AMD Publishes FSR 3 Source Code

socK said:
The emulator can't provide the information DLSS or Xess needs in a generic way. The emulator has no idea what's being drawn and it can't make assumptions because any and every game is free to do whatever it wants.
Click to expand...
Yep, this is what was said by one of the members of the Yuzu team.
https://github.com/yuzu-emu/yuzu/issues/12525#issuecomment-1872470058

"DLSS conflicts licence-wise, and the experimental build required using video encoding as motion vectors data.
FSR3 has the same limitation, without motion vectors data, it's useless, and Switch games don't export that information.
Frame generation on TVs has a little cost called input lag, good luck getting a game with a base framerate of 30 FPS into something playable if you delay for 66ms to get input data. And that's not even counting the ghosting.

It's not impossible to implement FSR3, but unless you find a way to extract motion vectors data in a sensible way, that works for any game, it will remain as not viable."
FSR3 and temporal friends cannot be implemented by the compositor in the Steam Deck either. Has nothing to do with source code, the information they need to function needs to come from the game - which is why Steam Deck can only ever spatially upscale for you.
Click to expand...
So far games that have FSR2 can have FSR3.

View: https://youtu.be/jst_zjiDCLw?si=MIC_418bnv4An39v
 
modders have emulated frame generation for the older RTX cards using the AMD FSR 3 source code.

Digital Foundry tested one of these mods on the RTX 20 and RTX 30 series GPUs and found that it improves in-game performance by up to 75%. That’s a generational upgrade right there. The downside of this mod is that it’s only supported in titles with native DLSS 3 support. It uses the same implementation path as DLSS 3 to patch in FSR 3, a free and open-source frame-generation technology. We tested FSR 3 at launch and found it to be on par or better than DLSS 3.

The DLSS 3 mod leverages two .dll files to inject the FSR 3 “Frame Gen” code into the game. A registry tweak allows RTX 20/RTX 30 users to run the frame generation algorithm on their hardware. You’re enabling FSR 3 using the DLSS 3 path by bypassing a Windows check for RTX 40 series GPUs.

https://www.hardwaretimes.com/dlss-...us-nearly-twice-as-fast-in-the-latest-titles/
 
I've been using the Nukem9 DLSS->FSR3 mod with Cyberpunk 2077 since December and it's pretty good.

There is in specific situations a bit of weird ghosting near the frame edges, could just be a thing with the mod though idk. Overall for me it's been a net positive- movement looks smoother with FSR3 enabled which is the point. I've mostly been running with slammed gfx & crowd density settings that result in 80~120fps with Framegen on (CPU-bound on the lower end) and the quality holds up, at least it looks smoother to me than with Framegen off. At more reasonable settings, 120~160fps with FSR3 FG enabled looks and feels very smooth. 175hz monitor fwiw.

Comparing modded-in FSR3 FG on my 3090Ti vs native DLSS3.5 FG on my 4070Ti, the quality of generated frames seems generally close enough that I doubt I'd notice the difference in motion except for the weird ghosting in the FSR3 mod with very fast movement combined with certain backgrounds. Framepacing of FSR3 with the mod left a good impression, I haven't noticed any horrible frametime issues and it looks (to me, on my setup) as smooth as DLSS3.5 in motion. That's good for me, because in some games (particularly Cyberpunk) I'd much rather use the 3090Ti with its luxurious 24GB than the 4070Ti with a more restrictive 12GB.

Also, unlike existing native FSR3 implementations, the modded Nvidia Streamline implementation is not tied to FSR upscaling which is a huge deal IMO. That means FSR3 Framegen can be combined with DLSS upscaling, DLSS Ray Reconstruction, and even DLAA. If the eventual native FSR3 implementation in Cyberpunk is locked to FSR upscaling then honestly I'll continue using the mod to stick with DLSS.

I've also tried the paid LukeFZ DLSS3->FSR3 mod which does not use NV Streamline and therefore works on any GPU but requires game-specific builds. It's kind of a pain to use but it did seem to do the thing on my 7900XT. Smoother movement than with FG off and looked pretty good combined with XeSS 1.2 upscaling. Sometime I'm gonna give it a try on my A770- I have a feeling that with Framegen enabled I could really make use of the 16GB VRAM ;-)
 
Last edited:
I'm now finding out there's a number of FSR3 mods for a number of games that exist. There's a BG3 and Palworld FSR3 mod as well. I gotta give these a try and see how long my Vega 56 can go.
 
