DOOM: The Dark Ages Performance Benchmark

[Krit]

RTX 5070 is slower than RTX 3080 and performs basically the same as Vanilla RTX 4070 it's just embarrassing


Gen to Gen improvements.

RX 9070 XT is ~63% faster than RX 7800 XT
RTX 5070 is ~1.5% faster than RTX 4070

 

[rusTORK]

Review mentioned latency measures. How it was done? Game show it?

 

[Legacy-ZA]

Thank you for adding the latency with MFG, its a fantastic metric to add for those of us extremely sensitive to it.

 

[LastDudeALive]

I was a bit concerned after Indiana Jones, but it seems id really put work into making the engine and the game work with hardware ray tracing requirements. It can't run on a potato anymore like the first two games, but it runs better than pretty much any other RT implementation, so good for them. Everything except the 3050 and the RX 6600 and 6600 XT can get a playable 60fps with quality upscaling.

I am a bit disappointing by the performance scaling, only ~10-20% gain from ultra to low settings, but I guess there's only so much they can do with RT capability being the primary limit, it's not a feature that can be cranked up and down.

 

[Legacy-ZA]

It’s amazing how the 9070 XT performs better in Australia than in the US, no wonder it’s so expensive there.
In Australia, the 9070 XT performs at the level of a 5080 or faster,
while in the US, it barely reaches 5070 Ti levels.

The Russian and German 9070 XT's comes from the same lineage.

View attachment 398846

View attachment 398845

Note: The above may contain sarcasm and might not make complete sense.

nGreedia drivers are simply a mess, there are even instances where the 4070Ti Super is on par or slightly faster than the 5070Ti, dispite the latter having superior hardware.

They should be ashamed.

 

[LastDudeALive]

No FSR 4 support? AMD be sleepin'?

FSR 4 isn't supported on Vulkan yet, so there's no way to include it.

 

[W1zzard]

Review mentioned latency measures. How it was done? Game show it?

NV App Overlay

 

[TomasK]

There is no FSR 4 support for Vulkan games at the moment. This game and No Man's Sky would greatly benefit from it. I hope the new SDK gets released soon.

FSR 4 isn't supported on Vulkan yet, so there's no way to include it.

Then I hope AMD manages to implement FSR 4 for Vulkan by 2027, that's likely when I'll be buying the game - once it gets at least 50% discounted. I'm not crazy to pay 80€ for a video game.

 

[W1zzard]

Maybe some special super overclocked 5060ti with 8gb was used

Both are stock cards at reference clocks, but yes, when it doesn't run out of VRAM, my 8 GB 5060 Ti is slightly faster than my 16 GB card

 

[LastDudeALive]

Both are stock cards at reference clocks, but yes, when it doesn't run out of VRAM, my 8 GB card is slightly faster than my 16 GB card

Probably because fewer VRAM chips means slightly more power budget for the GPU itself.

 

[lightofhonor]

Probably because fewer VRAM chips means slightly more power budget for the GPU itself.

Correct. W1zzard mentioned the same in the 8GB review.

 

[RadiatorTwo]

When you do image quality comparisons, at least turn film grain off...

 

[NSR]

9070XT is doing great. if you were one of the lucky ones to get it at $600 you should be smiling from ear to ear.

9070 also maintaining its -10% against the XT, pushing it above the 5070 Ti.

This is a rough driver diff.

 

[Luminescent]

View attachment 398841
View attachment 398843

Well, this is getting old with the benchmark rankings being different from one publication to another, why not make everything transparent and release save file and location where the benchmark is made ? that way anyone can test and see for himself.
Also the guys at computerbase said this "Nvidia has made a second Game-Ready driver available in advance with the GeForce 576.40, which is supposed to show a better frame rate on RTX 5000 cards. All RTX 5000 benchmarks were then run again with the GeForce 576.40. And indeed, it went out a few percent, but RTX 5000 remains surprisingly weak. There were no changes to RTX 4000 due to the new driver."

 

[Talon]

There's some crazy differences in performance between TPU's testing and HUB's testing.
5000 series look very underwhelming in HUB's testing and looks fine in TPU's testing.
8GB cards look typically fine in TPU's testing until like 4K, but break much quicker in HUB's testing.
HUB's testing shows that the XTX and 9070XT are evenly matched, whereas TPU's testing shows the 9070XT clearly ahead.

How very odd.

They used the first driver provided, but it was bugged. Nvidia provided the updated driver which TPU and DSOGaming used. Both TPU and DSOGaming show 50 series doing much better than HWU, so who's the outlier here?

DSOGaming also used 576.40 driver.

 

[NSR]

They used the first driver provided, but it was bugged. Nvidia provided the updated driver which TPU and DSOGaming used. Both TPU and DSOGaming show 50 series doing much better than HWU, so who's the outlier here?

DSOGaming also used 576.40 driver.

I see, thanks for clarifying.

Though, this is still an oddly tight knit family of cards which are all supposed to be further apart (well, at the very least the 9070 XT should be slower than the 5070 Ti and the 9070 shouldn't be so close to the XTX, but neither should the 4070 Ti S). The 9000 cards and the 40 series appear to be outdoing themselves.

 

[Talon]

I see, thanks for clarifying.

