AMD's Ryzen 3000 Boost-Fixing BIOS Leaks out

I installed in on my X570 ACE and performance seems to have returned to roughly what it was with the launch bios (after decreasing the last few releases).

A few quick benches with the "leaked" bios installed are below. The only thing I did was enable XMP and adjust the fans down to ~40% to keep noise in check.

99capture.PNG aida99.PNG cpuz.PNG

The bios itself seems a little rough around the edges, but I admittedly haven't spent much time playing with it yet.
 
  • Like
Reactions: erek
like this
Waiting for some detailed testing of the boost behavior of the 3800X after this change. I decided to save some money and pick up a 3700X because there was no real benefit to the 3800X outside of all-core overclocks, but I returned it to wait out this change.
 
I installed in on my X570 ACE and performance seems to have returned to roughly what it was with the launch bios (after decreasing the last few releases).

A few quick benches with the "leaked" bios installed are below. The only thing I did was enable XMP and adjust the fans down to ~40% to keep noise in check.

View attachment 186159 View attachment 186160 View attachment 186161

The bios itself seems a little rough around the edges, but I admittedly haven't spent much time playing with it yet.

What’s “rough” about it?
 
What’s “rough” about it?

There’s no longer an option to select amongst XMP profiles (used to have two to pick from, now just one). Thankfully the one that is there seems to work fine.

Also had some weirdness with setting fan speeds in BIOS at first (one of my CPU fans won’t go as low as it used to), but otherwise works OK.

Perhaps describing it as rough was a bit too harsh!

Edit: Now that I think about it the XMP change might be intentional. It took away the profile that disabled geardown mode, which would cause my computer to fail to boot. I bet that was causing a ton of customer support issues.
 
Last edited:
Toms mentioned that the temperature threshold was moved back up, can you check if it runs hotter when loaded ?
They also mention that you need specific version of Windows and updates for the Windows scheduler, I find that interesting.
 
C6H bios had the new SMU inserted in most recent released beta by The Stilt and it looks rather promising. PBO actually did something for all core OC, I am getting ~100 MHz more in multi-threaded applications, CPU stays around 4.2ghz vice 4.1ghz from all previous bios's. Also with default PB2 I got my highest CB20 single thread score of 529. Plus highest multi-threaded result with CB20. I hope the official bios will take it up even further.

Single Thread Test​

2019-09-10 (13).jpg

Multithread score plus you can see the boost of 4.6ghz while running the single thread benchmark
2019-09-10 (15).jpg

C6H Bios with new SMU (Note: Not Official)
https://www.overclock.net/forum/28120992-post43001.html
 
Last edited:
Gigabytes forums have an F7a beta bios available but it mentions nothing about new SMU microcode.

Thus, I am going to wait before flashing it. But stay tuned. Gigabyte will have it soon I am sure. I was hoping to get more than 4075mhz all core on my 3900x. I was hoping for around a 4200 all core.
 
Gigabytes forums have an F7a beta bios available but it mentions nothing about new SMU microcode.

Thus, I am going to wait before flashing it. But stay tuned. Gigabyte will have it soon I am sure. I was hoping to get more than 4075mhz all core on my 3900x. I was hoping for around a 4200 all core.
I was getting a little bit over 4.1ghz now 4.2ghz with CB20 benchmark with the new SMU 46.49.0. So it shows potential and that was using PBO with +200mhz clock but everything else in Auto. Still a lot more experimenting to be had. This is with an UnOfficial, user modified bios (The Stilt) which he indicated that he only changed out the SMU code for the bios - meaning to me it may be leaving out hooks and settings new to the SMU in the bios. We just have to see how the different motherboard makers make use of any changes inside the bios.
 
