Wonky system issues, Asus board-CPU overvolt concerns

[An0maly_76]

With all the brouhaha about AM5 thermals, and, more recently, GamersNexus showing horrendously wrong CPU voltage on Asus' AM5 boards, it's got me thinking about swapping my Asus Tuf B550-PLUS for a different brand. Mine's been skitzo enough from day one that it makes me wonder how long they've been getting away with that. First, it didn't like the RAM I used, which took awhile to sort out (wound up building another test system while the board was being RMA'd. Interestingly, Asus' RMA process went from 'waiting for parts' to 'returned - issue not duplicated', with no explanation.

Yes, I know AM4 is not AM5, but I noticed what appeared to be some light socket discoloration when reassembling. So in light of GN showing high CPU socket voltage on Asus' AM5 offerings, I wonder how long they've been getting away with this.

With AMD's statement of max boost clock of 4.8 Ghz for the 5900X, my benchmarks were surprising, seeing 4.975 Ghz. Yes, I have a monster cooler (Mugen 5), but it makes me wonder if my Asus AM4 board isn't doing the same thing their AM5s are. Because my experience with Ryzen systems has made me a big believer in QVL when it comes to pairing components, and my RAM (PVB432G320C6K Patriot Viper Blackout 4), at 3200, is not that fast.

Also, while Asus' QVL specified 16-18-18-36 timings, requiring manual timings, as the board's DOCP popped up a profile for 20-20-20-40 timings, and the system still had minor stability issues at both settings until I disabled DOCP and ran them at their base 2400 clock. About two months after I built this system, the screen went blank and the system went down with the fans went wide open.

I've also had an issue with audio playing peek-a-boo, the GPU's RGB defaulting to the rainbow puke after losing slot power, which has been nothing short of annoying (I have autism-related sensory issues, and certain types of light bother me). The BIOS also thinks my Pioneer Blu-Ray writer and WD Black HDD are a RAID array whenever the IPC is disconnected, even though RAID is disabled. Like I said, it's been one thing after another with this rig. After the cold-boot / power-loss issue, it runs fine.

Sure, there could be other reasons for all this, but it certainly could be some issue with the board, and poor CPU voltage regulation could certainly be a factor, methinks.

I'm looking at a Gigabyte B550 DS3H, which appears to support all my current components, including the RAM. Amazon has these on sale currently. And while they don't seem to have much of a heat sink (if any), the VRMs certainly look beefy, and even if they can't handle the base clock, the 5900X is kind of overkill anyway.

Thoughts?

 

[wNotyarD]

Is PBO enabled on your rig? That'd explain boosting over the nominal 4.8GHz when there's (thermal and power) leeway to do so.

 

[An0maly_76]

Is PBO enabled on your rig? That'd explain boosting over the nominal 4.8GHz when there's (thermal and power) leeway to do so.

Yes, also have an 850W PSU, but I thought 4.8 was the max boost using PBO?

Even if that's not out of the realm of possibility... The other issues would indicate something is not right. While I've had this chip nearly 2 years now, I've also heard that some owners won the silicon lottery. I should think I would have noticed far worse problems were that the case though.

 

[wNotyarD]

Yes, also have an 850W PSU, but I thought 4.8 was the max boost using PBO?

Even if that's not out of the realm of possibility... The other issues would indicate something is not right. While I've had this chip nearly 2 years now, I've also heard that some owners won the silicon lottery. I should think I would have noticed far worse problems were that the case though.

I mean, 4.8 is official boost. PBO is Precision Boost Override, and I do remember AMD marketing says it makes them Ryzens go above their nominal boost if able to, but hardly so on largely threaded tasks.

 

[An0maly_76]

Here is the info I have available for voltages...

 

[RJARRRPCGP]

I'm looking at a Gigabyte B550 DS3H

Anything with "DS3H" and "HDV" are bottom-of-the-barrel! It's likely that they can't handle anything more than a Ryzen 5 5600X!

 

[freeagent]

I dont like their TUF stuff. I ordered a TUF matx board from Newegg and it was DOA. But the quality was terrible. It even felt like a piece of junk. But I do have 3 Asus AM4 boards, and none of them had wonky voltages. Also, max boost before PBO is 4950MHz. Why many people say 4850 is beyond me. Power looks fine for a stock 5900X.