Marees said:
modders have emulated frame generation for the older RTX cards using the AMD FSR 3 source code.

Digital Foundry tested one of these mods on the RTX 20 and RTX 30 series GPUs and found that it improves in-game performance by up to 75%. That’s a generational upgrade right there. The downside of this mod is that it’s only supported in titles with native DLSS 3 support. It uses the same implementation path as DLSS 3 to patch in FSR 3, a free and open-source frame-generation technology. We tested FSR 3 at launch and found it to be on par or better than DLSS 3.

The DLSS 3 mod leverages two .dll files to inject the FSR 3 “Frame Gen” code into the game. A registry tweak allows RTX 20/RTX 30 users to run the frame generation algorithm on their hardware. You’re enabling FSR 3 using the DLSS 3 path by bypassing a Windows check for RTX 40 series GPUs.

https://www.hardwaretimes.com/dlss-...us-nearly-twice-as-fast-in-the-latest-titles/
Click to expand...

I did a little bit of testing with Cyberpunk 2077 after watching the DF video and I'm absolutely blown away how well it works. Running it @ 4K on my 3090 nearly maxed out with DLSS and it's barely making the card work and still getting a solid 60fps.

VERY impressed.
 
This week I've pivoted over to the system with the 7900XT and A770 to try FSR3 on non-NV HW, using the latest version of the LukeFZ mod on modded Cyberpunk.

I only briefly tried the 7900XT bc the hybrid mod needs work (VRAM getting real hot) but the gains were excellent. Tried both 3440x1440 XeSS Ultra Quality with max non-RT settings, and XeSS Performance with Medium RT- in both cases FSR3 very nearly doubled output framerate. Around 100-120fps with FG and noticeably smoother than native. XeSS Ultra Quality looks amazing! XeSS Performance is okay on the Radeon, better than FSR but worse than DLSS Perf at the same rez. I'll have to do more testing to confirm but I think the LukeFZ mod has less ghosting than the Nukem9 version?

I did several hours of actual gaming on the A770 with FSR3 and it's kind of a mixed bag. On one hand, the gains were less than I expected. At 3440x1440 XeSS Performance and Ultra non-RT settings, I was getting about 50-60fps with no FG, and enabling FSR3 only boosted it to around 80-90fps. On the other hand, responsiveness seems much better than I expected with such a low base framerate- 80fps with FG (so 40*2) feels extremely usable on the A770 and that's not been the case using FSR3 on the 3090Ti- there I really prefer 100fps+ with FG for things to feel responsive and look smooth. Dunno what the difference is, but in general the A770 has been impressing me with low latency and steady framepacing. On the Arc GPU, XeSS Performance looks great at 3440x1440 btw- definitely comparable to DLSS Performance at that rez and better than XeSS Perf on the 7900XT or NV cards. The XMX version of XeSS really does deliver the goods! And 16GB VRAM is quite useful even at this performance level! Loaded up with high-rez texture mods, I'm seeing 13-15GB used and the visual difference from stock textures is significant.

Didnt think I'd be having such a good time gaming on the slowest modern GPU I own but this A770 continues to impress. Even tho the FSR3 framerate gainz are on the low end it's super usable IMO. When I have a chance I'll check out the 7900XT + FSR3 combo a bit more too :)
 
GoldenTiger said:
Nice NattyKathy, great info there. Maybe Battlemage will be a huge surprise for the mid/low range?
Click to expand...
Based on leaked specs & rumors it seems likely both Battlemage and RDNA4 will be around 4070Ti/7900XT performance but at "midrange" pricing. If that pans out I'll be in for a B770 :) Intel's two-point extrapolated Framegen sounds interesting too, should have a much lower latency penalty than the three-point interpolation everyone else is doing.

Back On-topic...
I hope that AMD can get FSR3 FG in more games this year, the initial impressions may have been lackluster but IMO it seems to be in a really good state now if these modded-in implementations are anything to go by. I was ready for disappointment after all the noise NV made abt how their custom Optical Flow block was needed for DLSS3 FG, but it seems AMDs Async Compute solution does the thing.
 
