is nice, but often actually makes cores run hotter than they did stock - counterintuitive, yes, but anecdotally just higher freq even at the same Vcore can run hotter, and I see some cores draw more power and more VID the more aggressive the offset. For some of the unused cores (thanks CPPC) I'd rather use the -30 headroom to make it run cooler at the same clocks, than boost harder and never be used.

Now, back to testing.

 

[Mussels]

Oh i like how that works now. i didnt pay attention when it was new, as it seemed... unfinished

 

[sam_86314]

Swapped out the 2600X in my main rig for a 5800X

Had issues with temperatures at first, but thanks to @Mussels, I've managed to not only completely fix my temps, but I'm seeing sustained clocks of over 5GHz in games.

I've been playing a lot of heavily modded Minecraft (341 mods in total), and I've noticed a dramatic improvement to performance in the game since upgrading. Framerates are not only noticeably higher, but they're also noticeably more consistent too.

My only real concern is voltage now. Even with a -0.1125v offset, Ryzen Master says up to 1.479v is being fed to the chip during CB runs, and CPU-Z reports up to 1.37v while gaming and benchmarking. Seems a bit high.

 

[Mussels]

Swapped out the 2600X in my main rig for a 5800X

Had issues with temperatures at first, but thanks to @Mussels, I've managed to not only completely fix my temps, but I'm seeing sustained clocks of over 5GHz in games.

I've been playing a lot of heavily modded Minecraft (341 mods in total), and I've noticed a dramatic improvement to performance in the game since upgrading. Framerates are not only noticeably higher, but they're also noticeably more consistent too.

My only real concern is voltage now. Even with a -0.1125v offset, Ryzen Master says up to 1.479v is being fed to the chip during CB runs, and CPU-Z reports up to 1.37v while gaming and benchmarking. Seems a bit high.

use curve optimiser and not the offset, the offset can hurt clocks while the curve wont

 

[sam_86314]

use curve optimiser and not the offset, the offset can hurt clocks while the curve wont

Is there a generally agreed-upon curve to build off of?

 

[Mussels]

Is there a generally agreed-upon curve to build off of?

start with -10

I'm at -14 or so with +200MHz, and stable - but i may be lucky

My first two cores can do like -30, but i updated the BIOS and got too lazy for per-core again

 

[tabascosauz]

Is there a generally agreed-upon curve to build off of?

Every chip is different on quality. For the 6- and 8-cores it's easier to get away with a single offset across the board (e.g. -10, -15, even -30) but not guaranteed.

Best to go with the Corecycler script at its default settings, go slow say with -5 all cores at first, then note where specific cores start falling off as you start uniformly increasing the offset.

The bigger chips behave and respond a little differently to Curve Optimizer, but here's an idea of the kind of spread you could conceivably be looking at:



The easy mistake to make would be to have a chip that looks like CCD1 on mine, test it in corecycler at say -10 all cores, have it throw errors, and come to the premature conclusion that the entire CPU is bad because it "can't do -10". You get the idea.

almost forgot: GitHub - sp00n/corecycler: Stability test script for PBO & Curve Optimizer stability testing on AMD Ryzen processors

And congrats on the big upgrade!

 

[sam_86314]

Gonna try just setting everything to -10 and seeing what happens.

For other 5800X owners, what voltages do y'all see under various loads (CB, gaming, Prime95, etc.)?

EDIT: Got a BSOD after a CB23 run followed by a CB20 run. Gonna start backing off I guess.

 

[GoldenX]

The progress on mine, seems rock solid so far.
Remember to set LLC to the ideal level for your board, PBO to motherboard limits, and play with AutoOC depending of your CPU sample, mine is at +200MHz.
Use a known stable memory profile, else it will be hell to stability test.

 

[sam_86314]

For the time being, I've given up on CO due to instability. I still need to run Memtest for a while to see if my memory is stable at its rated speed (it wasn't on my old CPU).

Gonna keep running a negative offset for now. Also really don't know how concerned I should be about the voltages I'm getting.

On another positive note, I'm noticing performance improvements in even more real-world scenarios. When playing BotW in Cemu, I'm now GPU bottlenecked. That means I get a steady 70 FPS at 3440x1440. My old CPU was the bottleneck in that scenario, presumably due to the poorer single-threaded performance, which resulted in less consistent performance and slightly lower framerates.