 

[ir_cow]

Just update the BIOS to the newest.

 

[tabascosauz]

With all the brouhaha about AM5 thermals, and, more recently, GamersNexus showing horrendously wrong CPU voltage on Asus' AM5 boards, it's got me thinking about swapping my Asus Tuf B550-PLUS for a different brand. Mine's been skitzo enough from day one that it makes me wonder how long they've been getting away with that. First, it didn't like the RAM I used, which took awhile to sort out (wound up building another test system while the board was being RMA'd. Interestingly, Asus' RMA process went from 'waiting for parts' to 'returned - issue not duplicated', with no explanation.

Yes, I know AM4 is not AM5, but I noticed what appeared to be some light socket discoloration when reassembling. So in light of GN showing high CPU socket voltage on Asus' AM5 offerings, I wonder how long they've been getting away with this.

With AMD's statement of max boost clock of 4.8 Ghz for the 5900X, my benchmarks were surprising, seeing 4.975 Ghz. Yes, I have a monster cooler (Mugen 5), but it makes me wonder if my Asus AM4 board isn't doing the same thing their AM5s are. Because my experience with Ryzen systems has made me a big believer in QVL when it comes to pairing components, and my RAM (PVB432G320C6K Patriot Viper Blackout 4), at 3200, is not that fast.

Also, while Asus' QVL specified 16-18-18-36 timings, requiring manual timings, as the board's DOCP popped up a profile for 20-20-20-40 timings, and the system still had minor stability issues at both settings until I disabled DOCP and ran them at their base 2400 clock. About two months after I built this system, the screen went blank and the system went down with the fans went wide open.

I've also had an issue with audio playing peek-a-boo, the GPU's RGB defaulting to the rainbow puke after losing slot power, which has been nothing short of annoying (I have autism-related sensory issues, and certain types of light bother me). The BIOS also thinks my Pioneer Blu-Ray writer and WD Black HDD are a RAID array whenever the IPC is disconnected, even though RAID is disabled. Like I said, it's been one thing after another with this rig. After the cold-boot / power-loss issue, it runs fine.

Sure, there could be other reasons for all this, but it certainly could be some issue with the board, and poor CPU voltage regulation could certainly be a factor, methinks.

I'm looking at a Gigabyte B550 DS3H, which appears to support all my current components, including the RAM. Amazon has these on sale currently. And while they don't seem to have much of a heat sink (if any), the VRMs certainly look beefy, and even if they can't handle the base clock, the 5900X is kind of overkill anyway.

Thoughts?

The visual discoloration on the plastic PGA lid is normal. Every motherboard has that.

By now you should know that box boost spec is not actual spec. Ryzens are governed by Fmax as a max theoretical boost ceiling. 5900X Fmax is 4950, and silicon quality allowing can reach up to 5150 with +200 boost override. The box number is only vaguely suggestive of where Fmax lies, sometimes it's 50, 100 or 150MHz above box number. As always, what each core actually hits is a lottery, at a bare minimum even the stock Fmax (4950) is hoped for but not guaranteed.

HWInfo reports Fmax as Global Frequency Limit.

• AM4 chiplet CPUs should not be run past 1.2V VSOC daily anyway, as a general conservative rule. Renoir and later APUs can see higher VSOC relatively safely (but not always, again lottery).
• As long as Vcore peaks do not exceed about 1.55V in software when PBO (1.5V for bone stock no PBO), there's not a need to worry.
• SOC related issues (audio, USB) are VSOC related, bump it up a little and/or put your VDDGs where they should be.

Also, you're running 2400 JEDEC not your 3200 XMP? Being as low as 1.0V VSOC is nice if it's stable but if it's giving USB and audio issues you have plenty of headroom to work with.

TUF to HDV would be the downgrade to end all downgrades, considering the HDV is a board that can't even be trusted to run a 88W CPU without copious amounts of airflow.