NattyKathy said:
Based on leaked specs & rumors it seems likely both Battlemage and RDNA4 will be around 4070Ti/7900XT performance but at "midrange" pricing. If that pans out I'll be in for a B770 :) Intel's two-point extrapolated Framegen sounds interesting too, should have a much lower latency penalty than the three-point interpolation everyone else is doing.
Click to expand...
What exactly is midrange pricing? I consider midrange between $300-$400, and I really doubt AMD would sell a 7900XT for $400. Battlemage being able to at least match a 4070Ti is really juicy. Willing to bet Intel would sell it for $400 or less.
Back On-topic...
I hope that AMD can get FSR3 FG in more games this year, the initial impressions may have been lackluster but IMO it seems to be in a really good state now if these modded-in implementations are anything to go by. I was ready for disappointment after all the noise NV made abt how their custom Optical Flow block was needed for DLSS3 FG, but it seems AMDs Async Compute solution does the thing.
Click to expand...
It's a problem when the community has made patches for gemes to get FSR3. Good for the community, but nobody is going to know about them. FSR3 should be pushed harder by AMD because it's not like Nvidia is going to let off on DLSS.
 
Last edited:
DukenukemX said:
It's a problem when the community has made patches for gets to get FSR3. Good for the community, but nobody is going to know about them. FSR3 should be pushed harder by AMD because it's not like Nvidia is going to let off on DLSS.
Click to expand...
AMDs momentum with FSR in general has been slow- but steady. FSR2 had slow adoption too and for awhile needed mods to work in games like Cyberpunk, but now a couple years later it's everywhere. I have a feeling in a couple years FSR3 will get to that point too.

DukenukemX said:
What exactly is midrange pricing? I consider midrange between $300-$400, and I really doubt AMD would sell a 7900XT for $400. Battlemage being able to at least match a 4070Ti is really juicy. Willing to be Intel would sell it for $400 or less.
Click to expand...
I think unfortunately solid "midrange" these days is like $450~550 looking at perf/$ scaling and perf tiers.
$300~$400 "low-mid".

If B7x0 perf is closer to 4070, I imagine similar price tier to Alchemist. If perf is closer to 4070Ti, maybe more like $500. Nobody is gonna release a GPU with per/$ that's "too good" but given how aggressive Intel & their board partners have been with pricing with Alchemist I doubt they'll try to NV-levels of poor perf/$.

I hope we get more info on Xe FG soon, I am really excited for that.
 
NattyKathy said:
Based on leaked specs & rumors it seems likely both Battlemage and RDNA4 will be around 4070Ti/7900XT performance but at "midrange" pricing. If that pans out I'll be in for a B770 :) Intel's two-point extrapolated Framegen sounds interesting too, should have a much lower latency penalty than the three-point interpolation everyone else is doing.

Back On-topic...
I hope that AMD can get FSR3 FG in more games this year, the initial impressions may have been lackluster but IMO it seems to be in a really good state now if these modded-in implementations are anything to go by. I was ready for disappointment after all the noise NV made abt how their custom Optical Flow block was needed for DLSS3 FG, but it seems AMDs Async Compute solution does the thing.
Click to expand...
I'm sure they will... Probably will see a lot of new steam deck like devices
 
blandead said:
I'm sure they will... Probably will see a lot of new steam deck like devices
Click to expand...
Laptops and deck like devices are expected to make up the overwhelming bulk of “PC” gaming devices in the next 5 years, given the decline in desktop sales so….
It’s why AMD and Intel are so focused on it.
 
Lakados said:
Laptops and deck like devices are expected to make up the overwhelming bulk of “PC” gaming devices in the next 5 years, given the decline in desktop sales so….
It’s why AMD and Intel are so focused on it.
Click to expand...

Laptops will never become gaming devices for most, they lack the ability to have any upgrade path at all and the graphic cards are always weaker in laptop form. People are just not buying because prices are way up on computers in general so is everything else. People have reached spending fatigue and manufactures have dug in on not wanting to drop prices. Will se who wins the battle in the end.
 
