Update: There is new information on top of this article in the following link.
Update on AMD Ryzen 3000 Boost Clocks and more – somewhat technical
There has been a lot of controversy on Boost Clocks and how they behave across different AGESA based BIOS. Some AGESA BIOS versions not working properly, as reported in Anandtech and their BIOS version NPRP versus release for X570 ACE.
First, as Steve from Gamers Nexus said and I add myself too to his opinion, it would be nice to cut some slack reviewers that were affected on launch date. I have not even completed my review on RX 5700 XT to dedicate myself observing two anomalies mentioned on my review on Ryzen 9 3900X.
To bring up some light between behaviors, TheStilt has shed some light on differences between max boost clock per clock, but let’s see the odd behavior on some BIOS first.
This picture is from Anandtech and his tweet confirming some issues with NPRP BIOS compared to new BIOS release to public 7/7 on X570 ACE. They saw the core difference and had to re-do all their testing.
Returning to TheStilt’s information (you can check here) there is pretty much a high delta difference between the best two cores and rest of them, up to 300 MHz ~ on 3900X. According to his testing, his sample did not reach 4.6 GHz as marketed on the motherboard he used on latest AGESA 1003AB, being capped at 4.575 GHz.
“The two best cores reach 4.575GHz, while the ten other cores reach 4.325 – 4.4GHz peak.”
If you were wondering what is a safe voltage for high current loads, 1.325v is safe according to FIT measurements done by TheStilt, and single thread up to 1.47v.
He did mention the screen capture on our review was a product of some hack GBT been using, but since then, another user has replicated our results, with all cores boosting to max with the same AGESA code 1002 on a different board on another brand (ASUS X470 Crosshair VII Hero) on a Ryzen 5 3600 board.
A word on the VID on X570 AORUS XTREME; it is using a new type of PWM controller with true 16 phases and I am not sure if the VID voltages presented are actually high or normal, but that’s VID, which is not the same to VCore (voltage actually fed to the CPU).
We will ask some comments on AMD if the behavior or worst cores compared to best ones, have a delta up to 300 MHz. Also one thing to point out according to TheStilt’s test, there might be variance between chip to chip (silicon quality).
We have not done too much tests with our chip, but we believe our sample on All Core OC probably works out fine at 4.3 GHz, same as Gamers Nexus test on his.
However, TheStilt’s sample with his validation method for OC, 256-bit workloads, show safe voltage all-core overclock only reaching 4.125 GHz.
PCIe WHEA errors – NVIDIA cards
We reported WHEA errors on our review and we do not know if it affects numbers for other reviewers, or they found those errors. These still exist even on newest bios and latest chipset and we are working with NVIDIA LATAM, providing data to see this resolved ASAP with our two chip samples. Using both X570 GODLIKE and X570 AORUS MASTER/XTREME X470 A7GW resulted with same behavior as we mentioned on our review.
We are being sent an ASUS C8H and see if behavior remains the same. As soon as we get updated information, on the WHEA errors presented on NVIDIA cards we will have and update, but we are debugging those errors bit by bit (yet again).
As for our review, we have decided to e-mail AMD and hopefully to receive an official reply from them to see which is the official boost behavior (AGESA 1002 brings ourselves some doubts) for the new chips and if the Delta between the two best cores from the rest, can get up to 300 MHz~. Doing a quick test on CBR20 (1002 and newest 1003AB) yielded same or slightly better results.
If this the final behavior is as TheStilt has mentioned and the delta is as it is on the new Ryzen chips, we will see start doing our final test and even probably re-do productivity workloads. I am pretty much burnt out (yesterday finally we benchmarked RX 5700 XT) as the rest of reviewers, but well, we will inform any updates.
We have done a quick test CBR20 Multi-Thread and Single Thread on the new BIOS (does not reach max boost on all cores) but performance is pretty much the same, or somewhat better? Could be within margin of error. Waiting on those PCIe errors to dissapear to start testing on our side for one final time.
To end this article, did not really want to stir up commotions, just posted the oddities at the moment of my testing.
If you liked this article, you help us a lot by doing like, share. If you have any questions, you can follow us on our social networks: