Value of Hardware Unboxed benchmarking

[DavidGraham]

Provide a link of them ever saying FSR 1 or even 2 is equivalent to DLSS.

Sigh, here it is, they say here that "FSR1 and DLSS2 are competitive, one is not necessarily better than the other"!

Here they say FSR1 in reality is pretty decent and is competitive with DLSS2 at times! Then they claim FSR1 Ultra is close to native rendering and the IQ of FSR1 is only marginally behind DLSS2!

And about FSR2, here they have the thumbnail claiming it's a DLSS killer, without even testing any games!

Here they claim they are almost the same in Deathloop.

Please stop repeating their arguments ad nauseam, you just said DLSS2 and FSR2 offer the same performance, I provided a link disproving that, then you claimed they never said FSR1/FSR2 is equal to DLSS2, I provided several links proving they did. I hope you are now satisfied.

 

[DegustatoR]

In the Witcher 3 he did start enabling hairworks as a default once GPUs became more performant.

Which is a different way of saying that he started enabling HW once it stopped being a problem for AMD h/w in comparison to Nv. Which is a very weird thing to do for benchmarking, but still, if we apply the same logic to FSR2 then FSR2 does have a performance problem on non-AMD h/w. It will always have it because it doesn't use the h/w features of said h/w.

True, but that isn't a perfect proxy for reconstruction either because each solution has a potentially different (and non-trivial) cost, and I believe some post-processing effects are applied in the final reconstructed space, so their costs are proportional to final upscaled resolution.

Yeah but as I've said it is better to do "reconstruction benchmarks" less often than just use FSR2 in all benchmarks. The latter is straight up misleading.

Doesn't DLSS performance mode provide the same image quality as FSR quality mode at 4k?

Kinda, depends on a title. But yes, this is a valid point too.

Btw HUB's Hogwarts Legacy coverage is completely different from everyone else's - again. It's like Steve is specifically searching for areas where AMD does better than Nv in games he tests.

 

[pjbliverpool]

DLSS and FSR performance are typically within 1-2% of each other on an NV GPU.

Yeah see the first post in this thread. 1%-2% can be the difference between the NV GPU being faster or the AMD GPU being faster.

Although I guess they could always throw in another CoD MW2 test at different settings to swing the overall average back

 

[techuse]

Sigh, here it is, they say here that "FSR1 and DLSS2 are competitive, one is not necessarily better than the other"!

Here they say FSR1 in reality is pretty decent and is competitive with DLSS2 at times! Then they claim FSR1 Ultra is close to native rendering and the IQ of FSR1 is only marginally behind DLSS2!

And about FSR2, here they have the thumbnail claiming it's a DLSS killer, without even testing any games!

Here they claim they are almost the same in Deathloop.

Please stop repeating their arguments ad nauseam, you just said DLSS2 and FSR2 offer the same performance, I provided a link disproving that, then you claimed they never said FSR1/FSR2 is equal to DLSS2, I provided several links proving they did. I hope you are now satisfied.

In very specific scenarios he said FSR can be competitive with DLSS. Even in your provided links he follows it up with DLSS being superior in general though. It's a thumbnail with "attempt #2" in parenthesis.

 

[DavidGraham]

In very specific scenarios he said FSR can be competitive with DLSS. Even in your provided links he follows it up with DLSS being superior in general though. It's a thumbnail with "attempt #2" in parenthesis.

They don't, they generalize, they keep pandering to whatever AMD puts through, and try to twist words to make it equal to DLSS2 as best as they can. Even in your claimed specific scenarios their analysis is wrong, FSR1 is so hopelessly behind DLSS2, it's not even close.

 

[techuse]

They don't, they generalize, they keep pandering to whatever AMD puts through, and try to twist words to make it equal to DLSS2 as best as they can. Even in your claimed specific scenarios their analysis is wrong, FSR1 is so hopelessly behind DLSS2, it's not even close.

In every review he mentions DLSS as one of the strongest reasons to go NVidia.

 

[techuse]

Steve comes clean.

Quite funny and fitting that if he used DLSS the 4070ti would have actually performed worse in the final game summary. Nvidia fans would have raged even harder about how biased HUB was for using a higher quality upscale on Nvidia to make AMD look better.

 

[pjbliverpool]

Steve comes clean.

Quite funny and fitting that if he used DLSS the 4070ti would have actually performed worse in the final game summary. Nvidia fans would have raged even harder about how biased HUB was for using a higher quality upscale on Nvidia to make AMD look better.