Gideon said:
Laptops will never become gaming devices for most, they lack the ability to have any upgrade path at all and the graphic cards are always weaker in laptop form. People are just not buying because prices are way up on computers in general so is everything else. People have reached spending fatigue and manufactures have dug in on not wanting to drop prices. Will se who wins the battle in the end.
Click to expand...
Umm… hate to break it to you but Laptop gamers already outnumber desktop gamers almost 2 to one and the number is growing not shrinking.
 
Lakados said:
Umm… hate to break it to you but Laptop gamers already outnumber desktop gamers almost 2 to one and the number is growing not shrinking.
Click to expand...

Laptops outnumber desktops, never disputed that. Now people using them for gaming, much lower, even in my own group of friends, only 1 uses a laptop for gaming. Having a weak discrete gpu inside a laptop does not make it a gaming machine. Heck even my nephew is using a desktop over a laptop to game. And this article states it pretty clearly that Laptop gaming is not bigger than desktop gaming. https://playtoday.co/blog/stats/pc-gamer-demographics/
 
Gideon said:
Laptops outnumber desktops, never disputed that. Now people using them for gaming, much lower, even in my own group of friends, only 1 uses a laptop for gaming. Having a weak discrete gpu inside a laptop does not make it a gaming machine. Heck even my nephew is using a desktop over a laptop to game. And this article states it pretty clearly that Laptop gaming is not bigger than desktop gaming. https://playtoday.co/blog/stats/pc-gamer-demographics/
Click to expand...
Yes it is, between laptops and steam deck like devices they are expected to make up almost 90% of the pc gaming market by 2030 if current trends continue.

Every year we get the new dGPU’s shipped at a new all time low, the other side of that is gaming laptop sales reach a new all time high. AMD and Nvidia sell more laptop based gaming graphics cards than they do desktop ones by a wide margin.
 
Last edited:
Gideon said:
Laptops will never become gaming devices for most, they lack the ability to have any upgrade path at all and the graphic cards are always weaker in laptop form. People are just not buying because prices are way up on computers in general so is everything else. People have reached spending fatigue and manufactures have dug in on not wanting to drop prices. Will se who wins the battle in the end.
Click to expand...
The past 4 years had so many people buy graphic cards that the market for desktops is over saturated. People aren't buying desktops because peoples desktops are already more than powerful enough. Incremental upgrades are just not enough to get Bob to replace his 12GB RTX 3060. Up until recently, Bob was rocking a GTX 1060 for the past several years. The fact that the Steam Hardware Survey still has a lot of GTX 1060's, 1650's, and 2060's shows that people are just waiting for the right upgrade.
Lakados said:
Yes it is, between laptops and steam deck like devices they are expected to make up almost 90% of the pc gaming market by 2030 if current trends continue.

Every year we get the new dGPU’s shipped at a new all time low, the other side of that is gaming laptop sales reach a new all time high. AMD and Nvidia sell more laptop based gaming graphics cards than they do desktop ones by a wide margin.
Click to expand...
Going by the current Steam Hardware Survey, I don't see Laptops being a prominent source of gaming.
  • NVIDIA GeForce RTX 3060 Laptop GPU 3.86%
  • NVIDIA GeForce RTX 4060 Laptop GPU 2.61%
  • AMD Radeon Graphics 2.19% <-- probably laptop
  • Intel Iris Xe Graphics 2.01% <-- probably laptop
  • Intel(R) UHD Graphics 1.50% <-- probably laptop
  • NVIDIA GeForce RTX 3050 Ti Laptop GPU 1.10% <-- probably laptop
  • AMD Radeon(TM) Graphics 0.88% <-- probably laptop
  • NVIDIA GeForce RTX 3070 Laptop GPU 0.85%
  • NVIDIA GeForce RTX 3050 Laptop GPU 0.74%
  • NVIDIA GeForce RTX 4050 Laptop GPU 0.70%
  • AMD Radeon Vega 8 Graphics 0.57% <-- probably laptop

I wouldn't say that most people are gaming on laptops. Have you tried gaming on a laptop? Not a very fun experience. It comes in handy when you're away from your desktop. We tend to go into a cycle that's based on consoles capabilities. The PS5 and Xbox Series X are weak consoles compared to what was available on PC on release. Since most games won't go past the capabilities of these consoles, you don't need super amazing hardware to play games on PC. A lot of people prior to the release of this console generation was already capable of playing all the games, like a GTX 1060, 1070, 1080, with AMD having their RX 580, Vega 56/64, and etc. The equivalent to a PS5 in graphics output is a RX 6600 XT, or a RTX 2060 and those aren't expensive right now. People are getting better laptops because it's probably time for an upgrade. Even I upgraded my laptop to one with a Ryzen 5500U just so I could game when I'm away. It's not very good at it, but again I don't game primarily on a laptop.
 
