[St0RM53]

Hey @GBT-MatthewH
x570 aorus master owner with 3800x here


Definitely some issues with the 1.0.0.4B update
EDC limit although appears correct in ryzen master you lose points. After you set it to 0 (aka default) you gain points
Cinebench R20, 4981 with EDC at 999 (aka 600 mobo max), and 5025 with EDC at 0. With both power plans (AMD high perf and 1usmus). All other power limits at max.

The only difference i can observe with 1usmus power plan is that on single thread benchmark it uses "fastest core of the system" (gold star) and doesn't change cores randomly but clocks are the same.


However more importantly RAM stability is broken. I dropped ram coverage 7-8 fold..from over 7000% to under 800% with the same exact settings.
I will be using the bios switch to go back to 1.0.0.3ABBA to check if there is something else that's causing it but i doubt it.

 

[St0RM53]

Just wanted to mention that some of the default load "XMP" RAM timings have changed between bios versions for my kit. I manually tune every setting but if one were to just load xmp and forget about it the changed timings can cause instability for some kits. Also the tighter timings may be the main reasin for any performance gains.

I run manual timings and voltages (VDDP/VDDG/SOC etc). They are exactly the same. I re-tested the ABBA bios and i get lots of more coverage. Plus i get +50points in Cinebench R20 even after the EDC "fix" in the new one..1.0.0.4B is totally broken. Gigabyte needs to fix this along all the other problems it introduced. Might as well fix their other software as well, more specifically rgbfusion which crashes and restarts forever until you kill check_kill.exe manually and one in 50 times it will launch without crashing.

 

[St0RM53]

X570 AURUS MASTER updated to F10A with AGESA 1004 B.
XMP profile. The memory works well without errors. I would like to know, from the photo you can say that everything is fine in my system? Thank.

Yes it works, but for how long is the question.
You need to test stability with a program like Karhu Ram test and see how much coverage you can get before an error appears.

 

[St0RM53]

We are looking into this.


I have read alot of posts here, reddit, and else where. I am not convinced this power plan really helps with anything but "perception". Maybe it does help performance in certain situations? Not sure. Does not seem to be a magic bullet.



1004B introduced alot of changes to memory training. Things wont be the same (for better or worse) if you compare 1003 -> 1004.

Thanks a lot Matthew. Good to know you are working on it. AMD needs to release new chipset drivers as well.

I hear ya, I would love to use SIV to control my fans from within Windows but did you see the reddit post about the EDC bug in the latest AGESA 1.0.0.4? Basically you have to set PBO to manual then set EDC to 0 (which will set to board default)

Better detail here:https://old.reddit.com/r/Amd/comments/dqwcpo/agesa_1004_pbo_bug_and_how_to_fix_it/

AMD really needs to invest the $ from Zen 2 sales towards developers/ beta testers, etc

I only mentioned the changed timings when only loading XMP as I see a lot of users here just load XMP and plug in 1.35 for DDR voltage and call it a day. If you are not documenting all RAM Timings that can cause issues especially when none of the changes in bios are documented.

off topic but has anyone seen the Moore's Law video when he is @ Hot Chips and someone asked Jim Keller and Raja "what about Spectre and Meltdown?' They both busted out laughing so hard that they nearly soiled themselves. That's how they feel about their customers data. In fact Spectre and Meltdown has lead to record profits and 14nm shortages as Data Centers have to order 20-30% more servers to negate the 20-30% performance loss. And I hear NetCat is even worse but no one seems to care and media outlets are silent. But they report how AMDs Zen 2 chips not hitting advertised boost clocks must be a scam lead by AMD. The main editor from Paid Perspective has this board paired with a 3900X yet every week complains how "X570 still can't run my RAM or be stable" And when ABBA dropped he was like " well AMD will release it this month but it will take several weeks for vendors to get it and produce a bios and push it out to consumers. Yet failed to see the abba official bios already live on the website. Being a media person I'm sure he could have contacted a GB rep and got a bios before we did. If this isn't bias then its gross incompetence either way a shame.

Yes i know it was on my initial post a few pages behind:
https://www.overclock.net/forum/11-...570-aorus-owners-thread-372.html#post28183996

 

[St0RM53]

On my 3800X using F7b (1.0.0.3 ABBA) the power plan did increase my CB scores a little. Like less than 1%. Now in CB20 i get 5093/527
Using 1903 Win10. I have yet to test F10a.

are you on water?