 

[An0maly_76]

Anything with "DS3H" and "HDV" are bottom-of-the-barrel! It's likely that they can't handle anything more than a Ryzen 5 5600X!

Be that as it may, they do list the 5900X as supported. Of course, MSI said that about the B450M Bazooka I have as well, only to recant later.

So the general consensus is that HWinfo isn't showing anything abnormal?

 

[tabascosauz]

Be that as it may, they do list the 5900X as supported. Of course, MSI said that about the B450M Bazooka I have as well, only to recant later.

So the general consensus is that HWinfo isn't showing anything abnormal?

CPU support list only ever lists whether a CPU can boot with BIOS support. It never takes into consideration whether the board will perform poorly, become a fire hazard or shutdown on you due to OTP. Just because you can doesn't mean you should.

B450M Bazooka isn't exactly a shining specimen either but it's still orders of magnitude more capable than the HDV.

Your HWInfo looks fine, except for the fact that you aren't running the Fabric and UCLK speed you think you are running.

 

[An0maly_76]

CPU support list only ever lists whether a CPU can boot with BIOS support. It never takes into consideration whether the board will perform poorly, become a fire hazard or shutdown on you due to OTP. Just because you can doesn't mean you should.

B450M Bazooka isn't exactly a shining specimen either but it's still orders of magnitude more capable than the HDV.

Your HWInfo looks fine, except for the fact that you aren't running the Fabric and UCLK speed you think you are running.

Care to elaborate? I know some things this sort of software can only guess at, but I'm wondering why that is, or if there's another reason for this... Maybe just a Ryzen thing?

On the surface, it sounds like the board is fine, because it seems to me there would be discrepancies in the HWinfo that would jump out at others. I was just thinking that if some AM4s were doing the same thing, that perhaps this monster cooler is the only thing preventing my 5900X slowly getting cooked.

So what could cause the other issues? I have a Corsair Commander with six LL120s. Is it possible the iCUE software is fighting with Armoury Crate? I didn't want to install it, but it was the only way I could get the RGB to go dark on shutdown. Otherwise it would rainbow puke. But it still does it on the first boot after power loss.

 

[tabascosauz]

Care to elaborate? I know some things this sort of software can only guess at, but I'm wondering why that is, or if there's another reason for this... Maybe just a Ryzen thing?

On the surface, it sounds like the board is fine, because it seems to me there would be discrepancies in the HWinfo that would jump out at others. I was just thinking that if some AM4s were doing the same thing, that perhaps this monster cooler is the only thing preventing my 5900X slowly getting cooked.

So what could cause the other issues? I have a Corsair Commander with six LL120s. Is it possible the iCUE software is fighting with Armoury Crate? I didn't want to install it, but it was the only way I could get the RGB to go dark on shutdown. Otherwise it would rainbow puke. But it still does it on the first boot after power loss.

The cooked thing is strictly a [past?] AM5 issue. All silicon on the substrate changed significantly for AM5. There's really no need to keep compounding your worries lol. It's been 4 years of living with this Gloflo IO die, the OC community knows well what is safe and what isn't. Stay below 1.2V, most of the time you don't even need 1.1V for mundane speeds.

You can pretty clearly see in your screenshot that UCLK and Fabric are both 1200MHz. That's DDR4-2400 in 1:1.

The TUF is just fine. The X570s and B550s in both ATX and mATX have always been strong midrange choices.

 

[An0maly_76]

You can pretty clearly see in your screenshot that UCLK and Fabric are both 1200MHz. That's DDR4-2400 in 1:1.

The TUF is just fine. The X570s and B550s in both ATX and mATX have always been strong midrange choices.

Obviously, there's a lot I don't understand about this stuff ( I've learned a bit about Ryzen in the time I've owned them, but cannot really claim to even be a card-carrying parts assembler LOL), but help me to understand why the RAM running at 2400 works out that way. I guess I'm just getting overly concerned because of the other weird stuff going on, combined with the fact that I paid good money back in the 90s for someone to deliver me an OC'd Cyrix M-II machine and insist that it was fine, despite the fact I could literally smell it burning up while I used it.

