The Stilt, created a custom BIOS for my motherboard (Crosshair VII) that contains the SMU version (46.49.0) that is contained in the 184.108.40.206ABBA update. I'll test it out tonight and see if there are any noticeable changes in boost behavior on my end.
They are showing the improvements we desire, Tho I am hoping with better cooling the boost frequency duration can become more.... indefinite. My processor doesnt have a singe issue remaining cool enough to run on 3 cores at 4600mhz an entire Cenebench run from my testing.
I think we can stop testing now and wait for our motherboard makers to produce the bios.
Those are actually not based on AMD's AGESA ABBA update according to him, and TBO its kinda pointless for him to put out modded bios now. He made available modded bios for the Crosshair VIII series too... on the 10th.... after AMD's blog post tomshardware put out its findings on MSI's beta ABBA bios. I'd say thats too little too late, and according to some there is problems with those unofficial bios anyway.
They aren't full AGESA 220.127.116.11ABBA, but they do contain the SMU code version (46.49.0) that is in the new AGESA. The SMU binaries contain the code that controls boosting behavior and cannot be altered by motherboard vendors, so all UEFI releases with SMU 46.49.0 should boost the same.
Naturally the UEFI won't contain other changes made in 18.104.22.168ABBA like better idle voltages.
My CPU does hit 4.6 occasionally with the beta UEFI containing the updated SMU code.
The 521 CB score almost matches the 524 I saw by setting my CPU to 4.6 manually.
Userbenchmark UEFI 2703
The new SMU definitely helps. The official UEFIs coming in a few weeks should show a similar performance uplift.
You need to know NOT to use AUTO-OVERCLOCK as it will get you an average all core fixed frequency. Test should be on "default" with clear CMOS and single monitoring softwrae only, better to use HW Info
Hope new bios solves issues
ASUS discreetly released a beta bios for my Crosshair VIII Formula, 1001 that includes AMD's AGESA 22.214.171.124ABBA and the new firmware. It provided a decent uplift to the performance in Cenebench R15 and R20, but I'm sure it could be better if they would modify or let us modify the boost duration. Preferably to infinite.
Now they need to work on that boost frequency duration, and the all-core boost in respect to thermals. They should be giving me at least 4.35ghz all-core instead of the paltry 4.150-4.2 with the voltage they are setting for the all-core. Preferably they make all of these speeds and durations manually tuneable by the end user so enthusiasts can maximize performance without needing to manual OC. Or, be actually based on the thermal solution, for as we can see from my HWinfo readout, i'm not anywhere near the limit.
The beta bios for ASUS X570 boards can be found at https://www.hardwareluxx.de/community/f12/ultimative-am4-uefi-bios-agesa-ubersicht-12-09-19-a-122890... . You'll need to use Bios Flashback with them renamed appropriately for your board as they are still Beta.
BRILLIANT ! Enabling core performance boost and Performance enhancement (Level 3(OC)) on my board actually improved the boost duration !
Multi-core score is especially improved, by a good 200 points or so in R20.
Gonna check and see if only one of the options was necessary. - Dont need Core Performance Boost, just Performance Enhancer.
I don't think we'll see much improvement there. Due to the die shrink to 7nm you probably can't put much more voltage in with risking damaging the ICs with electromigration. AMD appears to have set the safe voltage for high current workloads at 1.3V and low current at 1.5V. These voltages aren't actually high enough to reach to socket power or any thermal limit, but the CPUs are probably right at the FIT limit for voltage. So it isn't going to matter how good the cooling solution is, the chips won't be able to go past a certain level of performance. In userbenchmark, the bottom 5th percentile and top 95th percentile in benchmarks are seperated by only 11.8%. That is how consistently the 3900X performs due to the chips hitting the FIT voltage limits before any other limiter. Meaning, even someone using the included wraith cooler is going to see similar performance to someone using a custom loop (you do have a quieter system though). Overclocking manual removes those limits, but it probably isn't worth slowly eroding the ICs with higher voltage. Electomigration will only become a bigger and bigger hurdle, as process nodes continue to shrink.
So.. updated to the new bios from gigabyte
version F6a 2019/09/09 Includes AGESA 126.96.36.199 ABBA
Once updated I just loaded safe defaults, set ram to xmp and 1.35 volts and rebooted.
As the linked imgur screenshots I gained a couple mhz but am still unable to get over 4350/4360 in single core CB20 runs.
Not sure how to proceed, is ABBA the "end" of fixes? Is the issue those of us who are stuck at 4300 not known by AMD?
Has anyone rma'd their board or chip and "fixed" the issue that way?