It really annoys me when media outlets, spread misinformation. I expected better from Tom’s hardware than just parroting such "news" without bothering to do some critical thinking and cross verification. So, the 3950X supposedly scores 32082 on Firestrike physics and that is supposedly 24.2% higher than the supposed score of 25838 of the 10980XE huh?
As it has already been pointed out if you go and check actual benchmarks from reviews of the 9980XE, the processor that the 10980XE replaces, it achieves a physics score of around 28200-29000 (depending on RAM, etc) at stock and around 32300-32500 when moderately OCed. This puts it above the supposed leaked score of 3950X. So much about the 3950X outperforming the 10980XE by 24.2% then…
Also Firestrike’s physics score scales with boost clocks. On Tom’s hardware own testing, going from stock 7980XE to stock 9980XE, Firestrike's physics score went from 25477 to 28214, a 10.7% increase. We should expect a similar increase with the 10980XE versus the 9980XE meaning we should be expecting stock Firestrike physics scores of 31200-32100 and of course even higher when moderately OCed. And, for the record, when overclocked to its absolute max the 9980XE scores a physics score of over 46000. So expect the 10980XE to breach the 50000 milestone this time around. We will see…
Last but not least, although media are always quick to point out that the 3950X achieves what it achieves while having 2 fewer cores, they never mention that it does have almost 70% more L2+L3 cache. So it is by no means with fewer overall resources.