 

[St0RM53]

Sooo..when are we getting a 1.0.0.4C update ready?

 

[St0RM53]

Whats new in 1.0.0.4C?

Not known if it's gonna be called like this but it's 1.0.0.4 with all the fixes;p

 

[St0RM53]

Well this should only really matter when a lot happens on the PCIe-BUS – the Chipset has 8-9 GB/s to share for the SSD in Slot 2 & 3.
As long as you don’t use everything at the same time this shouldn't be a big deal.
Of course a RAID 0 with all three M.2 Slots wouldn’t make sense…

…but I am running an Aorus Gen4 NVME-SSD in Slot 3 and when I first tested it – t measured speeds around 5000 MB/s read and write as expected when I installed it in Mid October – and now read speed dropped significantly(2400-2600MB/S) while write speed at least stays over 4000 MB/s.

So it seems something is going on here…of course I don’t know if it is really only related to the new BIOS or something with Windows since I don’t do always Crystalmark after every BIOS-Update…but something changed since F5b – when I first tested it in October.

Greetings,
Medizinmann

PS: And yes the 3rd M2-Slot is also Gen4 - it shares bandwidth with Slot 2 - but this only matters if accessed at the same time!!

it's because your drive is not almost empty now..it is normal all ssd's except the ones that are all SLC have this "issue"

 

[St0RM53]

Re-tried AGESA 1.0.0.4 with F11 on my x570 aorus master and still kinda of broken. First thing i've noticed setting VDDG and VDDP voltage control and voltage in favorites is broken. It only allows you to set "VDDP voltage" (if you have manual selected in "VDDP voltage control") and "VDDG voltage control" (while if set to manual doesn't allow you to add "VDDG voltage") in favorites. Trying to set one of these 2 parameters in favorites adds "PM_1V8" to favorites...


Also most importantly i run Cinebench R20 again..
Without EVEN enabling XMP (yes with like 90ns latency) and just running defaults in F11 i get close to the score i get in F7B with everything unlocked and enabled.
Enabling those options in F11 drops me around 60-70 points like something is REALLY REALLY BROKEN.
I need to play with the settings one by one to find *** is going on.


@GBT-MatthewH write that down ;p

 

[St0RM53]

Yo guys i made this tip:


PSA: Overclocking Infinity Fabric (and RAM) past 1800mhz and having random system crashes/reboots? It's because your specific CPU I/O die crashes. Solution inside.

From the time i started overclocking IF and RAM i was having this issue.

Randomly in the matter of minutes or days at any load my Ryzen 3000 system would crash and reboot.
I am not talking about BSOD but full system crash and restart. If you have BSOD it's usually the RAM's fault assuming you are not overclocking the CPU cores.

I've done several tests but i was always assuming i needed to refine my RAM timings further (because this is i was mainly doing), even though i was getting good coverage results.

I've changed and monitored various parameters but i finally figured it out the source of the issue and fully tested the solution.

Turns out even though i have one of the top mobos (Gigabyte x570 aorus master) the actual SOC voltage getting to the CPU would sometimes randomly drop about 40-50mV.
Not only that, the actual SOC voltage was 13mV lower than set one at a normal scenario (i.e 1.087V when set to 1.1V).

This in turn will either cause the SOC to run into instability or drop the set VDDG voltage further running the system to instability.

This is because the VDDG voltage (now split with the latest bios), is regulated from the SOC voltage and is always at most ~40mV lower than the SOC voltage.

I can't confirm which one of the two voltages causes the actual crash, but since they are related it's probably both.

Solution was to increase the SOC voltage (in my case from 1.1V to 1.15V. VDDG is (and was) set to 1.050V).

Since i cannot reproduce the crash on demand, my verification testing consists of a super user/engineering/VM/gaming load that in experience will crash. So far 5 continuous days uptime working perfectly

This will vary depending on the specific CPU you have. In my case my 3800X doesn't have a top end I/O die since i cannot even do 1900mhz and max at 1867mhz. My cousin's 3600X does 1900mhz no prbs.
Still you can't take frequency and voltages as a given, even at the same frequency voltages WILL DIFFER between 2 dies.

