Digital Foundry Article Technical Discussion [2025]

[DavidGraham]

There is lots and lots of stuff messing from current GPU reviews

Where is the Reflex tests? Reflex is found in over 150 games right now? Where is the latency measurment in these games?

Where is the IQ normalized testing? Reviewers test with DLSS and FSR as if they provide the same image quality. They test path tracing as it if it's the same quality despite other vendors lacking ray reconstruction .. etc.

In most UE5 games, users with one upscaler are playing at significantly higher quality than users with another upscaler, or at significantly higher fps if the IQ is normalized. That difference should be accounted for in the review, instead of being glossed over.

 

[Scott_Arm]

There is lots and lots of stuff messing from current GPU reviews

Where is the Reflex tests? Reflex is found in over 150 games right now? Where is the latency measurment in these games?

Where is the IQ normalized testing? Reviewers test with DLSS and FSR as if they provide the same image quality. They test path tracing as it if it's the same quality despite other vendors lacking ray reconstruction .. etc.

In most UE5 games, users with one upscaler are playing at significantly higher quality than users with another upscaler, or at significantly higher fps if the IQ is normalized. That difference should be accounted for in the review, instead of being glossed over.

I think dedicated/companion pieces are fine for this stuff. Repeating them in every review is kind of pointless unless something significant has changed.

 

[Sega_Model_4]

Spoiler: Summary

• (00:03) AMD had high expectations for CES, with rumors of new RDNA 4 GPUs and FSR 4, but these were not officially showcased due to "timing issues."
• (00:40) At the AMD booth in Las Vegas, there was no clear announcement or representation of FSR 4, but a surprise "research project" was discovered featuring machine learning-based upscaling.
• (01:39) The demo compared FSR 3.1 with a new machine learning upscaler, presumably FSR 4, showcased on two PCs, one with an RDNA 4 engineering sample.
• (02:55) The comparison was made using Ratchet and Clank, where FSR 3.1 exhibited visual artifacts (like Moiré patterns), while the new research project (FSR 4) appeared smoother with fewer artifacts.
• (03:29) The demo lacked official context, with no AMD representatives present to explain the technology in detail, leading to a somewhat "sneaky" presentation.
• (04:53) The new upscaling technology seemed to outperform FSR 3.1 in terms of image quality, with better handling of textures and reduced flickering or artifacting.
• (06:30) FSR 4 showed improved handling of background effects like ambient occlusion and textures, with fewer issues like the "Moiré pattern" that was noticeable in FSR 3.1.
• (08:27) The FSR 4 tech was sharper than previous iterations and handled particle effects (like confetti) better, displaying clear, non-trailing particles.
• (10:04) AMD's new upscaling solution, tested in performance mode, showed great quality improvements, even when scaling down from 4K, which FSR 3 struggled with.
• (11:43) The research project did not show characteristics similar to Sony's PSSR (PlayStation upscaling), meaning it was likely AMD’s proprietary technology. Further testing was needed to understand performance and final implementation.

 

[DavidGraham]

I think dedicated/companion pieces are fine for this stuff. Repeating them in every review is kind of pointless unless something significant has changed.

But repeating them in every review is crucial, on, for example on a 4060 you can do a 60fps with DLSS Performance with the same image quality as a 7700XT @FSR Quality running at 60fps as well, or vs a 7600XT doing 40fps at FSR Quality.

Normalized IQ per game may make or break certain GPU comparisons.

 

[Scott_Arm]

But repeating them in every review is crucial, on, for example on a 4060 you can do a 60fps with DLSS Performance with the same image quality as a 7700XT @FSR Quality running at 60fps as well, or vs a 7600XT doing 40fps at FSR Quality.

Normalized IQ per game may make or break certain GPU comparisons.

It's not possible to normalize quality that way. What's considered equal quality is going to be totally subjective. I think mentioning qualitative differences is important. Some discussion of upscaling methods, latency is important, but in depth is probably better handled in a side piece. I think in general reviewers kind of hand-wave these things away and just look at average fps and 1% lows and call it a day, and that undervalues those features, but really extensive coverage would be a lot for every single review.

I take more issue with reviewers reducing things to value in $/frame, and things like that, which totally neglect the other feature sets of the cards. People can decide their own value, based on their own preferences.

