Overclock.net banner
1 - 20 of 19439 Posts

· Registered
Joined
·
993 Posts
agesa 1.1.0.0 suppose to fix alot of bugs they introduced with these updates.

3900x, x570 master, only agesa 1.0.0.3 and 1.0.0.4 are usable for me. Anything after the voltages are complete garbage and not stable.

I can do 1800IF/3600 on auto voltages and 1900IF/3800 with vddp 920,vddg io 1029, vddg ccd 930 and 4066 async with agesa 1.0.0.4

I literally tested every bios after 1.0.0.4 and I can't even get stable with xmp 3200/1600IF.

Anyone tested voltages/IF with 1.1.0.0?
 

· Registered
Joined
·
993 Posts
Latest bios (F31) seems pretty stable on my zen2 (3700x)

IF1900

View attachment 2472493
Which version of F31 ?

Aorus x570 Master
I find F31x to be very stable for SOC, I even get 67ns with my 3900x at 1900IF , there's some weird issues where it randomly restarts or turns off.
While I did play with some bios settings to mitigate these, they eventually creep up again.


I'm back to F11, where 1800IF is 24/7 stable on auto voltages.
1900IF custom voltages is on the edge of stability. It passes P95,Memtest,realbench but games will BSOD if I increase SOC voltages.
 

· Registered
Joined
·
993 Posts
For those of us that are still rocking AMD 3000 series processors such as a 3600, 3700X, 3800X, 3900X, 3950X, etc on gigabyte x570 motherboards .......... Does newer BIOS releases such as F31 even offer anything for us at this point? Is there any reason to upgrade to F31 at this point if there is no AMD 5000 series processor in the equation?
I tried all f30 f31, not this new one posted recently because it's not on the gb site.

with my 3900x, I do get better score at stock,pbo and soc on auto supports 1900IF. But all of that is not worth the absolute garbage stability.

It just randomly restarts, bsod and turns off. With being so random you can't validate / test any bios changes.

With F11 it's rock solid even when running pbo,edc bug or 1800IF. Once I have Prime95,memtest,realbench,games passing 8h+ nothing will crash the system.

On f31 to me it seems the SMU, various voltages are borked and Gigabyte won't fix it. Just like prev f30.
 

· Registered
Joined
·
993 Posts
With latest F31 you can add up to 500Mhz boost, with 5900x normal EDC, PPT and TDC I got same results as before - rare 5150Mhz spikes and usual 5050Mhz. But if you put in bios EDC value 5, then I got 5298Mhz single core and quite stable 5221 all single CPU CB20 test. Sadly effective clock was low and result was only 515, but mybe somebody figure out what to change in other settings and got with EDC bug better results.
View attachment 2473014
Usually you run CB20 ST normal or PBO , and see what current you get for EDC.
for EDC bug you would choose values +1 or -1 of that value.

For my 3900x it's around EDC 16. Now for R20 ST scores under EDC, you will register lower because it's just single thread and you'll hover around that stock EDC for ST which might cause some throttling.

For games / other apps, this throttling doesn't kick in because other cores are also boosting, and you'll be higher on that EDC value.

I haven't run EDC bug for a while, it does add to the memory latency 1ns or 1.5ns worst under Aida (with already slow memory that's a big loss) and screws with AVX voltages. I know I needed a big +vcore offset to get it stable on Prime / [email protected]

Also because the CPU is not regulated for boost, you'll run into stability issues if you don't limit with PPT / TDC under high current workloads.
 

· Registered
Joined
·
993 Posts
You should have been able to do 1900 on 3000 a long time ago error free. I bought my 3900x a few months after launch and it did it on a Gigabyte X570 pro, then upgraded to a 3950x on a Asus X570i Strix and did 1900. My daughter Ryzen 5 3500 can do 1900. Out of all my ryzen 3000 chips only a 3700X could not do 1900. So maybe yours can't if you have not been able to do in the past. Now the question is can ryzen 3000 hit 2000 since it is the same mem controller according to AMD, is amd artificially limiting ryzen 3000 not to hit 1900+??
I don't think amd is limiting anything, ryzen 3000 still needs IF between cores inside the ccd, where 5000 it's a 8core complex no IF at play there.

There's a few that do 1933/1966 with ryzen 3000. I can do 1913IF with my 3900x but not 24/7 stable.
Running mine around 1905 so I don't get audio issues, crashes.
 

· Registered
Joined
·
993 Posts
Hello,
The same for me, I run under F11 bios about same config 3700x, gigabyte aorus elite x570, ddr 4 16 gb (Hynix, Patriot 3733), gtx 1080, win 10, NVME
no issue no WHEA, with IF 1900., (aida 64 Read 54500, Write 30400,Copy 49800, latency 67,5....very stable.)

Of course I monitor the benefits (or not) to switch to more recent bios... but I am afraid to get something worse.
I can tell you from experience that if you get IF 1900 working on F11 , no amount of tweaks will get it stable after F11.
For Zen2, after F11 the bios is complete trash / worthless code !

I'm 1904IF (small bclk) and I'm not switching from F11 with my 3900x.
 

· Registered
Joined
·
993 Posts
View attachment 2483011

Currently, Asus dark hero motherboard has a lot of successful cases without WHEA error in memory 4000mhz.

I have seen on GigabyteBoard as well, but which is very rare.

