Digital Foundry Article Technical Discussion [2021]

[BRiT]

I wouldn't say it's invalid, either. I'd say it's possible the results will change with driver updates.

 

[Deleted member 2197]

That's probably why comparisons involving WHQL drivers would be more meaningful, or in this case comparing just the AMD lineup. Just looking at some of GameGPU past reviews clearly indicates wide performance swings for both IHV's when no optimized drivers were available.

 

[techuse]

NVidia drivers rarely change performance. Especially in DX11 games.

 

[Deleted member 2197]

NVidia drivers rarely change performance. Especially in DX11 games.

Yeah, Nvidia's drivers are game optimized once the supporting WHQL drivers are available.

 

[dobwal]

Me neither. I hardly play games at all anymore, so when I do I always set them to "easy". I only ever get 30-60mins once a week to actually play. I used to be a hardcore gamer, but I've been replaced by two hardcore gamers.

...so yeah, any of the games deemed "hard" I avoid (though I enjoyed Bloodborne a few years back). Best games for me are short or multiplayer.

Yeah, I need games to be just hard enough for me to feel good about myself. Nothing more and nothing less. And game programmers are getting good at it. If you play a game and constantly find yourself making it through encounters with a sliver of health, its not you being lucky or skilled enough for those critical moments. Its programmers using psychology tricks to keep you engaged which is all I need. LOL.

 

[techuse]

Yeah, Nvidia's drivers are game optimized once the supporting WHQL drivers are available.

Do you think it’s likely a future driver changes anything WRT performance in Nioh 2?

 

[PSman1700]

Nice another DF video. Seems il have to settle for the PC version if i want the best experience.

ooo that's actually fairly long

Its alot of non-gameplay elements in those hours.

 

[Tkumpathenurple]

ooo that's actually fairly long, though shorter than Demon Souls likely.

There's a fair bit of optional content. So you may be able to blast through it somewhat swifter than 30 hours.

 

[Globalisateur]

NIOH 2 is NOT optimized for NVIDIA GPUs, at 4K the 6900XT is 20% faster than 3090, Vega 64 is as fast as 1080Ti, the 5700XT is as fast as the 3070. Comparing PS5 to NVIDIA GPUs is not really valid, a better point would be comparing against AMD lineup.

https://gamegpu.com/rpg/ролевые/nioh-2-the-complete-edition-test-gpu-cpu

When stacked against the PS5 the "tools are behind" narrative is also explaining the poor performance of Nvidia GPUs?

 

[Deleted member 2197]

Do you think it’s likely a future driver changes anything WRT performance in Nioh 2?

That's a hypothetical. Why not wait for a whql driver supporting Nioh 2 is available?

 

[goonergaz]

ooo that's actually fairly long, though shorter than Demon Souls likely.
Right now I get to choose between PoE 2: DeadFire, Disco Elysium, or GoW. The first 2 are pretty long lol. 60+ hr investments. I want to get through some of the Sony classics before I take on the juggernauts.


Thanks! I will enjoy it I'm sure!

www.howlongtobeat.com is your friend.

I have started using this to decide the order of my backlog - GoW main story is ~20hours

 

[dskneo]

When a game that runs well on AMD appears, its always "but its not optimized for Nvidia! not representative!!". I say its damn well representative of that game. There isn't a rule that Nvidia must always lead.

 

[techuse]

When a game that runs well on AMD appears, its always "but its not optimized for Nvidia! not representative!!". I say its damn well representative of that game. There isn't a rule that Nvidia must always lead.

I fail to see why people think Watch Dogs is a more valid comparison. As if Nvidia co-developed games have ever been the norm for multiplatform performance representation for the last 8 years rather than a very small fraction of the AAA titles.

Black Ops, Valhalla, Godfall, Nioh 2, Dirt 5 etc are all instantly dismissed while Watch Dogs is the current be all end all.

 

[Deleted member 86764]

The unidirectional argument; the idea that you only find data that supports your bias or hypothesis.

 

[PSman1700]

Its GPU is somewhat faster then a RX5700XT (in normal rendering). Thats the easiest and most fair way of 'comparing' the GPU if thats so important.
Where a 5700XT performs relative to a 2070/2070S will vary per title. Anyway have fun comparing to mid range 2018 products.

 

[snc]

Its GPU is somewhat faster then a RX5700XT (in normal rendering). Thats the easiest and most fair way of 'comparing' the GPU if thats so important.
Where a 5700XT performs relative to a 2070/2070S will vary per title. Anyway have fun comparing to mid range 2018 products.

5700xt is 2019 amd card from performance segment

 

[PSman1700]

5700xt is 2019 amd card from performance segment

While true, one could argue AMD just did not really (try to) compete in the high end segment in 2019. Fast forward 2020/2021 and they obviously compete with NV in the highest end range, atleast in normal rendering. Since Turing products are very popular in comparison benchmarks, 5700XT is around 2070/2070S across the board.
It doesnt matter but ok. 5700XT doesnt have ray tracing at all.

Going forward, a 6700 (10.37TF/2.2ghz) would be intresting as a PS5 matching setup for DF i think. Actually, from data (if correct), its rather similar to the PS5's GPU in raw specs. 36CU, 2250mhz, 384gb/s bw, rdna2.0 and die size for the gpu.
A 6700XT sits closer to the XSX at almost 12TF.

 

[Jay]

XS BC double fps tested on 5 games

 

[ChuckeRearmed]

Didn't get how they achieved that but insane!

 

[snc]

Didn't get how they achieved that but insane!

working with higher level directx gives more flexibility probably