Though, this is still an oddly tight knit family of cards which are all supposed to be further apart. The 9000 cards appear to be outdoing themselves.
View attachment 398866

Sure, we can look at any outlier game, but that doesn't change relative performance of a generation. Also a 9070 XT is priced exactly the same in the US. My local Microcenter has 1 model in stock at $940 and Newegg is $900. The 5070 Ti is $900 at Newegg. So they're not really far apart, not in terms of price anyways. The problem is that you're trying to price a card the same for less features.

 

[Nostras]

They used the first driver provided, but it was bugged. Nvidia provided the updated driver which TPU and DSOGaming used. Both TPU and DSOGaming show 50 series doing much better than HWU, so who's the outlier here?

DSOGaming also used 576.40 driver.

HUB states in the pinned comment that performance barely changed for the 5000 series cards with 576.40.

 

[Krit]

Sure, we can look at any outlier game, but that doesn't change relative performance of a generation. Also a 9070 XT is priced exactly the same in the US.

That's interesting because in EU RTX 5070 Ti is ~21% more expensive than AMD. RTX 5070 Ti is only ~6.4% (TPU) and ~5% (HWU) faster on average @1440p.

It depends on location but in EU AMD is better deal also AMD has better drivers than nvidia.

Massive 55 game test.

 

[NSR]

Sure, we can look at any outlier game, but that doesn't change relative performance of a generation. Also a 9070 XT is priced exactly the same in the US. My local Microcenter has 1 model in stock at $940 and Newegg is $900. The 5070 Ti is $900 at Newegg. So they're not really far apart, not in terms of price anyways. The problem is that you're trying to price a card the same for less features.

Oh yeah don't get me wrong, this is definitely an outlier. Still funny to see a group of cards so differently priced be one big happy family.

In the UK the 5070 Ti is at minimum £100 more expensive. At times maybe £75 if cheaper XT stock is sold. For some reason we're the only place who has these cards at their respective price gaps and close to MSRP (granted not an accurate MSRP since we're still charged extra above the US alternative).

 

[Talon]

Oh yeah don't get me wrong, this is definitely an outlier. Still funny to see a group of cards so differently priced be one big happy family.

In the UK the 5070 Ti is at minimum £100 more expensive. At times maybe £75 if cheaper XT stock is sold. For some reason we're the only place who has these cards at their respective price gaps and close to MSRP (granted not an accurate MSRP since we're still charged extra above the US alternative).

The price of these cards is disgusting. MSRP is such a joke now. Reviewers should stop doing the price per dollar graphs because they're almost borderline useless today in the US. The prices here are changing by the hour it seems lol. I saw B&H Video change the price of a 5090 Master 2x yesterday in the span of 1 hour.

 

[cpugpuram]

Looks like the game doesn't have 4k texture pack. Vram requirement should be higher.
I would love to see transformers vs cnn dlss tests for each generation of RTX in this game. I have 3080 ti and transformer dlss doesn't really work properly most of the time on my gpu. Enabling it can even reduce my native fps.

 

[Trunks0]

I was a bit concerned after Indiana Jones, but it seems id really put work into making the engine and the game work with hardware ray tracing requirements. It can't run on a potato anymore like the first two games, but it runs better than pretty much any other RT implementation, so good for them. Everything except the 3050 and the RX 6600 and 6600 XT can get a playable 60fps with quality upscaling.

I am a bit disappointing by the performance scaling, only ~10-20% gain from ultra to low settings, but I guess there's only so much they can do with RT capability being the primary limit, it's not a feature that can be cranked up and down.

Different teams. Indiana Jones was Machine Games, this is ID Software. ID of course can pour allot more into optimizing the engine itself, because it's in house.

That said... they are working with nVidia on bringing PathTracing to this game. Which honestly.... kinda disappointing. They are ID software, part of what they do is develop a game engine,... they shouldn't need nVidia to do it at all and they should be making there own custom solution.

 

[Luminescent]

Though, this is still an oddly tight knit family of cards which are all supposed to be further apart (well, at the very least the 9070 XT should be slower than the 5070 Ti and the 9070 shouldn't be so close to the XTX, but neither should the 4070 Ti S). The 9000 cards and the 40 series appear to be outdoing themselves.

In theory the 9070 xt should be at the level of RTX 5080 or faster, 9070 xt has 53,900 million transistors and RTX 5080 has 45,600 million transistors, 5070 ti has even less transistors being cut down/defective.
In practice, ray traced games don't use hardware ray tracing denoising on RDNA4, take a 10-20% performance hit there, then there is the texture compression Nvidia has and AMD doesn't, they have something now but no one uses it, make that another 5-10% disadvantage for AMD, add to that the usual FSR4 is not supported but we support even XESS and you get the picture.

No worries, i heard the game will get some path tracing/Nvidia money so AMD will run at 1 fps and every Nvidia GPU from rtx 3060 to RTX 5090 will beat anything AMD has from now till the end of times.

 

[Quicks]

"The driver situation has been quite fluid for this review. First, AMD and NVIDIA provided a set of drivers (25.10.09.01 and 576.31), but later updated them with new versions (25.5.1 WHQL and 576.40 Prerelease). When comparing performance data with other reviews, make sure to check whether old drivers are used. For the NVIDIA driver update we measured a performance increase of around 5%, which is pretty significant. NVIDIA confirmed that these performance improvements will be included in the public game-ready driver release on Monday, May 12th, 2025."