 

[Scott_Arm]

Spoiler: Summary

• (00:03) AMD had high expectations for CES, with rumors of new RDNA 4 GPUs and FSR 4, but these were not officially showcased due to "timing issues."
• (00:40) At the AMD booth in Las Vegas, there was no clear announcement or representation of FSR 4, but a surprise "research project" was discovered featuring machine learning-based upscaling.
• (01:39) The demo compared FSR 3.1 with a new machine learning upscaler, presumably FSR 4, showcased on two PCs, one with an RDNA 4 engineering sample.
• (02:55) The comparison was made using Ratchet and Clank, where FSR 3.1 exhibited visual artifacts (like Moiré patterns), while the new research project (FSR 4) appeared smoother with fewer artifacts.
• (03:29) The demo lacked official context, with no AMD representatives present to explain the technology in detail, leading to a somewhat "sneaky" presentation.
• (04:53) The new upscaling technology seemed to outperform FSR 3.1 in terms of image quality, with better handling of textures and reduced flickering or artifacting.
• (06:30) FSR 4 showed improved handling of background effects like ambient occlusion and textures, with fewer issues like the "Moiré pattern" that was noticeable in FSR 3.1.
• (08:27) The FSR 4 tech was sharper than previous iterations and handled particle effects (like confetti) better, displaying clear, non-trailing particles.
• (10:04) AMD's new upscaling solution, tested in performance mode, showed great quality improvements, even when scaling down from 4K, which FSR 3 struggled with.
• (11:43) The research project did not show characteristics similar to Sony's PSSR (PlayStation upscaling), meaning it was likely AMD’s proprietary technology. Further testing was needed to understand performance and final implementation.

Oliver did a really nice job filming with his phone. Looks like a great upgrade.

 

[Seanspeed]

That’s certainly not true when it comes to product reviews. In many cases the info shared is a subjective evaluation.



I disagree. DF is a standout in terms of evaluating frame consistency, stuttering, image quality etc. Most reviewers don’t go that far. You get avg and maybe 1% fps that’s it. They’re currently doing the least amount of work that would qualify as a review.

To bring this back to the original point. If reviews report only “raw” fps (which I agree they should) then they must either find some way to communicate the other capabilities of these cards or choose to ignore them. Those are the options. They have a very poor record of doing the former.

Digital Foundry's hardware reviews actually tend to be WAY lighter in terms of comprehensiveness and quantity of data than the other big outlets out there. It's such a bizarre example, I dont know what to say, because DF should be like target #1 if your issue is that reviewers aren't doing enough and testing every single kind of situation possible.

I'm getting the feeling what you're really asking for is reviewers to more heavily glorify Nvidia's features and advantages.

 

[Cyan]

Spoiler: Summary

• (00:03) AMD had high expectations for CES, with rumors of new RDNA 4 GPUs and FSR 4, but these were not officially showcased due to "timing issues."
• (00:40) At the AMD booth in Las Vegas, there was no clear announcement or representation of FSR 4, but a surprise "research project" was discovered featuring machine learning-based upscaling.
• (01:39) The demo compared FSR 3.1 with a new machine learning upscaler, presumably FSR 4, showcased on two PCs, one with an RDNA 4 engineering sample.
• (02:55) The comparison was made using Ratchet and Clank, where FSR 3.1 exhibited visual artifacts (like Moiré patterns), while the new research project (FSR 4) appeared smoother with fewer artifacts.
• (03:29) The demo lacked official context, with no AMD representatives present to explain the technology in detail, leading to a somewhat "sneaky" presentation.
• (04:53) The new upscaling technology seemed to outperform FSR 3.1 in terms of image quality, with better handling of textures and reduced flickering or artifacting.
• (06:30) FSR 4 showed improved handling of background effects like ambient occlusion and textures, with fewer issues like the "Moiré pattern" that was noticeable in FSR 3.1.
• (08:27) The FSR 4 tech was sharper than previous iterations and handled particle effects (like confetti) better, displaying clear, non-trailing particles.
• (10:04) AMD's new upscaling solution, tested in performance mode, showed great quality improvements, even when scaling down from 4K, which FSR 3 struggled with.
• (11:43) The research project did not show characteristics similar to Sony's PSSR (PlayStation upscaling), meaning it was likely AMD’s proprietary technology. Further testing was needed to understand performance and final implementation.

great tp know that AMD are improving a lot. Will watch the video later. That said, I can imagine how AMD fans are reacting, now what they call fake frames is going to be really cool, and upscaling technology will be he ideal technology to have.

