After seeing all the mixed reviews from current AM5 9000 series launch from tech t00bers and having played with the platform for a few weeks I wanted to see how Microsoft's current offering compares to Linux especially after seeing the Phoronix article and remembering playing with this during 3990X days and how big the difference was even back then.

I used a Windows 11 24H2 base that was tweaked vs a ootb Linux (Arch+GNOME) distro since I have no clue how to tweak things in that realm. I used Geekbench 3, 5, 6 (x86_64 & AVX2) with the exact same bios profile. The hardware consists of:

• AMD Ryzen 7 9700X
• ASRock B50M HDV/M.2
• G.Skill Trident Z5 RGB OCWC2024 (2x16GB) @ 8000c34
• EVGA 1600 T2 PSU
• Thermalright Phantom Spirit 120
• SK Hynix P31 / Samsung 870 EVO SSD's
• Galax GT630 Fishdick VGA

The only change in profiles was No Execute had to be enabled to boot and install the Linux operating system, so this was enabled for all tests not in Windows.

The overclock is a static 5.5Ghz all core with 2200 infinity fabric and 8000c34 memory for all screenshots running in 1:2 with sane voltages that can be run daily. Results as follows. . .

Geekbench 3 Windows 11

https://browser.geekbench.com/v3/cpu/9087680

 

Geekbench 3 Linux

https://browser.geekbench.com/v3/cpu/9087676

 

Geekbench 5 W11

https://browser.geekbench.com/v5/cpu/22787660

 

Geekbench 5 Linux

https://browser.geekbench.com/v5/cpu/22787531

 

 

Geekbench 6 W11 (x86_64)

https://browser.geekbench.com/v6/cpu/7379233

Geekbench 6 W11 (AVX2)

https://browser.geekbench.com/v6/cpu/7379332

 

Geekbench 6 Linux (x86_64)

https://browser.geekbench.com/v6/cpu/7378520

 

Geekbench 6 Linux (AVX2)

https://browser.geekbench.com/v6/cpu/7379381

 

TLDR: Gainz everywhere in every scenario

 

If you have any 7000 series chips lying around it might be interesting to see if there are similar gains to be had with that platform too, i.e. to rule out whether this is just a software efficiency thing or if it really is a platform issue on windows.

No matter how great some benchmarks are on linux it's a monstrocity to keep track of all the Kernel versions etc...

On 8/19/2024 at 5:09 AM, Leeghoofd said:

No matter how great some benchmarks are on linux it's a monstrocity to keep track of all the Kernel versions etc...

That's too bad because my main is CachyOS. I might have to buy another SSD for Windows. lol

For now I might share it here or start Discord or another forum for record keeping.

I had a lot of screenshots for i7 2600k, i5 4690k, Ryzen 7 1700, Ryzen 7 3800x and Ryzen 5900x all under CachyOS or Linux Mint. I bought couple of used stuff like i7 4790k and later next month I will be getting 7700k.

On 8/19/2024 at 5:05 AM, ObscureParadox said:

If you have any 7000 series chips lying around it might be interesting to see if there are similar gains to be had with that platform too, i.e. to rule out whether this is just a software efficiency thing or if it really is a platform issue on windows.

Its really Windows issue because they keep adding bunch of telemtry for data collection. You have to get strip down version and you can see the huge difference.

Even though strip down its not perfect there is missing stuff that cause unstability for specific software and there is still issue with RAM being full when dealing with 4GBs of ram or even 16GBs of ram.