By using this site, you agree to our Privacy Policy and our Terms of Use. Close
Cyran on 08 December 2022
Edit Thread
haxxiy said:
Captain_Yuri said:

Well if you look at the relative performance when the 5600XT launched, you can see that during that time, both Vega 64 and 5600XT were actually ahead of the 2060. Vega in general had a lot of driver issues with it's driver which made it under perform not to mention AMD was pretty horrid with DX11 titles. As we moved onto more and more DX12 titles and AMD fixed their drivers, Vega and AMD's arch as a whole benefited hence why you see them increasing their position in relative performance during the 5600XT launch vs their own launches.

But now if you look at the Callisto Benchmark as an example, both the 5600XT and Vega 64 are now behind the 2060:

And also remember that 5700XT was ahead of the 2070 and now that's behind as well:

That's just for one game that favors Nvidia vs. the average. If this was MW2 we'd be arguing the opposite.

I was just pointing out that Pascal is more likely being kneecapped by drivers instead of arch since the same isn't happening for older AMD archs, adjusting for the fact that AMD is worse in TCP across the board. If anything, GCN looks slightly better than the average vs. RDNA in TCP.

OT -

https://videocardz.com/newz/intel-raptor-lake-s-refresh-confirmed-for-q3-2023-sapphire-rapids-hedt-specs-leaked

Just like 10 nm six years ago, it's looking like the new "Intel 4" CPUs are a distant dream at this point... maybe the US government should just nationalize Intel's foundries at this point, since it's clear that their R&D isn't cutting the biscuit.

My guess is we might see Intel 4 "Meteor lake" mobile processes next year and then maybe a very limited desktop release in early 2024 (think Broadwell level).  For what ever reason I don't think the Intel 4 process scales well on bigger dies which why we saw earlier of  rumors that Meteor Lake being limited to less cores then current raptor lake (which maybe why they decided to do a raptor lake refresh instead).

I don't think we know till late 2024 if this going to be a 10nm situation or not.  If we get 20A Arrow lake CPU by end of 2024 on desktop then Intel might be fine going forward if not then there reason to start worrying about a repeat of 14nm.