New AMD Chipset Drivers Tested on Ryzen 9 3900X

[W1zzard]

Just a few days ago, AMD commented on the idle clock behavior of their new Zen 2 processors, which has been criticized by many. The company also released a chipset update for mitigation. We thoroughly tested the new version in our application and gaming test suite using the Ryzen 9 3900X on X570.

Show full review

 

[unluckymsk]

All i need to download new chipset driver? Or use ryzen balanced power plan in win 10 1903 also? Ryzen 5 3600 Here

I have new chipset drivers for my Mobo (Asus Prime X370-PRO) Version 19.10.22 2019/07/31 available

 

[Metroid]

Thanks for the review. No major changes, so not updating it just yet.

 

[W1zzard]

Or use ryzen balanced power plan in win 10 1903 also?

Yes you also need to use Ryzen Balanced Power Plan, which automatically gets installed and activated on chipset driver install

 

[bug]

Man, that "boost clocks" graph looks like it never ends

 

[TheLostSwede]

I actually gained quite a bit in CB20, although I also just got the new AGESA 1.0.0.3ABB for my board, which might have improved things.
5048 is with PBO, 5008 is without and 4824 is prior to new driver and AGESA.

 

[Axaion]

dont know about you, but with the new chipset driver my ryzen balanced defauled to 95% minimum processor state, causing me to be stuck on 1.45v on idle.

just setting it to 5% there lets me idle at 0.93v, dropping idle temps by 6c


This is on a 1.0.0.2 Bios and Windows 10 1809, (1809 cause, lol, creative drivers, and dpc)

 

[W1zzard]

95% minimum processor state

We will also be adjusting how the processor behaves in low or idle workloads. If a processor core is not power gated and sleeping, the processor core will sit at 99% of base clock for low or idle workloads. For example: an AMD Ryzen™ 7 3700X base clock is 3600MHz, so you would see “idle” frequencies around 3564MHz ± 0.08%. This 99% value keeps the active core on a razor’s edge so that your non-trivial applications can easily trigger CPU boost. While boosted, the processor will still control frequency selection in 1ms intervals.

 

[voltage]

This is not the first time that this has happened with their drivers

 

[Ferrum Master]

People still do not get the idea how these CPU work and bark, interfere with it's work...

 

[Axaion]

So what, all CPU's are supposed to "idle" at full baseclock almost, with 1.45v?, that doesnt sound right
Of course, i checked with ryzen master, cpu-z, Hwinfo64, by themselves without any of the others running meanwhile

was under the impression that it should go sub 1 volt for idle?

 

[TheLostSwede]

So what, all CPU's are supposed to "idle" at full baseclock almost, with 1.45v?, that doesnt sound right
Of course, i checked with ryzen master, cpu-z, Hwinfo64, by themselves without any of the others running meanwhile

was under the impression that it should go sub 1 volt for idle?

It does, just change your power mode from Best Performance to one of the other two and it drops.

 

[hzy4]

When not gaming or working, the average PC enthusiast's machine is running a web-browser, a chat application, peripherals application, and a monitoring app which polls hardware. The processor could interpret their load as a need for boost, if a processor and OS are talking to each other at 1 ms. This gives users the appearance of a processor that's stuck at a higher state than idle.

So AMD wants to tell us, the CPU is running under 0,7v at idle, we just cannost see it because of the 1ms update interval?
I am on the auto voltage in BIOS, newest BIOS, with the newest chipset driver, selected Ryzen Balanced power plan, Windows 10 1903, running Steam in the backround, using Ryzen Master to watch the voltage. I have never seen it to drop below 1V. I was using modded windows without the bloatware, without any monitoring software/MB software, AIO software...etc and then I saw 0,7V in CPU-Z. I dont think its even possible to let the chip do its idle thing like AMD intended with Windows 10 backround processes, steam and AIO, motherboard software running in the backround.

 

[Ferrum Master]

Forget the monitoring softwares, screw them all. It will be a rule for newer CPU's to have better powersaving, clocking and agility. Our perception of things is too slow vs the machine code execution rate.

Only way is to use dedicated voltage meters attached to the VRM itself with log function, then you can plot it in a graph basing on activities you did in that timeframe. Not using it via software layer... it is useless either way, as it doesn't represent the real voltage, but only the VID state.

 

[Imsochobo]

Thanks for the review. No major changes, so not updating it just yet.

There is one major change, voltage is 1.3 V for instance instead of 1.45V at idle.
it silences the internet who knows what right and wrong better than the engineers who's made the product.

 

[bug]

There is one major change, voltage is 1.3 V for instance instead of 1.45V at idle.
it silences the internet who knows what right and wrong better than the engineers who's made the product.

