Were we wrong about Zen 5?

[BigEZE]

This is the reason I built a new build using the AMD 5800X3D chip and the Gigabyte Aorus Master X570S motherboard as well as tried and true DD4 with Samsung 980 Pro SSD's. The new tech showed extremely little gains in FPS for a very large hardware cost up charge just to switch platforms and to go to DDR5 memory and to the AMD 7800X3D, which of course did NOT come down in price until much later. Glad I did. So far the 9000 series is starting off the same way.

 

[mistyrain]

Hmm, what about using the hidden admin account? Can Win be the problem?

 

[Cabuzzi]

Or perhaps forum writers actually know more about this CPU than Techspot staff? AMD said long time ago Zen5 should have much better AVX512 performance than Zen4. Now it's confirmed that is true. Techspot has written two articles where it failed to even mention AVX512! AMD "fanboys" keep facts together, like it or not.

Who the heck cares about AVX512? Literally no one here is talking about it, yet you keep bringing it up like it's relevant. The entire article is about gaming performance of the 9700X. All the comments are about gaming performance of the 9700X. Stay on target.

 

[HardReset]

Who the heck cares about AVX512? Literally no one here is talking about it, yet you keep bringing it up like it's relevant. The entire article is about gaming performance of the 9700X. All the comments are about gaming performance of the 9700X. Stay on target.

Entire article about gaming performance? Perhaps you should read article before writing such BS. If it's all about gaming performance, why they write about server workloads, efficiency, PBO, power consumption, TDP, AMD benchmark results, other reviews, memory speed, memory overclocking...

Basically, 90% of article is about something else than gaming performance.

 

[Endymio]

https://www.phoronix.com/review/ryzen-9950x-windows11-ubuntu/7

Windows or perhaps the software version is underutilizing both zen4 and zen4, 20% extra performance is not something I would want to leave on the table.

Not quite. They didn't test "Zen 4 and Zen 5". They tested 2 high-core count processors (16/32) against a suite of apps that scale perfectly linearly with core count -- a scenario that Linux usually beats Windows in. Nor was the performance differential 20%. A geometric mean showed 7% for the 7950x, and 14% for the 9950x. And a geometric mean isn't the appropriate metric in any case -- an arithmetic mean would have shown about half that.

 

[WhiteLeaff]

Not quite. They didn't test "Zen 4 and Zen 5". They tested 2 high-core count processors (16/32) against a suite of apps that scale perfectly linearly with core count -- a scenario that Linux usually beats Windows in. Nor was the performance differential 20%. A geometric mean showed 7% for the 7950x, and 14% for the 9950x. And a geometric mean isn't the appropriate metric in any case -- an arithmetic mean would have shown about half that.

The difference I mentioned is in Blender, I don't think such a large deficit is normal.
What you're right about is that the average calculated over a wide range of software doesn't work well for most users. Each person has different needs, if the person works with 3D editing or development the only software that matters is modeling/Engines/Compilation etc... this average becomes a useless distortion.