Testing boost frequency on AGESA 1003 ABBA
It’s been a while since I have written in English, but after Ryzen launch, I decided to take somewhat of a break and check on my health. Nonetheless, most people are aware about the impact Der8auer made with his survey, which made AMD speak at last about the boost issues we were seeing since day one launch.
Before we start on boost frequencies, WHEA errors were corrected with BIOS around the 12th August and I have primarily tested on 500 series boards, to be precisely on a X570 I AORUS PRO WIFI board (great little board, eventually should be reviewing it).
So yes, WHEA errors are GONE, so let’s talk about 1003 ABBA testing.
To begin with, the final public BIOS with new AGESA 1003 ABBA, should be released in 3 weeks due time (which might hit the mark of the rumored 30th September launch of the Ryzen 9 3950X) or maybe a little earlier than that.
Nonetheless, since BETA BIOS for GIGABYTE boards were made public on 500 series boards and AGESA 1003 ABBA and SMU 46.49, we already done some tests on the new BIOS.
The results look promising, since my retail chip of Ryzen 5 3600 never went above 4.125 GHz whatsoever on different boards/chipset. Was about the send to “warranty” back to Amazon, but decided to keep it for “testing” the new fix.
Screenshot Ryzen 5 3600 – X370 C6H
With the latest BIOS for the C6H, the chip did not go above 4.125 GHz as seen on the screenshots. The same happened with 500 series boards I tested with GIGABYTE.
This chip, would not boost above 4.125 GHz on two brands of motherboards and different BIOS version. Tested on 300 and 500 series.
Screenshot Ryzen 5 3600 – X570 I AORUS PRO WIFI – AGESA 1003 ABBA
Just like magic, new BIOS works great, one CCX seem to be boosting to 4.2 GHz on all cores (although there is preferred core inside this CCX) while the other CCX goes up to 4.15 GHz (which is fine, not all CCX are equal).
So that fixed boost being stuck from 4.125 GHz to 4.2 GHz. Moreover, compared to some internal testing I did with other BIOS (non ABBA) on single core workload (CBR20 SC) it now sustains pretty much all the time 4.2 GHz, something that did not happen before.
Information of setup on X570 I AORUS PRO WIFI
XMP >> Enabled
PBO >> Disabled manually on both options (yes there are two options)
Auto OC >> Disabled
Ryzen 5 3600 (same chip with issues) on AGESA 1003 ABBA and SMU 46.49 boosting correctly. Improvement of +75 MHz. By the way I’m dumb and did not disabled Spread Spectrum (hence the 99.8 MHz instead of 100 MHz).
On prior BIOS, while 4.125 GHz was best single core performance, when testing CBR20 SC it went down to 4.075-4.1 GHz.
TLDR; things are looking good, but final behavior may vary (hope even better) on the final release. There’s some good road at the end of the bumpy road Ryzen 3000 series has been (for those who read our R9 3900X review).
Auto-OC seems to be working now
So, after testing on stock (just XMP and PBO and Auto OC disabled manually) I decided to test Auto OC. I’m not sure the reason why my chip did not boost to 4.125 GHz (maybe it ain’t great chip) but the Auto OC function seems to be working. The gain is not much though (+25 MHz)… Can’t complain too much on this and probably need more testing with more chips (which I do not have) to see if Auto OC works fine and how much MHz extra it boosts with different CPUs within the same SKU.
Yes, did not disabled Spread Spectrum for screenshot. Auto OC + 200 MHz (tried different settings). Gained +25 MHz on all cores with Auto OC on. Tested CBR20 SC, it mantained 4.225 GHz most of the time.
Well I’ll give some data I have been tracking on the most accurate sensor I consider for the X570 I AORUS PRO WIFI, for those interested on voltages and this BETA version in particular.
Idle – 0.3v
CBR20 SC – 1.416-1.44v
CBR20 Multi – 1.36-1.38v
P95 Non AVX – 1.344v
P95 AVX – 1.244v
Bumpy road, a light in the end
Well, we sure reported some odd boost behavior in different BIOS and CPUs not reaching advertised boost. Der8auer did a great job making it viral so at last AMD would speak about the issue, so kudos for him. One thing as he said and we did on our day one coverage, was to report any issues seen at review and I think it is the way to go.
From our media and our position that criticized the launch and issues Ryzen had and few media covered, I guess it means there’s a light in the end on the testing we have done. Still, since they are not final public BIOS (Tweaktown forum has BETA BIOS with 1003 ABBA) there might be some changes. Still, seeing my retail Ryzen 5 3600 boost fix makes me quite happy.
Guess I was lucky I was within the 50% of R5 3600 users who had issues with boost (if we take Der8auer’s data) so I can show you the change with this new AGESA version. Hopefully next Ryzen launch (R9 3950X and Zen 3) will be much smoother.
Extra: CBR20 SC, 4.2 GHz hitting frequently, polling rate 500ms. (Oh yes, disabled Spread Spectrum for this one, 100 MHz on BCLK).
One last note, this is not directly a statement overall performance uplift (although in this chip particular, SC was really bad). For the data, need to compare prior BIOS and benchmark results and see uplift or regression with new BIOS.
Si te gustó este artículo, nos ayudas bastante haciendo like, share. Si tienen alguna duda, pueden seguirnos en nuestras redes sociales: