Digital Foundry Article Technical Discussion [2025]

Who is saying DF is getting preferential treatment? And also, what could be the motivation for such claims?
When 40 series came out Digital Foundry got to show off DLSS Frame Generation before anyone else. Presumably because DigitalFoundry actually does a goodjob covering such technologies. Similarly GamersNexus also got to show off 40 series cooler before anyone else, for similar reasons. Same reason why Intel went to Digital Foundry to show off Xess before anyone else. Also too lazy to look it up, but I believe Battlenonsense also got to show Reflex or something first before other channels.. again presumably because their previous efforrts on subject.
 
What makes you so sure that AIB's margins aren't tight?

https://videocardz.com/newz/nvidias...ure-on-board-partners-msrp-feels-like-charity

This is like the tenth time I have heard this story in the last years.

The story you are getting is from those with relations with the AiBs. You know how that poster brings up the conflict of interest potentially with reviewers/influencers and manufactuers like Nvidia? What do you think AiBs are? AiBs have historically had strong relations with tech media that targets DIY and enthuasists and marketing to that group since the start of the whole industry.

The EVGA example that gets cited is rather interesting because the primary reporting that people use for that is from an influencer with a relationship with said AiB including their owner.

Or I could be wrong and the people who comment on this do have their own information on this matter that isn't sourced just from influencers. But if not it seems like if you question the influence/relationship with Nvidia you'd do that with any other company and relationship with the information source?

It's worth mentioning we have 2 major AiBs that are publicly traded companies with publicly available financial data that primary focus on graphics and have GPU vendor links -

PC Partner is a Nvidia AiB (think Zotac etc.) -


TUL is primairly an AMD AiB (think Powercolor etc.) but now also does Intel -


Look at the gross profit over revenue.
 
Last edited:
I'm a dummy with this. What should we look at? Under Profit Margin one says 2.36% and the other says -3.09%.

If you take gross profit over revenue (gross proft / revenue) it's the gross margins.

I wanted to just present data and give as little commentary as possible but margins you can't really compare across industries (eg. an AiB vs. Nvidia's gross margins aren't comparable. Historically Nvidia and AMD at <40% gross margins had basically negative overall income). With this we're just looking at if an Nvidia AiB is suffering from lower margins than say the alternative.
 
@Andrew Lauritzen You may have never had hands-on with it, but I'm curious if you're aware of any effect Apple's "Dynamic Cache" has had on running maybe larger shaders, or an "uber" shader approach. This feature was added in the M3, and I was kind of expecting the other gpu vendors to follow suit. A number of caches, including the register file, can dynamically resize as needed during execution.


1738634469813.png
 
Tech conspiracists are so annoying. Always coming up with random tweaks they take on as religion. The worse being latency bros. The amount of settings they change for placebo and end up completely screwing up their experience is down right impressive. Remember it takes no credentials to have a YT channel.
Off topic but this comment reminded me: this type of nonsense is huge specifically in VR btw. You'll see people using crazy supersampling setups (like 150% scale resolution) swearing it "increases clarity" on their 1440p headsets. Some disable hyperthreading swearing it improves performance.

When I used to play DCS I'd get so frustrated with the self-appointed "performance experts" who would tell unsuspecting people to basically destroy their rigs in an effort to improve performance.

That said: I believe there are well documented issues with Nvidia's MPO implementation.
 
Question though, DX12/Vulkan PSOs were meant to solve the problems with compilation stuttering, knowing all the state ahead of time and being able to switch states more quickly... but that was only ever going to work if PSOs were created during load-time, right?. They had to have known this.
Yes, drivers still maintain the right to invalidate any sort of cache, so it always has to be a JIT. The idea was to do it at load time (similar to what most games do) or potentially at install time. The notion that some of this could potentially be crowd-sourced as well was discussed from the beginning too, but industry-collaboration kind of efforts tend to be harder to materialize.

I've noted previously but will note again that this isn't purely a DX12/Vulkan-introduced issue... if you go run Fortnite on DX11 you'll get a fair bit of shader compilation issues. It's a combination of the new APIs and the fact that games are just using a whole lot more shaders than they ever have, and thus the magnitude of the issue is a lot more than it was in the past.

At this point, I want the Fossilize "solution" because we simply can't rely on devs to a) do it in the first place.. or b) do a thorough job of it.. and I'd like the ability of the community to at least be able to mitigate it.
For sure, I think this is a no-brainer considering how well it appears to work on Steam Deck. Microsoft could do something like this at the OS level, but also nothing prevents Steam from doing fossilize with DX12 as well, and that would be another nice perk for Steam. Could be injected similar to fossilize, or could just be a simple Steam API to register your PSOs and a step that prebakes the cache when you launch or install a game (or on driver update, etc). There's a little bit of complexity with stuff like the agility SDK updates, but those are infrequent enough that it's fine to just wipe the caches in those cases.

@Andrew Lauritzen You may have never had hands-on with it, but I'm curious if you're aware of any effect Apple's "Dynamic Cache" has had on running maybe larger shaders, or an "uber" shader approach. This feature was added in the M3, and I was kind of expecting the other gpu vendors to follow suit. A number of caches, including the register file, can dynamically resize as needed during execution.
Yes I'm definitely aware of it and any advancements in that direction are very interesting. Unfortunately I haven't had a chance to experiment with it personally so can't comment on how well it works in practice.
 
When 40 series came out Digital Foundry got to show off DLSS Frame Generation before anyone else. Presumably because DigitalFoundry actually does a goodjob covering such technologies. Similarly GamersNexus also got to show off 40 series cooler before anyone else, for similar reasons. Same reason why Intel went to Digital Foundry to show off Xess before anyone else. Also too lazy to look it up, but I believe Battlenonsense also got to show Reflex or something first before other channels.. again presumably because their previous efforrts on subject.
Exclusive access/content is strongly correlated to increased views on the channel and potential channel growth. Essentially, the coercive relationship between manufacturer and reviewer incentivizes favourable behaviour with the carrot and stick model as it relates to content and hardware. It incentivizes excessive praise and minimal criticism.

You don’t need to be a genius to see how this reward structure is ethically questionable and frankly troublesome. Many times, several channels are recruited whether knowingly or unknowingly as part of an informal marketing arm.

It is why I’ll always advocate for reviewers to purchase their products. It helps alleviate and dispel any ethical concerns that may arise. Another way to address these concerns would be to review the product in a way that just shows the results of your findings without presenting opinions as to how one views the product. This way, the audience can draw their own conclusions. However, this may not be as engaging because part of the appeal of a channel is the personalities involved.

Anyway, it’s most certainly a difficult topic but, it’s a conversation that should be had both internally and externally. If it’s not discussed, then there’s no check and balance to prevent this relationship from spiralling into something like Linus Tech Tips. There, they just recommend whomever pays more/signs ad contracts with them… however, if it’s a small company, they just use them as a whipping boy in a feeble attempt to remain “balanced”….
 
Last edited:
Yes, drivers still maintain the right to invalidate any sort of cache, so it always has to be a JIT. The idea was to do it at load time (similar to what most games do) or potentially at install time. The notion that some of this could potentially be crowd-sourced as well was discussed from the beginning too, but industry-collaboration kind of efforts tend to be harder to materialize.
Right. So as games were naturally progressing towards streaming with minimal to no loading screens, there must have been a realization that much more of this was going to be needed to be done upfront (at first boot).. so I'm just curious as to whether the new APIs were being utilized like how you guys envisioned they would be when they were created, or whether you feel like developers didn't really adjust to it as you expected initially? I ask because from what I've read it seems like both APIs have changed quite a bit over the years through extensions and whatnot.
For sure, I think this is a no-brainer considering how well it appears to work on Steam Deck. Microsoft could do something like this at the OS level, but also nothing prevents Steam from doing fossilize with DX12 as well, and that would be another nice perk for Steam. Could be injected similar to fossilize, or could just be a simple Steam API to register your PSOs and a step that prebakes the cache when you launch or install a game (or on driver update, etc). There's a little bit of complexity with stuff like the agility SDK updates, but those are infrequent enough that it's fine to just wipe the caches in those cases.
MS could do something like this at the OS level would be the dream. The next best thing is Steam, as you said.. but I'm under the impression that if Steam was able to do it, they would have already. I didn't think it was possible for a 3rd party to create a Fossilize equivalent for DX12 due to the lack of plug-in (layers) support? The person who said that also said someone like Valve COULD implement a hack-y solution to make it work without Microsoft, but it would mess with anti-cheat/DRM and thus will likely never be done that way.

I prodded a little bit asking devs in the DirectX discord server why there wasn't a Fossilize type of feature for DirectX and the DirectX dev basically told me they're aware of it and aware of the issue, and stressed that they are working on things.. but obviously couldn't talk about anything. I think if Microsoft are serious about wanting to make Windows better for gaming and more handheld/UI friendly for gamers.. they should very much consider this an essential feature to get to consumers. It's the biggest perk that SteamOS handheld devices have over Windows devices, IMO.
 
Back
Top