Not commenting on whether there is any bias or not as the results shown do seem to demonstrate that using DLSS across the board wouldn't have been in NV's favour - at least in these games. But it is worth noting that DLSS was faster in several titles, and so had it been used in those titles only (i.e. using the fastest upscaling solution available to each GPU on a case by case basis), then the overall summary would have slightly tilted a little more in NV's favour. That said, I appreciate that's a level of additional work that it might not be reasonable to expect of the reviewers.

 

[davis.anthony]

The difference is that DLSS can be set to use a lower quality mode and still match/beat FSR 2.2 in image quality.

So in reality it should be FSR 2.2 quality vs DLSS balanced - That would give quite a big win to Nvidia in terms of frame rate and still give a slight IQ advantage.

The powerlines in FH5 at 2:50 show you how far a head DLSS is and how you can gain even more performance by dropping down to DLSS balanced mode.

So I still don't understand this video as it's still unrealistic to what Nvidia owners actually use.

 

[DegustatoR]

Unique Steve strikes again. He's quite possibly the only person in the universe on whose PC FSR2 runs faster than DLSS2 on GF GPUs.
Yes, I've compared them in pretty much all games where this is possible.

But hey, make a controversial video, get clicks, make another controversial video, get more clicks.
Dunno why he even bother testing anything.

 

[techuse]

Unique Steve strikes again. He's quite possibly the only person in the universe on whose PC FSR2 runs faster than DLSS2 on GF GPUs.
Yes, I've compared them in pretty much all games where this is possible.

But hey, make a controversial video, get clicks, make another controversial video, get more clicks.
Dunno why he even bother testing anything.

So you believe he's doctoring the data in some way?

 

[DegustatoR]

So you believe he's doctoring the data in some way?

No idea what he is doing. But his benchmarks are again and again different from everyone else's.

 

[techuse]

No idea what he is doing. But his benchmarks are again and again different from everyone else's.

Can you provide the benchmarks showing FSR and DLSS performance differentials that don’t line up with his results in these titles?

 

[Scott_Arm]

I'm assuming the video in the top left of each benchmark is the test scene? It's possible that in those particular scenes his results are correct. One of the problems with games is that from one scene to another performance can change and favour one gpu over another etc. It's why comparing reviews is kind of pointless, unless reviewers show their test scenes. Most do not. Then you have the question of whether a test scene is representative of the worst case in a game, or the average case, or if it's just the easiest to benchmark because it's a tutorial section, a checkpoint etc. Lots of reviewers basically structure their reviews for high turnover of content. Ultimately you need to find one reviewer that you think covers things in the way you'd like, or take the average of a whole bunch of different reviewers. Ignore the ones you think are outliers, or aren't testing in a way that's useful to you.

 

[DegustatoR]

Can you provide the benchmarks showing FSR and DLSS performance differentials that don’t line up with his results in these titles?

Check basically any other benchmark but Steve's.

 

[troyan]

Its a useless discussion. FSR 2 is specific designed to mimic DLSS - even from a performance perspective. So the cost will be around the same. But there is a huge difference in image quality and what settings you can use.

In the end this channel is devaluing nVidia's advantage to a point that their viewers will think that only performance is the difference:
No image comparision between FSR and DLSS, no FG, no Reflex measurements, no RT...

 

[Kaotik]

Check basically any other benchmark but Steve's.

The burden of proof is on the one who disputed his benchmarks, so you.

 

[DegustatoR]

The burden of proof is on the one who disputed his benchmarks, so you.

What proof? Check any benchmark comparing the two. Any.

 

[BitByte]

Yeah see the first post in this thread. 1%-2% can be the difference between the NV GPU being faster or the AMD GPU being faster.

Although I guess they could always throw in another CoD MW2 test at different settings to swing the overall average back

Almost nobody care about 1-2% differences except for people like you. A GPU is 1-2% faster? It’s statistically insignificant.

 

[BitByte]

No idea what he is doing. But his benchmarks are again and again different from everyone else's.

This shit should be bannable. It’s literally tales from your behind. You lazily make accusatory claims with absolutely no evidence and then concoct some bs in your mind about how they’re biased. Their data is different because they use different passes in the game. Some reviewers bench with the standard benchmarks and they at times create bench mark runs that they feel are more representative of typical game performance. As most sensible people understand, game benchmarks are often not representative of real performance.

If you want to make ridiculous claims like this, go purchase the gpus and do the same benchmark pass they do, then make your claims. I’m tired of seeing lazy internet whiners put down the work of others while not providing any evidence to support their claim. Again, this shit should be ban worthy. It’s libel.