We can't blame AMD for this because they only promised 1800mhz MAX, but it's a shame they don't (seem to) bin the I/O dies as well, considering the price difference.

Also make sure that you have SOC/Uncore OC mode enabled since it theoretically helps and select an appropriate SOC voltage line calibration setting.

Matthew is still on "holidays"?

Saw the new F12 beta but no change log.


Still didn't figure out what the heck is going on since F7b (testing F11 specifically) and PBO. Performance is lost; EDC "shows" to be fixed but is it really?

 

[St0RM53]

Today I did some testing and, as reported already, the correct vCore is -0.1 (about 1.4v under all core load).

Is there any trick to improve the actual VID under load?

Going below or above is always worse.
Raising the vCore only results in much worse thermal and lower VID; lower scoring, more unpredictable.
Dynamic vCore at +0.1 is actually VID at 1.3, doesn't really make sense.
Effective clock speed delta from perf clock/MP ratio also gets higher, up to 100 Mhz.

Tried adjusting the LLC but Medium is the most stable.
Anything above High for LLC gives more frequent and higher vdroops (which again doesn't really make sense).
Also LLC above Medium for me has proven to be unstable in the long term (random reboots).

PWM Phase Control doesn't have any impact.

Of course this is all about the scenario with PBO/EDC=0 enabled.

I believe your SOC and VDDG voltage drop slightly below stable voltage, increase by 50mv both

 

[St0RM53]

Is this the sticker you are talking about for production number? It's the only one I could see on my board.

Mine is
1/D/J5A105 04285


hope it's good

 

[St0RM53]

The cold-boot issues people constantly bicker about is Memory related. You board fails to train the memory on boot and resets settings.

100% the case I can say with confidence at this moment. People just don't want to bother trying to fix it and just change boards. Small issues become big issues.

Doesn't help the DRAM Calculator doesn't take notice that different brands of boards likes different settings for memory. They are quite accurate generally but they can and usually need tweaks for your particular board.

I've used my memory kit on 3 different boards and 2 cpu's and all needed different settings for best compatibility. No board wanted the same exact settings. Though the gigabyte boards I had were mostly similar the other brand of board was a different character entirely.

Just casually mentioning differences is they all wanted different procODT settings for boot. One board didn't want to boot with the memory with less than 53Ohm on Ryzen 1000 series while the other board was happy with 43Ohm.

CAD_BUS settings need to be tweaked... I never see anyone mention CAD_BUS!!!

CAD_BUS IS DAMN 90% OF YOUR ISSUE TO GET IT TO BOOT PROPERLY! Every board I've used need this different.
It's the [RZQ] values.
Gigabyte love [RZQ/6][RZQ/3][RZQ/1] & [RZQ/7][RZQ/3][RZQ/1] & [RZQ/7][RZQ/2][RZQ/1], numbers in this ballpark work best when using Micron rev.E memory. Other brands might need others but this is 99% true if you ask me if using this kind of memory on these boards.

I see often the [disabled][disabled][RZQ/X] used. It's literary horrible to boot with I've found out, training nightmare.
Don't use it unless you for certain know it works with your brand of memory and motherboard. The usual case being ASUS & Samsung B-die. Everyone just copies it.

well i sold my 3800x to my brother and build him a PC with the same 2x16gb e-die on an msi b450 gaming pro carbon ac with the same timings and frequencies (3733mhz/1:1/66ns latency - cpu can't do 1900mhz fclk) and it works just fine;p Funnily enough when i just put in the 3950x i got in my x570 aorus master it failed to post and i had to basically re-load the same settings again (it also did 3800mhz without changing anything proving indeed my specific 3800x couldn't handle it)

 

[St0RM53]

That's correct. Sorry in all the personal life changes I forgot to update you guys. My last day at GIGABYTE was right after CES. I wanted to thank everyone in this community for their feedback. With your help we got to make some amazing products even better as a direct result of your suggestions, feedback, bug reports, etc.

Bios flashback on every board? You guys asked for it. Finned heatsinks? You guys liked it and we agreed. Only passively cooled x570? We knew you'd like it Better vrm than even I thought possible 2 years ago? You guys paying attention to the details. Dropping orange ascents for a more neutral black? You guys saying you loved x product but not the color...

