GhostofWar
Regular
It may have been prudent to have another fortnite with hardware lumen in the test just to round out testing everything multiple times, especially with the conclusion saying hardware RT is going to be put out to pasture by ue5.
Value of Hardware Unboxed benchmarking *spawn
- Thread starter Phantom88
- Start date
It is pretty bizarre that HUB is still choosing to disable RT when testing current generation $1000 GPUs. No excuse for it at this point especially in games like Control, Cyberpunk and DL2.
They do include quite a few RT titles but the methodology seems completely arbitrary which is the problem. Why enable RT in Witcher 3 but not Control or DL2? Seems quite dumb.
DegustatoR
Legend
Because if they would do this it would present results which would be the opposite of everything which Steve has been saying for the last several years.
Yes but that’s not really relevant. They chose to test both “basic” and “ultra” settings only for one game which is inconsistent with how they tested all other games. It seems arbitrary at first. Then that game just happens to be a major win for AMD so people will have questions about the intent of that decision. Did they explain it? I haven’t watched the video yet.
DavidGraham
Veteran
Anyways, here is the proper corrected graphs, removed duplicate results, and separated RT from Raster.
Raster 4K, 7900XTX vs 4080:
Modern Warfare 2 Ultra: +34%
Borderlands 3 DX12: +20%
Battlefield V DX11: +18%
Death Stranding: +17%
Far Cry 6: +17%
Assassin's Creed Valhalla: +16%
Wolfesnstein Youngblood: +15%
Hitman 3: +15%
The Division 2 DX12: +12%
A Plague Tale Requiem: +10%
Cyberpunk 2077: +9%
Watch Dogs Legion: +9%
Ghostwire Tokyo: +9%
Total War Warhammer 3: +8%
Gotham Knights: +7%
Doom Eternal: +6%
Warhammer Vermintinde DX12: 6%
F1 2022: +6%
For Honor: +5%
Dying Light 2: +5%
Counter Strike Go: +5%
Strange Brigade: +5%
Dirt 5: +4%
Serious Sam 4: +4%
Sniper Elite 5: +4%
Kingdome Come: +3%
Control: +3%
Total War Three Kingdoms: +2%
The Callisto Protocol: +2%
Warhammer 40K Darktide: +2%
Monster Hunter World: +1%
Hunt Showdown: +1%
Horizon Zero Dawn: 0
Fortnite DX12 SW Lumen/Nanite: 0
Battlefield 2142: -1%
The Riftbreaker DX12: -1%
Rainbow Six Siege: -1%
Metro Exodus: -1%
The Witcher 3 DX11: -2%
Apex Legends: -2%
Star Wars Squadron: -4%
Resident Evil 3: -4%
Forza Horizon 5: -5%
Halo Infinite: -6%
Shadow Of Tomb Raider DX12: -6%
War Thunder: -6%
Asseto Corsa: -7%
Resident Evil Village: -7%
Tiny Tina Wonderlands: -8%
Rainbow Six Extraction: -9%
The Outer Worlds: -9%
PUBG: -14%
Fortnite DX11: -10%
World War Z: -21%
Gears 5: -21%
RT 4K, 7900XTX vs 4080:
F1 2022: -13%
The Riftbreaker: -14%
Metro Exodus Enhanced: -23%
The Witcher 3: -34%
So contrary to his findings, the 7900XTX is 3% faster (not 1% faster) at 4K raster, while being 18% slower in 4K ray tracing.
However, out of the 54 games tested, they had 23 games with RT, but he only tested 4 of them, worse yet, included weak RT games like F1 2022 and The Riftbreaker while ignoring major ones like Dying Light 2, Cyberpunk .. etc. This sort of pick and choose is a major form of selective bias, purposefully skewing the results into a specific outcome that is already predetermined.
Also, It's generally accepted and well known that if a game has multiple APIs that are equal in image quality, then you test with the best API for each hardware, but of course Hardware Unboxed doesn't do that, so they test Borderlands 3 with DX12 punishing the 4080 for no reason, despite the DX11 path being available in the game and offering better performance for all hardware.