So, if I'm reading this right (my understanding of these things is obviously limited), VID is voltage requested, so on the surface it looks like the CPU is requesting 1.03ish V at idle, which it's fully getting, but under load is requesting 1.525V, but the board is limiting output to 1.494V, both within limits. During which SoC is 0.962V - 1.006V, well within the 1.2V-1.3V I'm told is acceptable for that.

The cooked thing is strictly a [past?] AM5 issue. All silicon on the substrate changed significantly for AM5. There's really no need to keep compounding your worries lol. It's been 4 years of living with this Gloflo IO die, the OC community knows well what is safe and what isn't. Stay below 1.2V, most of the time you don't even need 1.1V for mundane speeds.

I did find something recently that showed impressive benchmarks with lower thermals by setting core ratio to 46.50 and VDDCR to 1.23750. Should I even bother with that?

 

[wNotyarD]

Obviously, there's a lot I don't understand about this stuff ( I've learned a bit about Ryzen in the time I've owned them, but cannot really claim to even be a card-carrying parts assembler LOL), but help me to understand why the RAM running at 2400 works out that way.

Is your RAM actually set at 2400? If yes, there ain't nothing wrong here then.

 

[An0maly_76]

Is your RAM actually set at 2400? If yes, there ain't nothing wrong here then.

Yep. It's rated for 3200, but I found disabling DOCP to be more stable than the timings recommended by Patriot or the ones recommended by Asus.

But I'm curious how this works. Is the RAM speed divided across the total # of DIMMs? And how does the ratio thing work with RAM? Also, how does the RAM speed / voltage intertwine with SoC, etc.?

 

[tabascosauz]

So, if I'm reading this right (my understanding of these things is obviously limited), VID is voltage requested, so on the surface it looks like the CPU is requesting 1.03ish V at idle, which it's fully getting, but under load is requesting 1.525V, but the board is limiting output to 1.494V, both within limits. During which SoC is 0.962V - 1.006V, well within the 1.2V-1.3V I'm told is acceptable for that.

Short answer - we don't look at VID on Ryzen. It doesn't indicate much useful info.

HWInfo often doesn't read Ryzen's idle Vcore properly. If you want to see what it's doing at true idle, open Ryzen Master.

The only hardware sensors aside from RM that will read Ryzen idle Vcore are die sense circuits on ROG Crosshair boards when set to Die Sense in BIOS. Replaces the Vcore reading under the HWInfo board sensors section with something more accurate.

I did find something recently that showed impressive benchmarks with lower thermals by setting core ratio to 46.50 and VDDCR to 1.23750. Should I even bother with that?

Nope.

 

[freeagent]

It's rated for 3200, but I found disabling DOCP to be more stable than the timings recommended by Patriot or the ones recommended by Asus.

Huge red flag there. It should *just* work. If you have to jump through hoops then that is a definite problem.

 

[An0maly_76]

Is your RAM actually set at 2400? If yes, there ain't nothing wrong here then.

Yep. It's rated for 3200, but I found disabling DOCP to be more stable than the timings recommended by Patriot or the ones recommended by Asus.

But I'm curious how this works. Is the RAM speed divided across the total # of DIMMs? And how does the ratio thing work with RAM? Also, how does the RAM speed / voltage intertwine with SoC, etc.?

Huge red flag there. It should *just* work. If you have to jump through hoops then that is a definite problem.

That's interesting, I wondered about that too, especially since this RAM is on Asus' QVL list for this board (previous Ballistix were not and caused stability issues).

I did find that the board's confusion about my HDD and ODD being a RAID array after power loss went away after replacing the CMOS battery. Not sure how, but apparently I confused an old one for the original after getting it back from RMA. This did not, however, fix the GPU's RGB rainbow puke under these circumstances.

One other thing. I've had intermittent audio issues (outlined in more detail in another thread), but basically it appears that my audio has been jumping to a different output nearly the entire time I've had this system together -- only recently did I discover it was apparently jumping to the rear output. Before I started using the rear output, it was simply disappearing. Now that I am using it, it will somehow use the rear output also when it's only supposed to be using HDMI. Makes me think this board is just flaky enough to be a PITA, but I've heard about an audio issue with certain Win10 updates... Could this be related?