AMD Firmware Update To Bring Boost Performance Optimizations

[phoronix]

Phoronix: AMD Firmware Update To Bring Boost Performance Optimizations

There has been a lot of talk recently of AMD Ryzen 3000 series processors reportedly not hitting their boost clock frequencies, whether stock coolers are adequate for hitting the boost frequencies, and other concerns around the boost behavior on these new Zen 2 processors. AMD issued a statement today they are rolling out a new BIOS/firmware update to help with boost clock frequency optimizations...

http://www.phoronix.com/scan.php?page=news_item&px=AMD-Zen2-Boost-Optimizations

[ms178]

It was quite a hot topic during this summer (der8auer and HardwareUnboxed covered this quite extensively), let's see how it turns out.

[birdie]

The whole Ryzen 3000 launch has been a myriad of screw-ups so far.

• CPUs not hitting their advertising boost clocks (not even a single core)
• Buggy X570 BIOSes
• X570 motherboards "feature" horrible chipset coolers and insane chipset power consumption (still no comment from AMD on this one)
• Weird/very high CPU idle power consumption (~20W at idle as reported by HWInfo64)
• Weird/very high idle/load CPU voltages - 7nm transistors @ 1.5V voltage ... what could go wrong? Oh, Ryzen 3000 CPUs are already dying.

Oh, and we've had Linux specific issues, like

• No ACPI tables for Linux (check for ACPI: [Firmware Bug]: BIOS _OSI(Linux) query ignored in dmesg)
• the k10temp driver not working at all (probably fixed in yet to be released kernel 5.3)
• motherboard sensors not working due to ACPI region conflicts (to which lm-sensors developers say you must use a kernel boot option acpi_enforce_resources=lax - wow! as if average people even know how to do that)
• lower multithreaded performance than in Windows 1903
• a lot higher power consumption than in Windows (no idea when and if that'll be fixed)

However Linux users might not care because GeekBench shows that barely anyone uses Linux to run Ryzen 3000 CPUs - from what I've seen less than 0.5% of people submit GeekBench results under Linux. It speaks volumes about Linux adoption but I'm pretty sure quite a few apologists will reply to this message and explain to me how and why I'm wrong and how Linux has over 10% on desktop.

What a fine day!

[shmerl]

The whole Ryzen 3000 launch has been a myriad of screw-ups so far.[*]motherboard sensors not working due to ACPI region conflicts (to which lm-sensors developers say you must use a kernel boot option acpi_enforce_resources=lax - wow! as if average people even know how to do that)

Do you have some link where this is discussed? Sensors work for me, but some readings look a bit fishy. For example chipset temperature never goes above 60°C which feels capped to me, since it shouldn't be so exactly flat.

[L_A_G]

Well this response seems rather odd considering this has previously been explained as AMD deciding to intentionally limit boost frequencies to improve the longevity of chips. Not sure if it was AMD themselves who explained the rather lackluster boosting like this, but it at least made sense even if it was a bit annoying considering the Zen1 core parts boosted quite nicely.

In the context of this latest update on the matter I get the feeling AMD may have just decided to backpedal on this and come up with the "firmware issue" as an excuse.

[birdie]

Do you have some link where this is discussed? Sensors work for me, but some readings look a bit fishy. For example chipset temperature never goes above 60°C which feels capped to me, since it shouldn't be so exactly flat.

Check https://github.com/lm-sensors/lm-sensors/issues

[HenryM]

sure would be nice to have a board that didn't rely on partners for firmware updates. most boards over two years old seem to get ignored, though Asus seems to be better about it.

[AsuMagic]

Weird/very high idle/load CPU voltages - 7nm transistors @ 1.5V voltage ... what could go wrong? Oh, Ryzen 3000 CPUs are already dying.

Wasn't that proven to be monitoring software incorrectly measuring voltage at idle (and measuring spikes of activity instead)?
Cause I can bet that the CPU ain't gonna consume only 20W at 1.5V...

[AsuMagic]

Do you have some link where this is discussed? Sensors work for me, but some readings look a bit fishy. For example chipset temperature never goes above 60°C which feels capped to me, since it shouldn't be so exactly flat.

Maybe that's when the fan kicks in? If it reaches 60°C but never ever reaches it though it is surely a cap.

[shmerl]

Maybe that's when the fan kicks in? If it reaches 60°C but never ever reaches it though it is surely a cap.

The fan is always active. In fact in the UEFI UI, that temperature fluctuates above 60°C a bit. But through sensors - only 60°C or lower.