Proper testing is to test Borderlands 3 with DX11, test more major RT games, and remove duplicate dumb results like COD Low, Witcher 3 DX12 (with no RT), and Fortnite DX12 (with no Lumen and Nanite).
Raster 4K, 7900XTX vs 4080:
Modern Warfare 2 Ultra: +34%
Borderlands 3 DX12: +20%
Battlefield V DX11: +18%
Death Stranding: +17%
Far Cry 6: +17%
Assassin's Creed Valhalla: +16%
Wolfesnstein Youngblood: +15%
Hitman 3: +15%
The Division 2 DX12: +12%
A Plague Tale Requiem: +10%
Cyberpunk 2077: +9%
Watch Dogs Legion: +9%
Ghostwire Tokyo: +9%
Total War Warhammer 3: +8%
Gotham Knights: +7%
Doom Eternal: +6%
Warhammer Vermintinde DX12: 6%
F1 2022: +6%
For Honor: +5%
Dying Light 2: +5%
Counter Strike Go: +5%
Strange Brigade: +5%
Dirt 5: +4%
Serious Sam 4: +4%
Sniper Elite 5: +4%
Kingdome Come: +3%
Control: +3%
Total War Three Kingdoms: +2%
The Callisto Protocol: +2%
Warhammer 40K Darktide: +2%
Monster Hunter World: +1%
Hunt Showdown: +1%
Horizon Zero Dawn: 0
Fortnite DX12 SW Lumen/Nanite: 0
Battlefield 2142: -1%
The Riftbreaker DX12: -1%
Rainbow Six Siege: -1%
Metro Exodus: -1%
The Witcher 3 DX11: -2%
Apex Legends: -2%
Star Wars Squadron: -4%
Resident Evil 3: -4%
Forza Horizon 5: -5%
Halo Infinite: -6%
Shadow Of Tomb Raider DX12: -6%
War Thunder: -6%
Asseto Corsa: -7%
Resident Evil Village: -7%
Tiny Tina Wonderlands: -8%
Rainbow Six Extraction: -9%
The Outer Worlds: -9%
PUBG: -14%
Fortnite DX11: -10%
World War Z: -21%
Gears 5: -21%
RT 4K, 7900XTX vs 4080:
F1 2022: -13%
The Riftbreaker: -14%
Metro Exodus Enhanced: -23%
The Witcher 3: -34%
So contrary to his findings, the 7900XTX is 3% faster (not 1% faster) at 4K raster, while being 18% slower in 4K ray tracing.
However, out of the 54 games tested, they had 23 games with RT, but he only tested 4 of them, worse yet, included weak RT games like F1 2022 and The Riftbreaker while ignoring major ones like Dying Light 2, Cyberpunk .. etc. This sort of pick and choose is a major form of selective bias, purposefully skewing the results into a specific outcome that is already predetermined.
Also, It's generally accepted and well known that if a game has multiple APIs that are equal in image quality, then you test with the best API for each hardware, but of course Hardware Unboxed doesn't do that, so they test Borderlands 3 with DX12 punishing the 4080 for no reason, despite the DX11 path being available in the game and offering better performance for all hardware.
Proper testing is to test Borderlands 3 with DX11, test more major RT games, and remove duplicate dumb results like COD Low, Witcher 3 DX12 (with no RT), and Fortnite DX12 (with no Lumen and Nanite).
Henry swagger
Newcomer
Hub was sneaky with the testing.. why is cod so amd biased ?
That’s pretty funny.
DavidGraham
Veteran
Call of Duty games are based on the IW engine, in it's DX11 form, the engine was pretty balanced on all GPUs, it was lastly seen in Infinite Warfare 2016, Modern Warfare 1 Remastered 2016, and Modern Warfare 2 Remastered 2020.