 

[iroboto]

Spoiler: Summary

• (00:03) AMD had high expectations for CES, with rumors of new RDNA 4 GPUs and FSR 4, but these were not officially showcased due to "timing issues."
• (00:40) At the AMD booth in Las Vegas, there was no clear announcement or representation of FSR 4, but a surprise "research project" was discovered featuring machine learning-based upscaling.
• (01:39) The demo compared FSR 3.1 with a new machine learning upscaler, presumably FSR 4, showcased on two PCs, one with an RDNA 4 engineering sample.
• (02:55) The comparison was made using Ratchet and Clank, where FSR 3.1 exhibited visual artifacts (like Moiré patterns), while the new research project (FSR 4) appeared smoother with fewer artifacts.
• (03:29) The demo lacked official context, with no AMD representatives present to explain the technology in detail, leading to a somewhat "sneaky" presentation.
• (04:53) The new upscaling technology seemed to outperform FSR 3.1 in terms of image quality, with better handling of textures and reduced flickering or artifacting.
• (06:30) FSR 4 showed improved handling of background effects like ambient occlusion and textures, with fewer issues like the "Moiré pattern" that was noticeable in FSR 3.1.
• (08:27) The FSR 4 tech was sharper than previous iterations and handled particle effects (like confetti) better, displaying clear, non-trailing particles.
• (10:04) AMD's new upscaling solution, tested in performance mode, showed great quality improvements, even when scaling down from 4K, which FSR 3 struggled with.
• (11:43) The research project did not show characteristics similar to Sony's PSSR (PlayStation upscaling), meaning it was likely AMD’s proprietary technology. Further testing was needed to understand performance and final implementation.

Alex looks like he's 6'+ tall. I was always under this impression Oliver was tall, but Alex is taller!

interesting results. I was heavily musing that PSSR would have been an earlier derivative of PSSR, but I suppose not. Looking forward to the release version.

 

[trinibwoy]

Digital Foundry's hardware reviews actually tend to be WAY lighter in terms of comprehensiveness and quantity of data than the other big outlets out there. It's such a bizarre example, I dont know what to say, because DF should be like target #1 if your issue is that reviewers aren't doing enough and testing every single kind of situation possible.

DF provides quality over quantity. I never said anything about testing every situation possible. Please reread my posts.

I'm getting the feeling what you're really asking for is reviewers to more heavily glorify Nvidia's features and advantages.

Ha so that’s what you’re actually on about.

 

[Seanspeed]

DF provides quality over quantity. I never said anything about testing every situation possible. Please reread my posts.



Ha so that’s what you’re actually on about.

No, that's not what I'm 'on about', that's what I'm suspecting you're on about, because nothing you're saying is remotely consistent here. DF is not about 'quality over quantity' necessarily, even if they'll occasionally talk about something others aren't. You're talking about reviewers doing 'bare minimum' and ignoring all manner of things that are supposedly important - well DF haven't covered the cost of filters or overlays ever, so they're every bit as guilty of what you're accusing others of. DF may fill in some occasional gaps, but they are not themselves more comprehensively covering the things others do, either. Almost like no one outlet can do it all, eh? DF aren't really major hardware reviewers anyways, they're typically a lot more focused on software/game analysis.

Also, quantity is very important in this realm. You cant just dismiss that. Judging performance based off a few results can be incredibly misleading, and it's why people should be checking outlets who do large sample testing, or at the very least checking out multiple different outlets to get a better overall picture.

 

[RobertR1]

Question now becomes, what the hell happened with pssr?

 

[Seanspeed]

Question now becomes, what the hell happened with pssr?

It's just one game. We need to see a lot more before we start jumping to conclusions.