Tested how the new boost behavior was in Resident Evil 2, I think it is rather good. Game looks like up 6 threads and Windows puts those threads on 6 cores. Boost rapidly switches the cores maintaining from my prospective high clock speeds. At least higher than I think I could OC all cores so to speak. Since more than one thread is being used Boost clock will not be as high but in many cases it is over 4.5ghz:

 
  • Like
Reactions: Nobu
like this
This is CB 20 rendering showing boost behavior with all cores. Initially it is over 4.3ghz but as temperature goes up it goes down. To get boost to drive all cores to over 4.3ghz requires some rather cool conditions is what I am finding. Another video here:


 
This is CB 20 rendering showing boost behavior with all cores. Initially it is over 4.3ghz but as temperature goes up it goes down. To get boost to drive all cores to over 4.3ghz requires some rather cool conditions is what I am finding. Another video here:




Is that stock settings? PBO on or manual OC, if not?
 
Gigabytes forums have an F7a beta bios available but it mentions nothing about new SMU microcode.

Thus, I am going to wait before flashing it. But stay tuned. Gigabyte will have it soon I am sure. I was hoping to get more than 4075mhz all core on my 3900x. I was hoping for around a 4200 all core.

I just run mine at 4.1 all core manually and let it crunch away.
 
Is that stock settings? PBO on or manual OC, if not?
That was PBO. PBO does seem to do more for multi-core then before but not much for single. Temperature really affects the boost speeds, even lower temperatures like 65c can start affecting the boost you see for multi-core.
 
i do not have a 3000 series processor at the moment ,but the new bios has had a effect on my 2600.i have it set to a manual overclock and the power settings are set to microsoft's balanced power plan not the ryzen plan. I now have 2 cores that constantly stay at 4.025ghz and sometimes as many as four.this much better than letting the processor boost on it's on
 
i do not have a 3000 series processor at the moment ,but the new bios has had a effect on my 2600.i have it set to a manual overclock and the power settings are set to microsoft's balanced power plan not the ryzen plan. I now have 2 cores that constantly stay at 4.025ghz and sometimes as many as four.this much better than letting the processor boost on it's on

realistically it doesn't matter what power plan you use.. the only real differences between ryzen power plan and windows power plan is that the windows one took longer to pull cores out of their parked state which effected benchmark scores. supposedly there was more added to the power plan for zen 2 but honestly don't see much of a difference.
 
I'm seeing a substantial difference in some benchmarks and the Hardware unboxed video is showing meaningful changes in some games. I am pleased with the update and satisfied with my chip.
 
realistically it doesn't matter what power plan you use.. the only real differences between ryzen power plan and windows power plan is that the windows one took longer to pull cores out of their parked state which effected benchmark scores. supposedly there was more added to the power plan for zen 2 but honestly don't see much of a difference.

Switched to the Ryzen power plan and all 6 cores went to 40.25 ghz and would not idle back down to 1.5ghz or there abouts.CB20 in single core went up only 6 points. I don't think that will make any difference. Going to leave the power plan with W10 balanced power plan as both clock speeds and processor voltage's go down.
 
I'm seeing a substantial difference in some benchmarks and the Hardware unboxed video is showing meaningful changes in some games. I am pleased with the update and satisfied with my chip.
I'm surprised there was a real.world difference, so good on AMD for not just brining the frequency up a tiny bit, but actually making it meaningful beyond a single boost #.
 
Switched to the Ryzen power plan and all 6 cores went to 40.25 ghz and would not idle back down to 1.5ghz or there abouts.CB20 in single core went up only 6 points. I don't think that will make any difference. Going to leave the power plan with W10 balanced power plan as both clock speeds and processor voltage's go down.

You might want to check actual power draw with a watt meter. Mine does the same, however there was no measurable difference from the wall. AMD's low power states work differently these days.

I took the Ryzen power plan because it consumed same power at idle, yet performed better.
 
so for shits and giggles decided to try the new bios from gigabyte with my 3600..

R20 single core went from 482 @ 126w to 486 @ 117w
R20 multi went from 3609 @ 180w to 3696 @ 189w

all scores are repeatable within +-1 point

single core clocks stayed the same at 4192mhz but seems to only bounce the load between Core 3 and 5 and sometimes core 0 instead of cycling through all 6 cores like it did on the previous bios.


so the changes to idle voltage are definitely working now. at the end of the day it doesn't really change anything for me but i'm happy with the idle power usage changes. probably going to mess with auto overclock and pbo on monday/tuesday when i don't have to deal with my work sleep schedule.
 