Overall, I am very happy with this upgrade so far.

 

[freeagent]

Woke up this morning to find my cpu running @ 90c. I have been folding for a few days and last night I decided to try out my Le Grand Macho RT and TY-143 combo on my 5900X. Prior that combo had been excellent on my last 4 CPUs. Turns out it wasn’t as good as my new cooler like I saw with my last 2 CPUs. I think the board inside the fan went, it should have been running at 2500rpm like it was last night.. but could not get more than 650rpm out of it. Funny that same thing happened to my TY-147A last winter.. left it at 1400rpm woke up to it at 250rpm. I really like their cooler, and their fans perform well but I had 2 die in 4 years.. guess which guy is looking at IPPC fans now

Edit:

Sorry the fans were 3 years old, I got them in 2018.. I have a TY-147B left on my z77 rig.. tick tock..

 

[sam_86314]

Ran Memtest86+ for about eight hours last night with no errors. Guess my memory instability at its rated speed was just a limitation of Zen+.

My PC is now perfectly fine with my memory at 3600MHz C18.

 

[Space Lynx]

Ran Memtest86+ for about eight hours last night with no errors. Guess my memory instability at its rated speed was just a limitation of Zen+.

My PC is now perfectly fine with my memory at 3600MHz C18.

yep the latest Zen always seems to run memory better. glad it worked out for you

 

[sam_86314]

Had a BSOD while playing BeamNG.drive and backing up a DVD at the same time with my voltage offset at -0.1185v and PBO set to 200MHz. Backed off both settings a bit to see if stability improves.

Also recently overhauled the cooling in my PC which I'll post about in the near future.

Actually just did some reading on how curve optimizer and stuff works, so I think I'm gonna start over with all of this, especially since I have much higher quality fans in my system now.

EDIT: Even with CO set to -5 on all cores and no voltage offset, I got a random crash while watching YouTube. Now at -3 on all cores with adjusted PBO settings.

I'm thinking I'll take the same approach I did with my old CPU if it keeps acting up; no PBO and minimal voltage adjustments.

 

[Mussels]

Had a BSOD while playing BeamNG.drive and backing up a DVD at the same time with my voltage offset at -0.1185v and PBO set to 200MHz. Backed off both settings a bit to see if stability improves.

Also recently overhauled the cooling in my PC which I'll post about in the near future.

Actually just did some reading on how curve optimizer and stuff works, so I think I'm gonna start over with all of this, especially since I have much higher quality fans in my system now.

EDIT: Even with CO set to -5 on all cores and no voltage offset, I got a random crash while watching YouTube. Now at -3 on all cores with adjusted PBO settings.

I'm thinking I'll take the same approach I did with my old CPU if it keeps acting up; no PBO and minimal voltage adjustments.

Use the tweaked PBO values in the thread i made, without messing with voltages (and if you have a truly poop CPU, without the +200)

(17) 5800x (and other Zen 3 chips) PBO settings/Temperature fix | TechPowerUp Forums

 

[sam_86314]

Use the tweaked PBO values in the thread i made, without messing with voltages (and if you have a truly poop CPU, without the +200)

(17) 5800x (and other Zen 3 chips) PBO settings/Temperature fix | TechPowerUp Forums

Been running 90/60/95 -0.1000v offset and no PBO override for a few hours now. 4850MHz on single-core loads, so far no crashes. Mid 60s to low 70s under gaming loads (dependent on ambient temp).

EDIT: Just switched to 105/70/95 with the offset set to "negative auto" (which I remember yielded good results on my old i7-2600K when overclocking) to see what happens. Currently idling around 35C to 40C with an ambient temp of 22C.

EDIT 2: Synthetic benchmarks with these settings are pretty much the same as when I was running CO and +200MHz. Seems that the extra 200MHz doesn't do anything for them. Gonna stick with these settings I think. Core voltage is considerably lower than it was before. I was seeing less than 1.3v under a single-core load.

EDIT the third: Forget what I said about temperatures. They're still pretty high under certain loads, though still better than stock settings. Interestingly my CPU got hotter while playing heavily modded Minecraft (mid-to-high 70s) than it did while running Cinebench R23 (low-to-mid 60s). Voltages went up to 1.45v, though I've read and am beginning to accept that those voltage spikes are normal.

