18 core Cascade Lake-X CPU scores 50% more than the Threadripper 2950X in Geekbench; almost 25% more than the Ryzen 9 3900X too

[Redaktion]

AMD may be storming away with multicore performance crowns, but Intel is not going down without a fight. Now, an 18 core Cascade Lake-X processor has appeared on Geekbench, and it has the AMD Threadripper series in its sights.

https://www.notebookcheck.net/18-core-Cascade-Lake-X-CPU-scores-50-more-than-the-Threadripper-2950X-in-Geekbench-almost-25-more-than-the-Ryzen-9-3900X-too.430886.0.html

[MrClip]

Just wanted to say that geek bench scores for 2950x in Linux are way higher than the scores notebookcheck has with Windows. In fact, they can be found to be mostly above 50,000 points on Linux easily. So either Linux is cheating the scores in AMD's favor or windows/geek bench in windows just doesn't like threadripper. The i9 9980xe also fares better on Linux or Mac OS though scores tend to vary more wildly from 40,000 to 70,000 possibly due to different cooling

[GoatGuy]

Ahem...

54500 ÷ 18 cores = 3027 per core  (Intel, this article)
44000 ÷ 12 cores = 3666 per core  (AMD, 3900)

Future AMD 3950 will have 16 cores.  Tho' straight scaling isn't terribly wise, it still offers a perhaps meaningful comparison.
 
44,000 × ¹⁶⁄₁₂ = 58,600 ...  estimate.

Also AMD's inside-track reporting has already cited that the 3950 is going to be one heck of a barn-burner ... GHz wise.  The actual on-the-streets results might actually exceed 64,000 ... apples-to-apples.

Just saying
GoatGuy  ✓

[Drax]

Strange, my stock Threadripper 1950X scored higher than the 2950X according to your numbers...

Check Geekbench 4 CPU result 14362501 (I can't post links due to low post count)

I also don't know why the "new" Intel benchmark results are using an old 4.3 version of Geekbench instead of the current 4.4 version.  Perhaps the old version shows AMD CPUs slower than they really are or Intel faster than they really are?

Just some oddities that make me question the results...all of the results...compared to my own benchmarks with a 1950X.