News AMD's Ryzen 3000 Boost-Fixing BIOS Leaks out, We Test With Ryzen 9 3900X and Ryzen 7 3700X

nitrium

Distinguished
Jul 27, 2009
1,444
10
19,665
194
I'm super happy with my 3700X (coming off a i5 760) - everything feels snappier. I have been checking the boost behaviour of this CPU, and mine for sure never boosts above 4266 MHz - i.e. 133MHz shy of the advertised boost speed. I'm curious to see if the new BIOS fixes this in a real manner, but the discrepancy can be addressed with just a 3% boost increase, which doesn't seem like a huge stretch.
 

redgarl

Distinguished
Still using an MSI motherboard? You know, the BIOSes from them are taking longer to release and are usually still a work in progress. The x370 BIOSes for 3000 series are still in BETA after 2 months.

As good MSI was for B450 and x470, they are the worst for the x570 chipset.
 

redgarl

Distinguished
"That's only 0.025 MHz short of the chips' spec, but many users have reported up to 300MHz deltas, and more. "

Should be 0.025GHz or 25MHz. Is it a Beta posting?

"While we see a higher peak boost, the chip runs at a lower frequency for the majority of the test. This might be a poor implementation of the firmware for this particular motherboard, or could be due to the beta nature of the BIOS. We'll have to wait until newer firmwares are released for this, and other, motherboards to get the full picture. We've pinged MSI for comment to see if the company can reproduce this result. "

Nothing to do with AMD, more about MSI... from a Beta BIOS... I don't know why you didn't wait like Anandtech, to get the whole picture before rushing to the start line to be first and reporting this. Unfortunately, this is the trend here now... be the first and don't ask question... JUST POST IT!

"This firmware appears to be a work in progress, but shows that the bugs are being ironed out and are fixable. "

You would have known that if you read my first comment on the matter on your initial hit piece.

https://i.ibb.co/KwtmD4b/toms.jpg
 
Last edited:
Sep 8, 2019
2
1
15
0
Is there any info that except fixing boost frequencies will new agesa fix isssue with WHEA errors on PCI-E which still occur and causing totally unreliable work/gaming PC platform?
 
Aug 4, 2019
55
15
45
2
I feel like bad to write this again, but just because some users can't reach the advertised boost frecuencies, doesn't mean that all those chips are bad. Many users simple don't have the expertise or experience to deal with BIOS options, Windows intallation and drivers, and whatever unknown software that could be eating thier cpu power on the background.

Also it seems that the CPU is the only one to blame when not reaching the boost frecuecy, when motherboards, chipsets, RAM, and the CPU COOLER colud be all part of the problem many users are havíng. Just a simple loose screw on a stock or after market cooler could be leading them to believe thier CPU is defective.

Im not syaing that BIOS, SMU and microcode in general does not need to be worked out more, better, and updated. Juts pointing out the "issue" in some cases could just be the user.

Cheers
 
Reactions: King_V

bigdragon

Distinguished
Oct 19, 2011
501
13
18,985
0
Looks like the new firmware still needs refinement. I expected a little better than the observed marginal improvements. I think AMD is going to need to take a more active and engaged role in firmware for its platforms. I think Intel learned years ago that firmware is hard and partners will only do the minimum to make things work. More resources need to be thrown at this boost clock problem.

Personally, I'm thinking of migrating to a 3900X or 3700X. It may finally be time to retire the i7-4930k. Not sure if the upgrade cost is worth it though.
 

Gurg

Honorable
Mar 13, 2013
266
9
10,795
1
Notice you used a H115i cooler rather than the bundled Prism Wraith RGB cooler AMD claimed allowed for some overclocking headroom. TH gushed over AMD including that $43 cooler in the bundle. Guess this is getting serious now that these AMD cpus aren't even hitting advertised boost.
 
Last edited:

Giroro

Reputable
Jan 22, 2015
500
31
5,040
13
Now that there's proof your test methology can record the boost frequency, will you please update the 3600x core breakdown so the conclusion reflects the data?
 

joeblowsmynose

Distinguished
Jun 14, 2011
203
72
18,760
0
Was there really no interesting boost data in the first 60 seconds of the test?

Does boost performance under load over a period of time not change at all ... like it does on say a 9900k or a 9700k after you load them up and run for a minute or two?

(99900k throttles its boosting behaviour after X amount of time under full load - it seems most people aren't even really aware of this, and apparently its not a concern ... but it does allow Intel CPUs to run better in short load tests like Cinebench and most testing, but doesn't show true full load performance -- this was very apparent in PCWorld's 3900x vs 9900k testing as they only did long multicore tests to show what the 9900k performs like under a sustained load)

Anyway, we could just as easily nitpick at Intel about their full load boost behaviours - but since Intel's CPUs do hit advertised boosts I suppose it is different, but I digress.