Just looking at this, you can see that WHEA has a relationship with the motherboard (especially the BIOS).

I think the GIGABYTE doesn't make the bios for the Zen 3 properly. (maybe it doesn't have a competent staff. 😢)
I never had whea on my x570 master board with the 3900x. What's the upper limit for IF on Zen3 ?
Currently running 100.34bclk with 1900IF, auto voltages(VDDG,VDDP).
2483017
 

· Registered
Joined
·
993 Posts
Have you guys noticed how much the SOC Power Consumption in Idle increases when enabling XMP ?

With XMP off I get 3W SOC Power and with XMP on I get 22W SOC Power in Idle. This is with 1.2V Vsoc vs 0.975 Vsoc. I also tried with 1.1V Vsoc and it gets just a little under 21W.

This contributes to a 4dgr C idle temp increase on the CPU.

Mathematically, if the voltage is 1.2 times increased and the frequency let's say it's 1.8 times increased, I would have expected a 2x power consumption increase, not a 7x.

Ryzen Master screenshots in Idle, same settings except XMP off / on:

View attachment 2483893

View attachment 2483894
I have 11w soc idle with 1900IF, 1.2v . There's a bios option that forces IF/ soc to always run full power.
It will set to ON, when you use xmp, just set it off.
 

· Registered
Joined
·
993 Posts
That's funny you have to run a whea suppressor utility lol

It's somewhere in one of these "event trace sessions", might need to change permission on the registry to remove that whea trace. *Too lazy to boot into my other windows copy (no changes to the logging)

I usually stop all of them, just pages of non-sense logging / traces wasting cpu cycles.

2490602
 

· Registered
Joined
·
993 Posts
With Prime95, do LargeFFT no avx with all cores. You'll boost higher and it'll wreck the shiat out of your IMC and fabric.

Close all programs before you do LFFT though, it will consume all of your memory (by design).
Prime large ffts loads falls more toward mem voltage, primary mem timings, imc voltage, heat, etc

IMC load will be higher with proper memtest and IF load will be higher under mixed traffic with load cycle (gpu pcie traffic + avx load) like asus realbench
 

· Registered
Joined
·
993 Posts
anyone know of a good methodology to find out which pbo scalar setting give the best performance for my chip? where should i look at in hwinfo when stress testing ?
My understanding of the scallar is that the higher the number the higher it will hold voltage during boost.

On my 3900x, R20 scales very little, about x4.
Games scale a bit better(non-avx load).
At x1 I can have 4.2 all core, x7 seem to have up to 4.4 for games.

Benching that will be hard, depends on temps and type of workload. Even from 4.2 to 4.4 all core for games you won't find a fps diff on Zen2.

My 3900x also likes high voltage, early agesa was 1.55v on auto with 4.7 max boost vs 1.5v with new agesa 4.5 max boost
 

· Registered
Joined
·
993 Posts
Just upgraded to an X570 Aorus Master + 5900X this evening, and I'm already about to throw it in a dumpster after about 2 hours.

First off, it will randomly go to a black screen when sitting in the BIOS (F32 BIOS on a rev 1.2 board if it matters) and require a reboot. WHY???

Second, how do you get Windows 11 installed on this disaster of a motherboard/platform? I had downloaded the latest Windows 10 dev insider ISO (why they don't post the Win 11 ISO on the insider page, lord only knows) and had no issues installing it. But the second I opened Windows update, my computer froze. Upon rebooting, I was greeted with repeated "INACCESSIBLE BOOT DEVICE" green screens. Apparently some sort of AMD NVMe driver issue (driver issues with AMD... WHAT A SHOCKER)?

Currently on a normal release copy of Windows 10 and praying that it installs all the latest updates without crashing into a brick wall. Really wishing I bought a 10850k like a sane person right about now...
welcome to AMD :) where you spend most of the time tweaking and troubleshooting

at least it's been like this for me with various cpus,mobos since Zen started
 

· Registered
Joined
·
993 Posts
I've a Gigabyte Aorus Master Rev 1.2 and I've noticed that the chipset is getting quite hot.
There are two readings in hwinfo for the temperature.

When gaming say red dead redemption 2 I get 75C on one and 90C on the other temperature reading. (I guess one is a sensor in the chipset die and another is just placed around it)
While my VRM temps get to around 62C.

I was reading that the board does use a thermal pad on the chipset instead of thermal paste.

Would you guys recommend swapping that out to proper thermal paste?

Also there are thermal pads on the back of the board too, should I upgrade those too maybe?

That's my setup:

View attachment 2518061
The chipset pad quality is horrible. Paste on the chipset will get tops 55c under load with no fan spinning.

Pads on the back are not worth changing, not much heat for the back of the vrm.
The only reason to chnage them is if you don't want silicon oil, they do leak like crazy.
 

· Registered
Joined
·
993 Posts
Alright, thank you for describing it in detail.

Does the silicon oil have any negative effect? Also if I would replace them, which thickness is it?

Btw. are the m2 pads of the same bad quality? If so, do you know a recommended thickness ?
Silicon oil, no negative effects. I don't have the thickness, I didn't replace mine.

M2 pads are fine, M2 even PCIE4 don't put out high heat to warrant premium pads. All my 3 heatsinks on the Aorus Master are very hot to the touch.
 
1 - 20 of 19439 Posts
Top