AMD Updates Ryzen Chipset Drivers with Greater 3D V-cache Optimization?

[btarunr]

AMD released updated Ryzen Chipset software, version 5.08.02.027. In addition to bug fixes for its various components, the drivers introduce application support for a "New program support and some new drivers added." The package now includes three new components—AMD SFH 1.1 driver, AMD PMF-7040Series driver, and a curiously named AMD Interface driver. SFH, or Sensor Fusion Hub, is a mobile-relevant component where various sensory devices, such as cameras, motion-sensors, etc., can be presented to the OS to drive applications that need them, such as Windows Hello. AMD PMF-7040 driver provides Platform Management Framework, and is specific to the Ryzen 7040-series "Phoenix" mobile processors. "AMD Interface driver" is a mystery.

AMD's first processor with 3D V-cache, the Ryzen 7 5800X3D, didn't face optimization issues, as it was a single-CCD processor. The company's latest 7900X3D and 7950X3D are dual-CCD processors, where one of the two CCDs has 64 MB 3D V-cache on top of the 32 MB on-die L3 cache, and the other is a regular "Zen 4" CCD with just the on-die 32 MB L3 cache. To ensure the right kind of workload is scheduled to the right kind of CCD, AMD has been using a driver that assists Windows scheduler with this exact task. "AMD Interface driver" could serve as a backend to an application that probably gives some kind of control for this 3D V-cache control driver; or perhaps it is just a backend for Ryzen Master. If AMD has to give end-users the ability to tinker with 3D V-cache application scheduling, the logical way it would go about doing it is add functionality to the Ryzen Master app. If there is a separate front-end application planned for the task, then it is probably non-public, and meant for AMD and its ISV partners.

DOWNLOAD: AMD Chipset Software 5.08.02.027



Release Highlights

• New program support and some new drivers added.
• Bug fixes.

Known Issues

• Some driver names may appear in English on a non-English OS.
• Uninstall summary log may incorrectly show the uninstall status as failed.

View at TechPowerUp Main Site

 

[Hyderz]

this is pretty nice for v-cache users

 

[Konyalee]

Maybe they will announce/launch something related next week at gamescom, I heard speculations about FSR 3.0 coming out so we might see more than hardware launches next week.

 

[Eskimonster]

Thanks AMD, looking forward to do various test on the driver updates

 

[Space Lynx]

this will be nice, i planned to go back to windows 10 soon, so this will be good to compare numbers vs win 11 on cinebench, shadow of tomb raider benchmark, etc.

i am curious if these latest updates will still give win 11 the winning edge. i hope not, cause I want to go back to windows 10.

 

[S73fan]

Does this mean that this update is not necessary for the 5800x3d?

 

[Space Lynx]

Does this mean that this update is not necessary for the 5800x3d?

not needed, but recommended, as you should see some minor increase in performance, we don't have benchmark numbers yet

 

[Kohl Baas]

Are these X3D specific optimizations only for Win11? I just bought such a CPU and am still on Win10. Should I go for the 11?

 

[Space Lynx]

Are these X3D specific optimizations only for Win11? I just bought such a CPU and am still on Win10. Should I go for the 11?

this is a good question, hopefully some website will probably post benchmarks and let us know soon enough

@Mussels get testing! *cracks whip*

 

[ixi]

this is pretty nice for v-cache users

It is unknown what the current "optimization" does. Maybe it is not "nice"

 

[Deleted member 185088]

Some driver names may appear in English on a non-English OS.

That's a huge problem in windows, it simply ignores any language/region settings.

 

[Kohl Baas]

That's a huge problem in windows, it simply ignores any language/region settings.

Well, now you'll know that in this specific case it won't be Windows ignoring your preferences but the driver itself.

 

[Bruno Vieira]

They finally added Chipset drivers for 7040 CPU users, so far you could only find in third party sites. Now all 7040 drivers (GPU+Chipset) are available on AMD's site

 

[R0H1T]

It is unknown what the current "optimization" does. Maybe it is not "nice"

 

[ncrs]

"AMD Interface driver" is a mystery.

There's a ReadMe.txt in the directory of that driver:

AMD Interface
===================================
Version 2.0.0.14 Release Notes

[...]

1 Overview
------------

Windows thinks PCI as a device and require a driver for it. Therefore, we provide a null driver package, a .inf file without .sys file, to satisfy OS requirements.

