A Look At The Windows 10 vs. Linux Performance On AMD Threadripper 2990WX

[Grinch]

Another very interesting test, thumbs up! I was disappointed that there were no Go benchmarks this time around though, since I find them the most telling due to not being affected by any distro specific C/CXX flags.

[Michael]

Another very interesting test, thumbs up! I was disappointed that there were no Go benchmarks this time around though, since I find them the most telling due to not being affected by any distro specific C/CXX flags.

If you click through to the OpenBenchmarking.org page, there are some. I only included the results in the article that were heavy CPU tests.

[angrypie]

And they're extremely correlated, but not like someone like you would know that.

All you know about me is my username, so it's obvious you're the clueless mongrel here. I suggest you join Reddit and gather some sweet internet points.

[haasn]

What cooling system are you using?

Block: Phanteks Glacier C399a
Pump: EK-XTOP Revo D5 PWM
Radiator: Hardware Labs Black Ice Nemesis GTS 360 X-flow
Fans: 3x 120mm NB-eLoop B12-P Bionic
Tubing: Tygon E3603 (15.9/11.1mm), with 16/11mm compression fittings from Bitspower

I run the pump at around 50% PWM constantly, and the fans dynamically at low speeds (which works fine, the radiator is designed for low-speed operation).

[Flodul]

I think Windows is unable to recognize the strange NUMA-layout (2 nodes with memory and 2 nodes without) of the threadripper.
It would be interesting how the results would look when SRAT is disabled, page-interleaving is on and the two compute-dies are disabled.

[edwaleni]

In my research across the various sites that benchmark products, I am seeing some wild variations in test results with little or no explanation on some key configuration details.

One site can have Windows rock the world and another site say Linux rocks the world on TR.

I have been trying to piece together the commonalities with all of these benchmark results and while they all attempt to provide results in a consistent fashion, they seem to be missing some key details.

PTS is nice as Michael has gone all out to make it cross platform. Many sites stick to the stock Windows apps as 98% of their readership expects that.

PCWorld has run into several testing anomolies that they don't try to explain, just a "something aint right" response to the results. As Michael has found on certain applications, they aren't designed with 32, let alone 64 threads in mind and their behavior is inconsistent. Also PCWorld has found that TR runs "better" when it has more load (more than 2 apps threading).

This has been seen in Xeon's where the thermal management and speed steps are too aggressive and it won't activate the additional cores and allow the threads to run. This might explain the wild variations in bench results in different places.

[switchperson]

In my research across the various sites that benchmark products, I am seeing some wild variations in test results with little or no explanation on some key configuration details.

One site can have Windows rock the world and another site say Linux rocks the world on TR.

I have been trying to piece together the commonalities with all of these benchmark results and while they all attempt to provide results in a consistent fashion, they seem to be missing some key details.

PTS is nice as Michael has gone all out to make it cross platform. Many sites stick to the stock Windows apps as 98% of their readership expects that.

PCWorld has run into several testing anomolies that they don't try to explain, just a "something aint right" response to the results. As Michael has found on certain applications, they aren't designed with 32, let alone 64 threads in mind and their behavior is inconsistent. Also PCWorld has found that TR runs "better" when it has more load (more than 2 apps threading).

This has been seen in Xeon's where the thermal management and speed steps are too aggressive and it won't activate the additional cores and allow the threads to run. This might explain the wild variations in bench results in different places.

I also wonder what are the differences. That's quite strange the Linux is THAT much faster all across the board (not saying it ain't true). It seems that a thorough investigation is required

[varikonniemi]

How long until microsoft throws in the towel and submits their inferior kernel? Use Linux and make windows the premier DE. Everyone would win.

[Weasel]

All you know about me is my username, so it's obvious you're the clueless mongrel here. I suggest you join Reddit and gather some sweet internet points.

Nope. Based on your (previous) post in this thread, I can say I know enough to make that statement.

[angrypie]

Nope. Based on your (previous) post in this thread, I can say I know enough to make that statement.

If you're happy with a pretense of knowledge, then more power to you.