You hate it when people do that, don't you?

Are you gaming in 4K?

Tried switching to a 27UK650-W this year, but it didn't work out, kept running into some bug that LG swears is the monitor.

www.techpowerup.com

Hitman 2 to Get DirectX 12 Renderer Through a Patch Later Today

the results are very repetitive across all sites,may be the benchmark skews the results in favor in nvidia watching actual gameplay comparisons on youtube, RVII still is outperformed by 2080 and 1080Ti,but nowehere near what is shown in the benchmark. this is just sad.should be pulled from the...

www.techpowerup.com

 

[HenrySomeone]

Hehe, with this launch it seems AMD decided that since they weren't going to fall as far from the hype as usually (apart from the frequencies, of course), to just leave beta testing to the general population and somehow they are mostly getting away with it. I can't imagine if Intel did something similar - Internet would probably explode

 

[EarthDog]

For games, less than 1.1x% and less than one percent is an improvement? Seems the same to me considering run variance, etc...

Doesn't look like this does much at all outside of a few select applications. Otherwise its margin of error....

 

[bug]

Hehe, with this launch it seems AMD decided that since they weren't going to fall as far from the hype as usually (apart from the frequencies, of course), to just leave beta testing to the general population and somehow they are mostly getting away with it. I can't imagine if Intel did something similar - Internet would probably explode

The CPUs are fine. It's just RDRAND that's messed up, but since it's a new instruction, there isn't much code out there to make use of it.
It would have rendered my Linux install unbootable, since I like to keep up with software changes, but I haven't switched yet.

 

[Jism]

The interval for changing clocks, voltages and multiplier works so fast that there is no tool that is able to keep up let alone show the correct graph. Only ryzen master is capable of that. 1.45v at idle is perfectly normal too. Light weight workthreads need more voltage basicly and its SAFE for the CPU. Its not safe for the CPU when the same voltage (1.45V) is applied and your running a all core load. Thats where the CPU´s are being damaged. If you run a all core workload that same voltage will lower to around 1.34v. Thats normal. But a stock CPU does what it is supposed todo.

So much misinformation on the internet on how the new Ryzens work. Their power states are brilliant. You dont need manual overclocking as well. Just ramp up a cooler thats sufficient to keep the CPU under 60 degrees and trust me, youll have constant boost states when having the PPT limits and such extended.

My 2700x hangs all the time on 4150Ghz all core boost, 4.35GHz single core. Its because i undervolted a slight bit (too much undervolt will HURT performance, not make it unstable) and cooled with a 360mm rad including 6 fans and a 100% working waterpump.

 

[vega22]

Hehe, with this launch it seems AMD decided that since they weren't going to fall as far from the hype as usually (apart from the frequencies, of course), to just leave beta testing to the general population and somehow they are mostly getting away with it. I can't imagine if Intel did something similar - Internet would probably explode

Intel have been doing it since Conroe dude, just the changes have been such small steps that the issues each gen have brought have been small scale.

 

[olymind1]

This is on a 1.0.0.2 Bios and Windows 10 1809, (1809 cause, lol, creative drivers, and dpc)

What kind of sound card do you have?

On my end my X-Fi Titanium HD (in MSI B450 Tomahawk) with the latest driver (FRL_PCDRV_L11_3_00_2022) is working great without a fault on Windows 10 1903.

 

[shmuck]

Not impressed with the update. Overall CPU usage does not go above 2% (which classifies as idling in my books, irrespective of any spikes that are so quick they aren't reflected in task manager), yet the voltage stays at 1.4V+. What's worse, when I chose the Energy saving plan before this update, core voltage maxed out at 0.9V if I am not mistaken, now it sits at 1.4V with very light usage

dont know about you, but with the new chipset driver my ryzen balanced defauled to 95% minimum processor state, causing me to be stuck on 1.45v on idle.
just setting it to 5% there lets me idle at 0.93v, dropping idle temps by 6c

In my case it was 99%. However, changing it to 5% does not seem to have any effect whatsoever.

 

[TheLostSwede]

Not impressed with the update. Overall CPU usage does not go above 2% (which classifies as idling in my books, irrespective of any spikes that are so quick they aren't reflected in task manager), yet the voltage stays at 1.4V+. What's worse, when I chose the Energy saving plan before this update, core voltage maxed out at 0.9V if I am not mistaken, now it sits at 1.4V with very light usage


In my case it was 99%. However, changing it to 5% does not seem to have any effect whatsoever.

It might help you let us know what hardware you have. It seems AGESA 1.0.0.3ABB helps.

 

[shmuck]

B450 Tomahawk with 3700X. 7C02v1A BIOS and 1.07.29.0115 chipset driver.

The slider made no difference either.