At the end of the day a minor memory tune will get you far better performance increase than 50-100 extra mhz on a one core boost. Performance won't magically increase some huge amount when (if) AMD gets this "fixed".

From the article ...
"That was followed by Intel citing a report ..." -- Do we have a link to this "report"? Or was it someone's opinion that Intel found in a forum post and tried to pass it off to you as some legit claim? Cause I'm pretty sure it was the latter ...
 
Reactions: King_V and RodroX

TJ Hooker

Illustrious
Herald
Was there really no interesting boost data in the first 60 seconds of the test?

Does boost performance under load over a period of time not change at all ... like it does on say a 9900k or a 9700k after you load them up and run for a minute or two?

(99900k throttles its boosting behaviour after X amount of time under full load - it seems most people aren't even really aware of this, and apparently its not a concern ... but it does allow Intel CPUs to run better in short load tests like Cinebench and most testing, but doesn't show true full load performance -- this was very apparent in PCWorld's 3900x vs 9900k testing as they only did long multicore tests to show what the 9900k performs like under a sustained load)
The first and 2nd plots in the article show clock speeds for the entire test duration for an old and new BIOS. Do those not show what you're looking for?

Although I wish they'd just graph the max clock across all cores rather than each individual's core speed. The graphs look pretty messy as-is, especially for the 3900X.
 
Last edited:

Gurg

Honorable
Mar 13, 2013
266
9
10,795
1
"However, the shortcomings vary based on silicon quality, and many users have reported much larger deltas with the 3700X (up to ~250 MHz). .......... That's only 25 MHz short of the chips' spec, but many users have reported up to 300MHz deltas, and more. "

Were these TH review AMD CPUs sourced directly from AMD, or were they purchased from retailers? If directly from AMD these CPUs couldn't have been high binned chips specifically chosen to send to reviewers could they?

"That means there is more to this fix than just dialing the firmware back to the same limits as the first round of "reviewer BIOS's," which were faster than the BIOS's made public after launch."

So did AMD give reviewers binned CPUs with faster Bios than available to the public just to get positive reviews?

Could it conceivably make sense to analyze these boost deficiencies and revised Bios on retail purchased AMD CPUs?
 

jimmysmitty

Polypheme
Moderator
"That's only 0.025 MHz short of the chips' spec, but many users have reported up to 300MHz deltas, and more. "

Should be 0.025GHz or 25MHz. Is it a Beta posting?

"While we see a higher peak boost, the chip runs at a lower frequency for the majority of the test. This might be a poor implementation of the firmware for this particular motherboard, or could be due to the beta nature of the BIOS. We'll have to wait until newer firmwares are released for this, and other, motherboards to get the full picture. We've pinged MSI for comment to see if the company can reproduce this result. "

Nothing to do with AMD, more about MSI... from a Beta BIOS... I don't know why you didn't wait like Anandtech, to get the whole picture before rushing to the start line to be first and reporting this. Unfortunately, this is the trend here now... be the first and don't ask question... JUST POST IT!

"This firmware appears to be a work in progress, but shows that the bugs are being ironed out and are fixable. "

You would have known that if you read my first comment on the matter on your initial hit piece.

https://i.ibb.co/KwtmD4b/toms.jpg
Man you just can't be happy about anything can you?

I feel like bad to write this again, but just because some users can't reach the advertised boost frecuencies, doesn't mean that all those chips are bad. Many users simple don't have the expertise or experience to deal with BIOS options, Windows intallation and drivers, and whatever unknown software that could be eating thier cpu power on the background.

Also it seems that the CPU is the only one to blame when not reaching the boost frecuecy, when motherboards, chipsets, RAM, and the CPU COOLER colud be all part of the problem many users are havíng. Just a simple loose screw on a stock or after market cooler could be leading them to believe thier CPU is defective.

Im not syaing that BIOS, SMU and microcode in general does not need to be worked out more, better, and updated. Juts pointing out the "issue" in some cases could just be the user.

Cheers
No one is saying the chips are "bad" however a user should not need to know how to tinker in the BIOS to reach advertised stock speeds. Yes there is a lot you can do to improve it as most motherboards stock settings tend to be meh, same with overclocking as most boards would err on the side of higher voltage when you can tweak it to be lower, but again advertised stock settings should be what the CPU can do when it is installed with all normal drivers in a stock setting.

Yes there will be variation, hence what we call the silicon lottery. Some may only reach the minimum stated speeds others may have the potential for more. Some may be able to run on lower voltages and lower temperatures at stock while others may do so higher etc.