[...]

i. "AMD PCI".
ii. "Tiered Memory Page Migration"
iii. "Secondary Tiered Memory Page Migration"
iv. "AMD Root Complex Event Collector"
v. "AMD SMBUS"
[...]

It contains no active, executable code. Looks to be just the equivalent of most of the Intel Chipset driver package that just identifies hardware.

 

[persondb]

As someone with a Phoenix CPU, I wonder what the PMF-7040 does...

They finally added Chipset drivers for 7040 CPU users, so far you could only find in third party sites. Now all 7040 drivers (GPU+Chipset) are available on AMD's site

The GPU drivers are still a huge mess, they removed support of the 7040 series in the lastest drivers(the Ratchet and Clank one).

There's a ReadMe.txt in the directory of that driver:


It contains no active, executable code. Looks to be just the equivalent of most of the Intel Chipset driver package that just identifies hardware.

Could you take a look to see what the PMF-7040 one says?

 

[ncrs]

Could you take a look to see what the PMF-7040 one says?

Sure, but I won't fully reverse engineer it. I'm no driver expert, I only have a passing familiarity with Windows drivers.

It is a "proper" driver with both system and user-mode services. Looks to be responsible for adjusting power and performance ratios focusing on mobile devices:

SubGroup = {C763B4EC-0E50-4B6B-9BED-2B92A6EE884E}, "AMD Power Slider", "Adjust AMD Power Slider options"
Setting = {38cab4d5-db09-449f-9db5-1c91c909b6d4}, "PMF Controller", "Controls the thermal power solution"
Value = 1, "Better battery", "Enable some AMD power savings feature to increase battery life", %REG_DWORD%, 0x1
Value = 2, "Better performance", "Enable more performance gaining features", %REG_DWORD%, 0x2
Value = 3, "Best performance", "Enable all possible AMD performance gaining features", %REG_DWORD%, 0x3

There is another difference from the 6000-series PMF driver - a component called "System Idle State Detection Service" has been removed, but it's possible that it was integrated into the 2 services added for 7040.

The same subsystem is supported under Linux, so you can look into the code if you're interested in the implementation details.

 

[bug]

Here's hoping AMD didn't just say "screw it", exposed an interface and let applications deal with scheduling themselves.
Because clearly proper release notes isn't something of interest for them these days.

 

[Makaveli]

Installed these last night so far so good.

 

[Pooch]

I dont know much about hacking but is it possible that this interface driver if connected to optimization of the cpu v cache, could it be used as a backdoor/exploit area?

 

[bug]

I dont know much about hacking but is it possible that this interface driver if connected to optimization of the cpu v cache, could it be used as a backdoor/exploit area?

Of course it can, anything can be used to snoop around. But if you have an attacker talking to your driver's interface, you're already in trouble, regardless of whether AMD exposes this particular interface or not.

 

[persondb]

Sure, but I won't fully reverse engineer it. I'm no driver expert, I only have a passing familiarity with Windows drivers.

It is a "proper" driver with both system and user-mode services. Looks to be responsible for adjusting power and performance ratios focusing on mobile devices:

SubGroup = {C763B4EC-0E50-4B6B-9BED-2B92A6EE884E}, "AMD Power Slider", "Adjust AMD Power Slider options"
Setting = {38cab4d5-db09-449f-9db5-1c91c909b6d4}, "PMF Controller", "Controls the thermal power solution"
Value = 1, "Better battery", "Enable some AMD power savings feature to increase battery life", %REG_DWORD%, 0x1
Value = 2, "Better performance", "Enable more performance gaining features", %REG_DWORD%, 0x2
Value = 3, "Best performance", "Enable all possible AMD performance gaining features", %REG_DWORD%, 0x3

There is another difference from the 6000-series PMF driver - a component called "System Idle State Detection Service" has been removed, but it's possible that it was integrated into the 2 services added for 7040.

The same subsystem is supported under Linux, so you can look into the code if you're interested in the implementation details.

Thank you!

I hope that AMD releases a tool to easily adjust those things.

 

[Punkenjoy]

I installed it but I am not sure if there is any gain, the AMD 3D V-Cache optimizer version haven't changed.

 

[Double-Click]

Unless something is a game changer I'm usually of the "if it ain't broke" mindset.
So if you Zen 3 X3D owners out there could just post more about it on Saturday...that'd be great.

 

[Pooch]

Thanks Milt. Great... thanks.

Yea ..great.