Can a BIOS update improve undervolting performance? (AMD Curve Optimizer)

[YoungBlade]

I did a BIOS update a few weeks ago on my PC in an attempt to fix an NVMe issue (the update was actually a bad experience, but I don't need to get into that story right now) and on a whim today, I decided to check to see if my Curve Optimizer results had changed. They have, actually, and seemingly much for the better. I can now apparently run at the full all-core -30 offset with no issues. My CPU is the Ryzen 9 5900X and my motherboard is the MSI B550 Gaming Edge WiFi. I went from BIOS version 7C91v16 to version 7C91v19.

 

Is it common for BIOS updates to improve undervolting performance for CPUs?

 

Obviously, the VRMs are on the motherboard and they clearly do have a roll to play in regulating the voltage, but I had thought that the primary determining factor for undervolting with Curve Optimizer was CPU silicon quality. I didn't realize that a simple BIOS update could make such a big difference in this regard.

 

If you'd like further details on what I've done or the BIOS update fiasco, I've put them here:

Spoiler

Here's what the CO results I had before were:

 

For my test today, I tried taking the two worst cores to -20 and -15 and they were both now stable. I then tried to see if I could get OCCT to produce the errors it used to, and put all cores to -30. But OCCT didn't produce any errors this time for any core I tried, which was odd to say the least.

 

The first time I tried -30 several months ago, the initial problem I found was that Windows Defender crashed the computer by doing a scan. That was a repeatable issue, so I tried that again, too. There was no crash this time.

 

Cinebench can also run fine on a loop - it occasionally crashed with an all-core -30 before. And I ran the benchmark for SotTR with no issues. I also ran an OCCT all-core stress test to make sure that's okay (it was always fine before, but the status quo has clearly shifted) and I'm going to try Prime95 next. Any other suggestions for a test is appreciated, also, as I want to be as sure as possible that my system is stable.

 

As for the BIOS update, the long-story-short version is: I had been having an occasional issue where Windows would crash and the system would reboot, and upon reboot, the primary NVMe drive housing Windows would be missing from BIOS. This issue would go away after an apparently random period of time, although one time I seemingly fixed it by reseating the drive. I was unable to diagnose the issue, so I figured I'd try a BIOS update since the board had had several since I built the system.

 

I decided to just run the BIOS update from within Windows, because I was too lazy to get a flash drive, which yes, is a bad idea, and yes, it turned out to be a mistake. Believe me, I'll never do it again.

 

The system wouldn't post after the BIOS update. Lights and fans came on, but the debug LED stopped at "CPU." I was understandably pretty concerned. Luckily, my board has BIOS flashback, so I got a flash drive, went to my secondary system, put the BIOS on there, and then used the flashback feature. It took a pretty long time (over 15 minutes) but the flashback completed successfully and while I lost all my BIOS settings, I had a working system again.

 

I'll also always get a motherboard with BIOS flashback and/or a secondary BIOS chip from now on, too. Those features are absolutely invaluable.

 

[LIGISTX]

16 minutes ago, YoungBlade said:

I didn't realize that a simple BIOS update could make such a big difference in this regard.

I think this is the biggest part here, BIOS updates are anything but simple. 
 

BIOS is what regulates all of the mobo functions and interactions with the devices on the board. How these interactions and controls are implemented and improved can have very real and very noticeable affects. 

[YoungBlade]

1 hour ago, LIGISTX said:

I think this is the biggest part here, BIOS updates are anything but simple. 
 

BIOS is what regulates all of the mobo functions and interactions with the devices on the board. How these interactions and controls are implemented and improved can have very real and very noticeable affects. 

I didn't mean "simple" in the sense of "a BIOS update should do nothing" - after all, I did the update in the hopes of fixing an NVMe issue, so I understand that the updates can impact system stability. I meant "simple" in the sense of "it was easy to do." I didn't need to buy a new board or CPU or PSU. All I had to do was download an update.

 

Mainly, I wasn't expecting to get a better result in something like undervolting specifically, as I figured that was primarily CPU silicon dependent. Basically that, as long as the motherboard (and power supply) were of a minimum quality level, they were not really factors. This seems to be in line with the prevailing wisdom - motherboards don't really impact performance (unless the VRMs throttle or they don't allow sufficient power limits, but that's again the "minimum quality level" issue).

 

Is this idea that motherboards don't impact performance just flat out wrong and folks like Linus need to stop spreading misinformation that motherboard choice is about features and not performance? Or is even basic OC/UV an exception to that advice? (Obviously, extreme OC depends heavily on the motherboard, and basically every other component in the system, too.)

 

And if a BIOS update - not even a board swap - can regularly result in noticeable performance improvements, then why is the advice to only update the BIOS if you're having an issue? Shouldn't enthusiasts be recommending constant updates to the latest version to get better performance?

[Blue4130]

1 hour ago, YoungBlade said:

I didn't mean "simple" in the sense of "a BIOS update should do nothing" - after all, I did the update in the hopes of fixing an NVMe issue, so I understand that the updates can impact system stability. I meant "simple" in the sense of "it was easy to do." I didn't need to buy a new board or CPU or PSU. All I had to do was download an update.

 

Mainly, I wasn't expecting to get a better result in something like undervolting specifically, as I figured that was primarily CPU silicon dependent. Basically that, as long as the motherboard (and power supply) were of a minimum quality level, they were not really factors. This seems to be in line with the prevailing wisdom - motherboards don't really impact performance (unless the VRMs throttle or they don't allow sufficient power limits, but that's again the "minimum quality level" issue).

 

Is this idea that motherboards don't impact performance just flat out wrong and folks like Linus need to stop spreading misinformation that motherboard choice is about features and not performance? Or is even basic OC/UV an exception to that advice? (Obviously, extreme OC depends heavily on the motherboard, and basically every other component in the system, too.)