 

[trinibwoy]

You're talking about reviewers doing 'bare minimum' and ignoring all manner of things that are supposedly important - well DF haven't covered the cost of filters or overlays ever, so they're every bit as guilty of what you're accusing others of. DF may fill in some occasional gaps, but they are not themselves more comprehensively covering the things others do, either.

Well clearly I disagree with all of this. No other outlet even bothers to investigate frame timing, stutters or IQ differences in detail.

Also, quantity is very important in this realm. You cant just dismiss that. Judging performance based off a few results can be incredibly misleading, and it's why people should be checking outlets who do large sample testing, or at the very least checking out multiple different outlets to get a better overall picture.

If bar charts of avg FPS over 30 games adequately convey your personal experience with using a graphics card then be happy that your needs are being met.

 

[Seanspeed]

Well clearly I disagree with all of this. No other outlet even bothers to investigate frame timing, stutters or IQ differences in detail.



If bar charts of avg FPS over 30 games adequately convey your personal experience with using a graphics card then be happy that your needs are being met.

HUB has made tons of videos investigating differences in IQ of various techniques offered by Nvidia, AMD and Intel and all that.

Stuff about stutters and whatnot - that's not hardware review material, that's specific game performance analysis. Which yes, is DF's speciality. You're continuing to prove my point here about your inconsistency in what you're actually asking for.

 

[Blade Brown]

Question now becomes, what the hell happened with pssr?

What happened to it?

 

[trinibwoy]

HUB has made tons of videos investigating differences in IQ of various techniques offered by Nvidia, AMD and Intel and all that.

Stuff about stutters and whatnot - that's not hardware review material, that's specific game performance analysis. Which yes, is DF's speciality. You're continuing to prove my point here about your inconsistency in what you're actually asking for.

How would you like to see features like upscaling and framegen covered in GPU reviews?

 

[Seanspeed]

How would you like to see features like upscaling and framegen covered in GPU reviews?

I've made my thoughts on those clear earlier already. You even responded directly to the post where I did so. Let's not repeat ourselves.

You haven't made yourself clear on what your issues are with coverage of these things, though. You tried to make it seem like it was actually a bunch of stuff about filters and overlays that you were concerned with instead when pressed on it.

So it really was about upscaling and framegen that you were most unhappy with in review coverage?

 

[Cyan]

Spoiler: Summary

• (00:03) AMD had high expectations for CES, with rumors of new RDNA 4 GPUs and FSR 4, but these were not officially showcased due to "timing issues."
• (00:40) At the AMD booth in Las Vegas, there was no clear announcement or representation of FSR 4, but a surprise "research project" was discovered featuring machine learning-based upscaling.
• (01:39) The demo compared FSR 3.1 with a new machine learning upscaler, presumably FSR 4, showcased on two PCs, one with an RDNA 4 engineering sample.
• (02:55) The comparison was made using Ratchet and Clank, where FSR 3.1 exhibited visual artifacts (like Moiré patterns), while the new research project (FSR 4) appeared smoother with fewer artifacts.
• (03:29) The demo lacked official context, with no AMD representatives present to explain the technology in detail, leading to a somewhat "sneaky" presentation.
• (04:53) The new upscaling technology seemed to outperform FSR 3.1 in terms of image quality, with better handling of textures and reduced flickering or artifacting.
• (06:30) FSR 4 showed improved handling of background effects like ambient occlusion and textures, with fewer issues like the "Moiré pattern" that was noticeable in FSR 3.1.
• (08:27) The FSR 4 tech was sharper than previous iterations and handled particle effects (like confetti) better, displaying clear, non-trailing particles.
• (10:04) AMD's new upscaling solution, tested in performance mode, showed great quality improvements, even when scaling down from 4K, which FSR 3 struggled with.
• (11:43) The research project did not show characteristics similar to Sony's PSSR (PlayStation upscaling), meaning it was likely AMD’s proprietary technology. Further testing was needed to understand performance and final implementation.

the difference between FSR 3.1 and FSR 4 is impressive, FSR 3.1 simply destroys the confetti, while with FSR 4 you see everything, more numerous confetti and in greater detail.

FSR 3.1

FSR 4

 

[techuse]

Question now becomes, what the hell happened with pssr?

It has to operate in a more performance and memory access confined environment.