 

[outpt]

any idea when this will be avalible?

 

[Asryan]

Hello anyone here using a dark hero motherboard with a 5900x who would be able to tell me the whole settings for setting up the oc with dynamic oc switcher? Thanks

 

[tabascosauz]

Been running 90/60/95 -0.1000v offset and no PBO override for a few hours now. 4850MHz on single-core loads, so far no crashes. Mid 60s to low 70s under gaming loads (dependent on ambient temp).

EDIT: Just switched to 105/70/95 with the offset set to "negative auto" (which I remember yielded good results on my old i7-2600K when overclocking) to see what happens. Currently idling around 35C to 40C with an ambient temp of 22C.

EDIT 2: Synthetic benchmarks with these settings are pretty much the same as when I was running CO and +200MHz. Seems that the extra 200MHz doesn't do anything for them. Gonna stick with these settings I think. Core voltage is considerably lower than it was before. I was seeing less than 1.3v under a single-core load.

EDIT the third: Forget what I said about temperatures. They're still pretty high under certain loads, though still better than stock settings. Interestingly my CPU got hotter while playing heavily modded Minecraft (mid-to-high 70s) than it did while running Cinebench R23 (low-to-mid 60s). Voltages went up to 1.45v, though I've read and am beginning to accept that those voltage spikes are normal.

Up to 1.5V has always been allowed to sustain high single core clocks, but Ryzen Master is basically the only software that can accurately read single core boost Vcore. The +200MHz override only works if your boosting cores are easily able to exceed 4850MHz - it only tweaks the absolute Frequency Limit (in HWInfo) to set a theoretically higher boost ceiling, other facts like core quality can intervene before then to limit boost.

For undervolt settings you need to be testing your curve settings thoroughly through CoreCycler or OCCT v8.0 which now has a similar function. Point is to push boost clocks as high as they can possibly go to verify stability under best-case boost. Obviously a certain level of stability is *probably* good enough for most users, but you never know where you stand until you test.

Interesting you should mention Minecraft, I stopped running Optifine+shaders because of just that. Vanilla MC sails at a breezy 165fps at all times with CPU at like 60C and GPU at basically idle. MC with any shaderpack struggles to reach 100fps in busy scenes, with the 2 preferred CPU cores balls to the wall at 75C and GPU at 100%. Optifine on its own is still fine, though. I bet the shaders implementation is just optimized like garbage, there's no combination of graphics settings that can be tweaked to regain the loss in performance from even lowest quality shaders.

 

[sam_86314]

Up to 1.5V has always been allowed to sustain high single core clocks, but Ryzen Master is basically the only software that can accurately read single core boost Vcore. The +200MHz override only works if your boosting cores are easily able to exceed 4850MHz - it only tweaks the absolute Frequency Limit (in HWInfo) to set a theoretically higher boost ceiling, other facts like core quality can intervene before then to limit boost.

For undervolt settings you need to be testing your curve settings thoroughly through CoreCycler or OCCT v8.0 which now has a similar function. Point is to push boost clocks as high as they can possibly go to verify stability under best-case boost. Obviously a certain level of stability is *probably* good enough for most users, but you never know where you stand until you test.

Interesting you should mention Minecraft, I stopped running Optifine+shaders because of just that. Vanilla MC sails at a breezy 165fps at all times with CPU at like 60C and GPU at basically idle. MC with any shaderpack struggles to reach 100fps in busy scenes, with the 2 preferred CPU cores balls to the wall at 75C and GPU at 100%. Optifine on its own is still fine, though. I bet the shaders implementation is just optimized like garbage, there's no combination of graphics settings that can be tweaked to regain the loss in performance from even lowest quality shaders.

So, in theory, the +200MHz shouldn't affect stability since it just lets PBO go higher if conditions are right?

Your performance in Minecraft with shaders; is that on your 4650G and its iGPU?

As I said a few posts ago, I'm running the "All The Mods 3" mod pack with a couple of extra mods that I like thrown in (341 mods in total). I can sustain my FPS limit of 144 pretty much all the time now, but my CPU temps are higher than synthetic loads.

I haven't done any real tests in the vanilla game yet.