When assembling a new PC or upgrading your CPU, a common situation to be in is: “I need to upgrade the BIOS to support this CPU, but I can’t boot the computer to the BIOS to do that.” Outside of finding an older CPU to temporarily use, many motherboards have a BIOS flashing utility that doesn’t require you to boot the PC. Just make sure you have the PSU power switch toggled on.
While benchmarking for our new 9000 series video, I noticed it can be a little fickle to know exactly what to do for these. This information in motherboard manuals are meh at best depending on the brand.
Does your motherboard support Ryzen 9000?
On the BIOS page, you’ll find an overview of all supported CPUs. Like this one from the AsRock page:
There’s a good chance not just any flash drive will do. For best compatibility, here is a list of things to look out for when picking a flash drive:
Format flashdrive to FAT32 - Windows uses NTFS for instance, so not all storage partitions are the same.
Use a flash drive that’s 8GB or less - Large storage devices are not very compatible ith BIOS even though they’re formatted to FAT32. They might work, they might not.
Use a USB 1.0 or 2.0 flash drive - Look at the color of the USB end of the flash drive. White and black are the most common colors for USB 1.0 and 2.0. If it’s blue, red, etc. It’s probably USB 3.0+. Again, USB 3.0 might work, might not depending on the board.
One way to tell the bios flash has failed is if it seems to complete very quick (in just a few seconds). The back BIOS FLASH button should blink for a couple of minutes before stopping.
Formatting your USB
You can do this in Windows by just right-clicking → format
My USB worked with these settings below:
Which IO port do I use?
Thankfully, with more and more modern motherboards, they’ve started putting squares around the correct BIOS flashback ports.
For brands like AsRock or MSI, depending on the model series of motherboard, you have to rename the .ROM file. For the Taichi that I was flashing, this was “CREATIVE.ROM”. Once renamed, drop the file onto your flash drive. What you need to rename the BIOS file will change depending on what series of motherboard you have. The specific name should be in your motherboard manual.
ASUS has a specific “file renamer” you have to use before copying the file to your USB. You can download this from rog.asus.com
Have you tried enabling PBO? DerBauer has shown massive performance improvements by simply removing the 65W power limit on those parts. Perhaps running them at the same limits as 7700X and 7600X would be interesting.
Would love to know how the 7700 and 7900 vs their new counter parts when they are released being 9700 and 9900 (non X). When using them in Eco mode, how much of a power saving difference is there. Currently got the 7000 series ones in a colocation, and if the newer 9000 series is quite a bit lower, then would help not having to pay the extra power costs, as I’m just under the 1 amp usage, and woud need to pay 2 amp if it starts going over 1 amp on average. So switching processors could be a good ROI, and then sell the old 7000 series.
Shall have to keep a look out if anyone does any power testing with the CPUs in eco mode, and their results.
Half the answer’s already available since the 9700X is, like the 7700, 65 W TDP. Wendell and quite a few others put up 9700X reviews yesterday, see Phoronix and TechPowerUp in particular for power efficiency. The drop to 4 nm seems likely not to be what you were hoping for.
For the 9900X you’ll get, I dunno, another quarter of the answer seemingly next Wednesday as it’s 120 W TDP.
Hey there, quick stupid question about pet peeve of mine, how are uefi boot times and memory training times?
I am on AM5 with 2x16 Dimms and even with memory context restore on its unpleasant. Is there any change due to more mature uefi+agesa or due to better IMC on zen5?
Kind of hard to find a 8GB or smaller USB 2.0 flash drive these days unless you already have one from olden times or dig around a recycle/trash bin. Could be just me, but I have never encountered issues using 16-32GB USB 3.0 flash drives on my AM4 Gigabyte boards with flashback functionality.
It might be the certain board I was flashing, I tried a 3.0 32GB and was unsuccessful. Of course you can try whatever you have, I was just listing “best compatibility” options
Really, couple of seconds after training? How would you describe it in detail? Would you mind recording with stopwatch and doing cross comparisons with multiple boards? Video with stopwatch in frame alone would be enlightening.
Even better the new fast board and ryzen 7000 series cpu?
These boot time are driving me nuts and I dont know what is exactly the problem or better how off my boot times are from general experience due to lack of any hard data.
Is it my board? Is it gigabyte uefi implementation in general, or just this b650 board?
Only thing I have tried is hard cmos reset via screwdriver.
EDIT: Why I am so focused on this:
a) its friggin annoying, especially when you try to quickly iterate over some setting that require reboot
b) older platforms were extremely faster without any tweaking (UEFI capable early zenbook for example):
And that was pretty ancient ivy bridge dual core cpu. Modern dell xps suffers from massive firmware slowdows too, regaredless.
I.e. its infuriating how slow uefi is doing whatever it does, and its completely unprofileable.
Maybe DDR5 training with MCR is not even as much culprit as we suspect.
Have not tried any AM5 yet, but in AM4 my Gigabyte boards are noticeably faster than my MSI boards. It could be just the UEFI implementation on the boards.
No problemo, I did cmos reset and it seems the cumulative agesa updates brought some improvement to memory training times. No hard data on my initial observations when setting this up year back or so.
It just seems this board baseline optimization is very shitty. Question is how shitty are these boot times compared to other AM5 boards from 7000 era? Or Intel boards? Both have to deal with memory training.
Below measurements are from fedora 40 boot target, emitted by systemd-analyze. I boot via UKI and fedora is system default target, so there is not inserted delay by user interaction. Simply power on and then login once able (unlike windows partition).
scenario
notes
firmware boot time (s)
user commentary
starting baseline on up to date firmware
expo off since melting x3d chips, fastboot on, no logo
27.32
full cmos reset
expo off by default, fastboot off by default
27.32
wtf?
turning on fastboot
fastboot level = fast
19.549
turning fastboot to max
fastboot level = ultra fast
19.076
turning off boot logo
fastboot level = ultra fast AND no logo
19.054
disabling boot logo is pointless
reset bios to optimized defaults + apply fastboot = ultra fast
reset to default, fastboot level = ultra fast
19.058
no need load optimized defaults anymore
lets disable memory context restore
last + MCR = DISABLED
34.666
much better than on release agesa versions
and now explicitly enable memory context restore (measurement on second boot after)
last + MCR = ENABLED
26.597
something changed, but measurement does not make sense
I strongly suspect something fucky is going on with this boards firmware as measurement above show.
Or there is some factor I am missing. It seems to me that either some setting are applying inconsistently or there is strong and unpredictable variability in memory training.