In fact if you made a thoughtful post and asked us to do/change/make something there is a good chance I took a screen shot of your post, put it in a internal presentation, and said "this is what they want and I agree". So again thank you to everyone who pushed us to make the best products possible.

Finally thanks for all the support for me personally - sitting behind a screen often let's us forget there is a person on the other end. I can tell you that every time someone benefited from a change we made and took the time to say thanks, keep it up, etc, all of that helped push us to keep doing it.

This is probably the last post you will see from this account. My new position is much less public facing (AMD for those wondering) but I'll still be lurking around, taking note of suggestions, etc. I still plan on being an advocate for you guys internally but perhaps not as publicly. Its been fun guys. Be good to each other!

I've seen your goodbye on reddit, hopefully it goes well for you, but you forgot to tell us if someone else is taking this position you had or are we gonna be left in the dark?;p

Same happened to me but if you read some posts back, a friend here replaced the battery and here is ok now so we really don't know for sure if it is a hardware problem or a simple battery swap.
You have the old motherboard?

i never had the cold bug so definitely defective or unstable hardware related

 

[St0RM53]

Yeah, that's fine. That's not the issue. The issue is it takes ~1 min from the time I click sleep to when my PC actually goes to sleep.

Here is the Power Options I have set right now. I disabled hibernation, which is why you can't see it there now.

I don't think this is my issue. I do have my OS on an M.2 SSD alongside a SATA SSD and a normal HDD for storage. But I don't hear my HDD spin up when I click sleep.

Run powercfg /requests to see what is preventing it going to sleep. For example i had firefox had some tab "playing" something preventing sleep/turn off monitor.

 

[St0RM53]

Are you sure the RAM is 100% stable?

I have found that even with slight instability it will restart without leaving a blue screen or mini dump to troubleshoot further.

Solution: https://www.reddit.com/r/Amd/comments/end4d8/psa_overclocking_infinity_fabric_and_ram_past/


We should pin this

 

[St0RM53]

I saw the following after running the command:

Running "powercfg -requests" shows:

DISPLAY:
None.

SYSTEM:
None.

AWAYMODE:
None.

EXECUTION:
None.

PERFBOOST:
[DRIVER] Legacy Kernel Caller
Power Manager

ACTIVELOCKSCREEN:
None.

Googling this led nowhere, but I had an idea to turn hibernation back on:

powercfg /h on

After doing so the following two options came back in power settings. I set "Allow hybrid sleep" to "Off" and "Hibernate after" was already set to "Never". This worked like a charm! My PC now instantly sleeps--YAY!

Glad i kinda helped ;p

 

[St0RM53]

forgot to ask how to you stop windows 10 putting my internal drives into the safty remove tool as i have nearly clicked on my internal drives, I dont want them in the list as the c drive is not listed but my other drives are

my c drive is a NVMe drive which is not listed
d,e drives are ssd which are listed which i dont want as i just want external drives and usb sticks and so forth listed

thanks

you can selectively disable with registry option without disabling hot-plugging..google search

 

[St0RM53]

Does anyone knows which USB ports are the ones coming straight from a Ryzen 3000 CPU's on X570 Aorus Master?


I remember i've seen a diagram showing which lanes go where on here but i can't find it. I don't think it had the USB ports on it though


edit: Nevermind, found it, they are: 4*Rear USB3.1 *2, USB3.0 *2

 

[St0RM53]

Don't forget that MSI is working on their updated Bios versions whilst Gigabyte is sitting dumb like it doesn't matter anymore to them. Not good enough to only release hardware then walk away from it offering no support whatsoever.



Have my eye on anything from either Asus, Asrock or MSI built boards but absolutely NOTHING from Gigabyte, not good enough to supply only 16Mb bios ROMs alone and then charging Aorus Master and Xtreme price tags.

That was indeed scum-bag like behaviour in doing that and then trying desperately to justify their dirty decision in doing that on expensive boards. Justified on smaller priced hardware here with lacking overclocking potential but not with those models.

Not an MSI shill here but at least they never shafted the end user with substandard bios updates, they simply work whilst Gigabyte's efforts was completely shameful and pathetic.

Yeah Gigabyte is really flipping the bird lately..let's not forget they've changed GBT_Matthew's position and they didn't replace the reps..and 0 updates since that point..it's not gonna end well for them like this