Carzy Zarx’s PC Gaming Emporium - Catch Up on All the Latest PC Gaming Related News

[Bofferbrauer2]

Techpowerup checked how much only having 8 lanes slows you down, by comparing 8 lanes of PCIE 4.0 with the same amount of lanes on PCIE 3.0, 2.0 and even 1.1.: https://www.techpowerup.com/review/amd-radeon-rx-6600-xt-pci-express-scaling/

The result is... a whole lot of nothing, really. On average at 1080p, where the drop is the largest in most cases, having to resort to PCIE 3.0 instead of 4.0 costs you a whole 2% of performance - and that's both rounded up and with 2 outlier with Hitman 3 (13.5%) and Death Stranding (6.8%). In most cases, the difference was barely measurable, let alone felt. Going from DDR4-3200 to 3600 has probably a higher effect on the FPS than having just 8 lanes of PCIE, 4.0 or otherwise. 

In fact, even just using PCIE 2.0 would only result in a 7% performance drop (again mainly due to the outlier), and only PCIE 1.1 where the difference was really felt and the average performance drop was 17%. By extrapolating the curve, they come to the conclusion that PCI4.0 x16 would have added that whole lot of 1% to the performance average, so really nothing to lose sleep about.

That being said, there's no guarantee that it will stay that way. It could be that in the future, those speeds could in fact really bottleneck enough to be truly felt. But that probably won't be before a couple years down the road.

Captain_Yuri said:

Intel Core i7-12700 non-K Alder Lake CPU is almost as fast as Ryzen 7 5800X in leaked Geekbench score

https://videocardz.com/newz/intel-core-i7-12700-non-k-alder-lake-cpu-is-almost-as-fast-as-ryzen-7-5800x-in-leaked-geekbench-score

I'd ignore the rocket lake results because they are skewed by having AVX512 while AMD and ironically, Alder Lake S doesn't. Overall, it's good performance for a non-K sku. If the K skus are 15-20% ahead of Ryzen 5000 and not priced terribly, I think it will be a good CPU. But with Intel, pricing can always be funky.

How long is the Geekbench test?

If it's short enough, then it could have been performed completely or almost entirely under full boost, which is only marginally slower than on the K models. This is also one of the reasons why I take Geekbench results with a giant grain of salt, as they are pretty far removed from reality half the time.