RAM shouldn't have much to do with the top frequency of the CPU. Cooling, yes. Thats a huge part of it but as I have said before, if you include a stock cooler then the stated stock speeds should be obtainable by that cooling method not requiring after market.

"However, the shortcomings vary based on silicon quality, and many users have reported much larger deltas with the 3700X (up to ~250 MHz). .......... That's only 25 MHz short of the chips' spec, but many users have reported up to 300MHz deltas, and more. "

Were these AMD CPUs sourced directly from AMD, or were they purchased from retailers? If directly from AMD these CPUs couldn't have been high binned chips specifically chosen to send to reviewers could they?

"That means there is more to this fix than just dialing the firmware back to the same limits as the first round of "reviewer BIOS's," which were faster than the BIOS's made public after launch."

So did AMD give reviewers binned CPUs with faster Bios than available to the public just to get positive reviews?

Could it conceivably make sense to analyze these boost deficiencies and revised Bios on retail purchased AMD CPUs?
Considering how many other manufactures of hardware have been seen doing this, hell we have seen golden sample PSUs sent to reviewers and the retail units don't measure up as well, its plausible they did.

Not saying they did, don't want people to jump the gun, but again I have seen companies do this time and time again and well AMD is a company so who knows in the end.
 

Giroro

Reputable
Jan 22, 2015
500
31
5,040
13
As a semi related side note: everybody should avoid using AMD Ryzen Master, Because holy heck did I have a weekend.

In anticipation of the new update for my 3700x I was setting up my computer to check what my current max boost was on the newest BIOS compared to launch, if PBO affected it etc. And -something- broke nearly irreparably. It is not currently possible to totally reset OC settings made with AMD Ryzen Master or to cleanly un-install the software.
What happened was my Vcore was being throttled to 1.1V and my CPU was limited to the 3.6Ghz base clock no matter what, which destroyed performance, obviously. The only thing that seemed to fix the issue was to manually reset the CPU settings with AMD Ryzen Master (which I would have to do every single boot as they kept defaulting back to the broken values).
Since it would always reset on reboot, I thoght it might be a problem with the voltage regulation on my MSI x570 board (which, BTW I very much don't recommend). I spent 2 days doing every possible combination of bios flashing, CMOS resets, reseating the cooler, pulling the CPU, removing everything overnight etc. I also did many uninstalls/reinstalls of the Ryzen Master app as well as the chipset drivers and basically anything made avaliable by either MSI or AMD. I tried a variety of 3rd party driver removal tools, although I think most of them were actually supposed to be for GPU drivers. None of this did much of anything. I "refreshed" windows, this also did nothing (except obviously ruin all the app installs and configuration I had just finished spending the last 2 weeks doing from my clean windows install with the new cpu/mobo) .
Eventually I noticed that the other account on my computer actually wasn't having this problem and the CPU and voltage seemed to be correct.
I deleted my main account and remade it as a limited user - no change. Set that account to admin - no change.
Entirely new account with a totally different name - success (maybe)! Whatever was broken, was linked to the name of the account, and stored somewhere outside of what is initialized on an account deletion/creation.
I spend a little while getting stuff reinstalled and run into an issue with NVidia drivers: Control panel is missing, get it from the windows store.... I don't know why they force you to use the windows store for that way. I try to install it - Windows store is totally broken (on all accounts). I try for a couple hours troubleshooting and trying to find a workaround because, forget the windows store, no luck, other than possibly the "standard" drivers they wouldn't let me install (thanks for a horrible software distribution strategy nvidia!)... Also it seemed like windows update was totally broken. I eventually tried some manual cumulative update to Windows out of hopes that, idk, anything changes. It did not.
It was around this time I noticed that my CPU clock Wasn't fixed, although the voltage seemed OK(?) at 14.5i-sh Volts which seemed common before ... but my CPU still wasn't boosting when using that account, it was just fixed at 4.275GHz, but occasionally after a restart 4.25GHz.
I reinstall Ryzen Master to try a 'reset' again, because Why not. I had gone back to it many times before this new account on a refreshed windows, but had long moved past letting it screw with my system. AMD Ryzen Master said there was nothing to reset, and indeed the core behavior in that app (and only that app) looked normal.

W.T.F.

At some point, the HWInfo, CPUZ, and Task Manager all lost the ability to read the CPU ratio -OR- AMD Ryzen Master was straight-up lying. Who knows.
I gave up, reformatted, and reinstalled Windows from scratch. Things seemed totally fixed but I'm taking a break from that PC for a few days.

So moral of the story, "refreshing" windows doesn't work to fix a broken overclock and will cost you a lot of time in the long run. Also you should absolutely avoid AMD Ryzen Master, and possibly any windows-based OC tool.
 

ASK THE COMMUNITY

TRENDING THREADS