New BIOS just showed up on ASUS website for the X570-E board. Running my "before" Cinebench/HWInfo tests now.
 
ok, just got done testing, in a nutchell, not a whole lot has changed.

CB20 single core was up about 15 points, which is well under 1% so well within the margin of error
CB20 Multi was down about 100 points but that could be because I had more monitoring tools up

PBO still does nothing

New updated did allow me to "see" ALMOST 4.6GHz, it got up to 4.5xx on 3 cores for a split second. When I say split second I mean just that, I didn't actually see it in real time, but HWInfo reported it under the "max" column. New updated does not appear to be able to maintain any higher clocks in single or multi core loaded instances.
 
I just updated to 1001 on my C8H, fuck all difference as far as I can tell, if anything, boost behaviour is a little worse. Also, something that's been bugging me since I built this machine. In Ryzen Master, C06 on CCX 1 is tagged yellow (gold if you will). My understanding is this is supposed to be the fastest core. However C01 and C02 on CCX 0 are the ones boosting. Doesn't make sense to me. Voltages all over the place too, seeing 1.46V in Ryzen Master, which broadly matches HWinfo. Max clock in HWinfo is 4342MHz during R20 single core.
 
I just installed f7a aorus master bios last night

I'll do testing tonight after workouts over.
 
I'm seeing a substantial difference in some benchmarks and the Hardware unboxed video is showing meaningful changes in some games. I am pleased with the update and satisfied with my chip.

Where are the before and after screenshots?
 
I just updated to 1001 on my C8H, fuck all difference as far as I can tell, if anything, boost behaviour is a little worse. Also, something that's been bugging me since I built this machine. In Ryzen Master, C06 on CCX 1 is tagged yellow (gold if you will). My understanding is this is supposed to be the fastest core. However C01 and C02 on CCX 0 are the ones boosting. Doesn't make sense to me. Voltages all over the place too, seeing 1.46V in Ryzen Master, which broadly matches HWinfo. Max clock in HWinfo is 4342MHz during R20 single core.

mine doesn't use the so called "golden" core either, not really sure how it decides which is the best one, if it just requires less voltage or what but mine tends to stick to core 3 and then switch to 5 and then occasionally it switches to core 1(which it claims is the good one) for a couple seconds then goes back to core 3. i doubt windows scheduler can even tell the difference but i guess you at least know that if you want to manually lock a program to a specific core then that would be the one to lock it too, i dunno.
 
I ran Prime95 Blend Test overnight and checked the results this morning. Went from zero cores hitting 4.4GHz on my 3700X to five core hitting the targeted 4.4GHz boost speed on an ASUS Strix 570-E running the official ABBA release. This is on the stock Wraith Prism cooler as well, pretty happy with the result and glad AMD addressed this.
 
  • Like
Reactions: N4CR
like this
I ran Prime95 Blend Test overnight and checked the results this morning. Went from zero cores hitting 4.4GHz on my 3700X to five core hitting the targeted 4.4GHz boost speed on an ASUS Strix 570-E running the official ABBA release. This is on the stock Wraith Prism cooler as well, pretty happy with the result and glad AMD addressed this.

Could you upload that Prime95 I want to try out that one.
 
I'm gonna p95 mine later. I'm running 7x120mm radiator lol. I predict that I have excessive cooling but Temps will still hit 80c.
 
I have the F6a BIOS on my x570 Ultra. It's stable. Boost clocks are definitely higher, but I'm still not hitting 4.4 on any cores. 4374 is the most I've seen on only one core. The rest are in the 4200 range. CB20 scores are 476 single core and 4796 multi.

I'd love to know why I can't control CPU voltage properly. For instance. If I leave CPU voltage on auto I get nice low idle, around 1.3v all core load and around 1.45v single core load. BUT, if I set my voltage manually to, say, 1.4v, shit gets all fucked up. Idle voltages are fine, but single core load voltage goes to 1.29v and multi core load it goes to 1.1v, which results is completely shit clocks. I just don't get it. Why would my voltage be so low if it's set to 1.4v?

This new Ryzen platform confuses the hell out of me.
 
Back
Top