Lakados said:
Yes it is, between laptops and steam deck like devices they are expected to make up almost 90% of the pc gaming market by 2030 if current trends continue.

Every year we get the new dGPU’s shipped at a new all time low, the other side of that is gaming laptop sales reach a new all time high. AMD and Nvidia sell more laptop based gaming graphics cards than they do desktop ones by a wide margin.
Click to expand...

You said they are already 2 to 1, which is clearly untrue as 2030 was never mentioned till now. What are we defining as a gaming laptop, because I have seen integrated graphics called a gaming laptop, when clearly they really are not. So let me know how the definition is, because it seems like it keeps changing.
 
https://focusonbusiness.eu/en/news/...to-jump-by-25-and-hit-62-million-in-2024/4007
Shows mobile gaming silicon moving 2x faster than Destop-based silicon starting as early as 2020

An older report from Forbes in 2022 showed that nearly 40% of dedicated PC gamers then were using a laptop as their primary, with a significant number using it as a secondary.
https://www.forbes.com/sites/gigabyte/2022/02/01/next-generation-gaming-on-the-go/?sh=2024bb9b4e1f

Statistics show that gaming notebooks represent the market’s largest segment, with 22.3 million sold units and a 44.9% market share in 2020. By 2024, this segment is expected to reach almost 50% market share and 30.2 million sold units globally, a 35% jump in four years.

Gaming desktop PCs are forecast to account for nearly 30% of shipments in the global gaming PC market this year, with 14.8 million sold units. Statistics show the market share of this segment is expected to drop to 25.5% in 2024, although shipments will rise to 15.8 million units.

So in 2022 they were predicting 3x as many gamers playing on laptop devices as PC devices by 2024 and based on sales trends and dGPU sales figures that seems to be on track.

If the mobile silicon sales trends keep at the same pace they currently are, and the dGPU sales continue in their existing downward trend then by 2030 we are looking at a 90% laptop gaming user base, so a 9-1 ratio in laptop gamers vs desktop gamers. That is of course if the trend continues.

WFC article on dGPU shipment figures over time.
https://wccftech.com/gpu-market-rebounds-q2-2023-amd-nvidia-intel-increased-shipments-discrete-gpus-up/#:~:text=GPU Shipments Per Segment For Q2 2023&text=The integrated segment had a,in the entry-level segment.

this is their numbers then on desktop vs laptop GPU sales and you will see that the High End for gaming laptops then moves nearly 2x as many units, that is a snapshot for just 1 quarter but just about every quarter for the past 4 yeras has held pretty close to this pattern.
1707954956997.png



Now whether you want to consider any of this a good thing or not is irrelevant (desktop to the end for me), but that is the way the trends are going, which explains much of AMD's and Intel's decisions for their releases and leaks.
Intel has not had much planned out past 3 generations of dGPUs aka another 5 years or so (taking them to 2029) as well as many of AMD's APU and mobile chip decisions as well as their gGPU strategies. It also frames Nvidia's commitment to transitioning to an AI company because if this trend continues we are only going to be seeing Nvidia GPUs paired with Intel mobile CPUs because you know AMD isn't going to let Red and Green exist on their platform together if they have their way. So Nvidia is planning for the event they get squeezed out of the consumer market, which is smart, not something I like the idea of because then we are just staying with a 2 player game but it is something they need to consider unless the PC gaming market suddenly gets ARM friendly.
 