 

And if a BIOS update - not even a board swap - can regularly result in noticeable performance improvements, then why is the advice to only update the BIOS if you're having an issue? Shouldn't enthusiasts be recommending constant updates to the latest version to get better performance?

What took you minutes to apply took engineers and software devs hundreds of hours to make. It's no trivial thing, a bios update. Yes, bios and boards do affect performance not all 550s will achieve the same level of performance. 

 

And I think the reason why updating the bios is not recommended is because the possibility of catastrophic failure is higher than something like a ram timing tweak. 

[LIGISTX]

1 hour ago, YoungBlade said:

I didn't mean "simple" in the sense of "a BIOS update should do nothing" - after all, I did the update in the hopes of fixing an NVMe issue, so I understand that the updates can impact system stability. I meant "simple" in the sense of "it was easy to do." I didn't need to buy a new board or CPU or PSU. All I had to do was download an update.

 

Mainly, I wasn't expecting to get a better result in something like undervolting specifically, as I figured that was primarily CPU silicon dependent. Basically that, as long as the motherboard (and power supply) were of a minimum quality level, they were not really factors. This seems to be in line with the prevailing wisdom - motherboards don't really impact performance (unless the VRMs throttle or they don't allow sufficient power limits, but that's again the "minimum quality level" issue).

 

Is this idea that motherboards don't impact performance just flat out wrong and folks like Linus need to stop spreading misinformation that motherboard choice is about features and not performance? Or is even basic OC/UV an exception to that advice? (Obviously, extreme OC depends heavily on the motherboard, and basically every other component in the system, too.)

 

And if a BIOS update - not even a board swap - can regularly result in noticeable performance improvements, then why is the advice to only update the BIOS if you're having an issue? Shouldn't enthusiasts be recommending constant updates to the latest version to get better performance?

Typically mobo doesn’t matter, yes. But AMD keeps implementing new agesa microcode updates and stuff, so as those get incorporated into bios updates, it does make a difference. But those updates will be applied to any mobo bios, so theoretically all mobo’s will improve as yours did. 

[freeagent]

Are you running the newest AGESA? If so you will notice you aren't getting much more than 1.425v. Your load temps should be at least 5c cooler, give or take.. But it wont be boosting as hard as before. At least that's what I saw.

[YoungBlade]

27 minutes ago, Blue4130 said:

What took you minutes to apply took engineers and software devs hundreds of hours to make. It's no trivial thing, a bios update. Yes, bios and boards do affect performance not all 550s will achieve the same level of performance. 

 

And I think the reason why updating the bios is not recommended is because the possibility of catastrophic failure is higher than something like a ram timing tweak. 

I am not trivializing the work that goes into a BIOS update. This forum is also a product of hundreds of hours of work. The OS that I'm using is the product of thousands of hours of work. The keyboard that I'm using to type this is the product of hundreds of hours of work. The desk it's sitting on probably took hundreds of hours to make. And the Internet by this point likely has millions of hours that went into it arriving at its current state. Literally all aspects of modern society today are the products of hundreds, thousands, or even millions of hours of human labor.

 

If I had to take the time to acknowledge all of the efforts that went into everything I do and use every day, I'd have no time for anything else. It's not that I don't understand the significance of our technology - believe me, I marvel at it constantly. Our lives today are miracles beyond imagination of those who lived just a century ago. The fact that we aren't losing access to utilities every 5 minutes is nothing short of astounding and a testament to the work of countless individuals. Plenty of folks around the world today live in places where access to electricity and clean water and phone service is intermittent at best and unavailable at worst.

 

In no way am I trivializing the time that was put into a BIOS update by pointing out that it is simple from the perspective of the end-user. If anything, that's higher praise, because it means that the engineers and devs did such great work as to make it easy. It's only through their work that I didn't have to get out a specialized tool and hook up wires to my motherboard.

 

16 minutes ago, freeagent said:

Are you running the newest AGESA? If so you will notice you aren't getting much more than 1.425v. Your load temps should be at least 5c cooler, give or take.. But it wont be boosting as hard as before. At least that's what I saw.

I am running the latest non-beta BIOS, using AGESA 1.2.0.6c, and I haven't noticed a difference in temperatures, but I've been running my CPU at the point where it begins to lose clocks due to thermal limits in the winter, and it's getting warm here now. Since I was thermally soft-limited already, I'm not surprised that my thermals haven't improved.

[freeagent]

27 minutes ago, YoungBlade said:

am running the latest non-beta BIOS, using AGESA 1.2.0.6c

Whoops forgot to mention, that version also limits vcore. It doesnt matter a whole lot, as you can still put out well over 200w PPT.. closer to 240. 

[Somerandomtechyboi]

7 hours ago, YoungBlade said:

Is this idea that motherboards don't impact performance just flat out wrong and folks like Linus need to stop spreading misinformation that motherboard choice is about features and not performance?

It is true mobo can affect performance but only if you overclock, and as most ppl know cpu overclocking has been dead since 9th gen so its basically useless anyways unless you run a beefy waterloop or a chiller loop

 

Motherboard choice is mainly about features as overclocking is not someting everyone wants to do, if you are looking at oc boards then b550 unify x or aorus master as those seem to have the best ram oc capabilities and that does somewhat help with performance assuming you got good rams that do 3800-4000 cl14/13 1.5-1.75v or 4600+ cl18/17 1.6v+ (samsung b die and micron rev e/b) and you are not overly conervative with volt, but for the mostly not noticable 10% performance that nets you it really isnt worth it buying an over 200$ board and you are better off with a cheaper option like aorus pro p, tomahawk, steel legend/extreme4/pg velocita, etc.