Call of Duty Modern Warfare Remastered тест GPU | Action / FPS / TPS | Тест GPU
Игра Call of Duty: Modern Warfare, удостоенная множества хвалебных отзывов от критиков, возвращается с поддержкой высок
Call of Duty Modern Warfare 2 Remastered тест GPU/CPU | Action / FPS / TPS | Тест GPU
Действие кампании начинается сразу же после драматических событий революционной игры Call of Duty 4: Modern Warfare, зав
After that, the engine switched to DX12 and completely obliterated all NVIDIA GPUs, the DX12 implementation in the engine is borked, and is so CPU limited, that a 4090 can lose in 1080p to any AMD high end GPU.
Last edited:
That's a brilliant summary and shows how laughably skewed that review sample is.
I was already suspicious of bias and I'd barely scratched the surface of what was actually going on.
And did he really conclude that UE5 is going to put hardware RT out to pasture? Despite hardware RT being demonstrably visibly superior in both available examples of the engine for only a marginal performance impact?
Last edited:
Did Dying Light 2 and Cyberpunk become good games already?
Cynicalking
Regular
What does that have to do with anything?
Benchmarks could be more useful to consumers if they focus on games people like to play.
Cynicalking
Regular
So then they should focus on Fortnite and Call of duty at minimum settings since no games can compete in player base with them?
Dying Light 2 and Cyberpunk have excellent RT implementations, and what's being discussed are RT benches. That's why they where mentioned. If you think they are good or bad is irrelevant, still, they have solid sales numbers and reviews.
DegustatoR
Legend
Are we pretending that no one plays CP2077 now?
I for example couldn't care less about any Call of Duty title but that doesn't mean that no one is playing these games.
I for example couldn't care less about any Call of Duty title but that doesn't mean that no one is playing these games.
DavidGraham
Veteran
In that case a lot of the tested games in that review are practically useless, stuff like Strange Brigade (seiously WTF?), Riftbreaker, For Honor, Star Wars Squadron and others.
Cyberpunk is the most popular single player game on Steam in 2022 and by far, especially after the anime show and the constant updates that fixed almost everything and added lots of new features.
Dying Light as a series has a huge following, and the first game received dozens upon dozens of story and multiplayer add ons that went on for years, the game received multiple Definitive and Complete editions, and Dying Light 2 looks set to repeat the same story as Dying Light 1.
If you -as a tester- are going to test ray tracing in Riftbreaker, instead of testing it in Cyberpunk, or Dying Lights 2, or Control, or Doom Eternal, or Watch Dogs Legion or Hitman 3 or Warhammer Darktide, then you either have a malintent or something is wrong in your head.
That's a really strange take. Cyberpunk is widely considered to be a very good game outside of its early technical issues.
I don't know much about Dying Light 2 but I've not encountered any negative reactions to its gameplay.
Theoretically yes but only if applied consistently for all games. Either way it doesn’t have anything to do with disabling RT in games that support it.
Svensk Viking
Regular
Here's what got me pondering regarding you folks who snear at the comments of reviewers being excused for focusing on popular games...This thread is obviously focusing on high end raytracing performance, and Hardwareunboxed's reviews typically are not. That's well known since their RX 5700 review and there has been multiple backs and forths about it, for instance in this thread's predecessor.
Still, they get linked to in this thread for whatever reasons, and criticized according to the criteria set up by users in this thread.
Then one user comes along and suggests there might be valid reasons for their test choices, and immediately there's a response suggesting that line of thought is fully wrong.
Why not just keep it simple and clean and just pick out the raytracing tests they did run?
Why start hacking away on a review piece by piece for not meeting criteria which it never did set out to meet? All it amounts to is turning the thread into a mess.
Always this attitude from you, only capable of seeing things in black and white. No, a tester does not need to have malintent or something wrong in the head just because they test games not from your list. A forum is a place of discussion of varying opinions, you should have learned this by now.
A tester is free to test raytracing in whatever games he likes for whatever reason he has. And yes, one sensible reason for doing so could be because he wants to draw readers/viewers by featuring a game he knows is popular.
The only case where he is not free to do his tests freely? If he has signed a deal dictating the testing procedure to get the review hardware.
D
Deleted member 2197
Guest
So are you suggesting HUB is perhaps receiving an incentive for testing the way they do?