Last edited:
Lakados said:
https://focusonbusiness.eu/en/news/...to-jump-by-25-and-hit-62-million-in-2024/4007
Shows mobile gaming silicon moving 2x faster than Destop-based silicon starting as early as 2020

An older report from Forbes in 2022 showed that nearly 40% of dedicated PC gamers then were using a laptop as their primary, with a significant number using it as a secondary.
https://www.forbes.com/sites/gigabyte/2022/02/01/next-generation-gaming-on-the-go/?sh=2024bb9b4e1f

Statistics show that gaming notebooks represent the market’s largest segment, with 22.3 million sold units and a 44.9% market share in 2020. By 2024, this segment is expected to reach almost 50% market share and 30.2 million sold units globally, a 35% jump in four years.

Gaming desktop PCs are forecast to account for nearly 30% of shipments in the global gaming PC market this year, with 14.8 million sold units. Statistics show the market share of this segment is expected to drop to 25.5% in 2024, although shipments will rise to 15.8 million units.

So in 2022 they were predicting 3x as many gamers playing on laptop devices as PC devices by 2024 and based on sales trends and dGPU sales figures that seems to be on track.

If the mobile silicon sales trends keep at the same pace they currently are, and the dGPU sales continue in their existing downward trend then by 2030 we are looking at a 90% laptop gaming user base, so a 9-1 ratio in laptop gamers vs desktop gamers. That is of course if the trend continues.

WFC article on dGPU shipment figures over time.
https://wccftech.com/gpu-market-rebounds-q2-2023-amd-nvidia-intel-increased-shipments-discrete-gpus-up/#:~:text=GPU Shipments Per Segment For Q2 2023&text=The integrated segment had a,in the entry-level segment.

this is their numbers then on desktop vs laptop GPU sales and you will see that the High End for gaming laptops then moves nearly 2x as many units, that is a snapshot for just 1 quarter but just about every quarter for the past 4 yeras has held pretty close to this pattern.
View attachment 635119


Now whether you want to consider any of this a good thing or not is irrelevant (desktop to the end for me), but that is the way the trends are going, which explains much of AMD's and Intel's decisions for their releases and leaks.
Intel has not had much planned out past 3 generations of dGPUs aka another 5 years or so (taking them to 2029) as well as many of AMD's APU and mobile chip decisions as well as their gGPU strategies. It also frames Nvidia's commitment to transitioning to an AI company because if this trend continues we are only going to be seeing Nvidia GPUs paired with Intel mobile CPUs because you know AMD isn't going to let Red and Green exist on their platform together if they have their way. So Nvidia is planning for the event they get squeezed out of the consumer market, which is smart, not something I like the idea of because then we are just staying with a 2 player game but it is something they need to consider unless the PC gaming market suddenly gets ARM friendly.
Click to expand...

Once again no one is disputing more laptops sell then desktops especially the last few years, they often make the most sense for students and business people on the move.

Your second article is saying 38% use a gaming laptop, which makes a certain amount of sense because I am sure most college students use a gaming laptop. But what are we defining as a gaming laptop, which they don't go into at all. Which is not surprising since it's a paid article for a Gigabyte laptop. This is still far short of that 2 to 1 you claimed which made me comment again. The rest of the issue, is steam does not track this supposed explosion in gaming laptops, many laptops get disposed as you can't upgrade them so sales numbers being higher is a necessity. Were long past the days of people buying desktop gateways and such. Discrete shipments are in the toilet because of pricing and many people already upgraded to a card that is still getting the job done these days. If this changes than you will see a large jump in discrete sales.

So lets now take this chart. We have no midrange laptops for some reason just for desktops which already makes me question this chart. So we have to combine midrange and high end for desktops which puts us at 4.78 to 4.65 for notebooks. Slightly higher but definitely shows there is a large appetite for gaming laptops but not amazingly higher, but does give you credit that there could be a market shift happening at some point. Entry level and below I am struggling to count as a gaming machine, but even if we do it's barely changing anything. Integrated is a work computer and I will not count as a gaming computer, I don't think anyone would but you are welcome to argue why it should be counted. But yeah work and student laptops definitely are far higher then the desktop sales without a doubt.

Good or bad, really doesn't matter to me but I think were a bit early to state Laptops will dominate the future, especially when cpu's and gpu's keep getting hungrier for power to get the experience people want. As for info AMD, Intel or Nvidia puts out, yeah about all we ever get to see is about 5 years of detailed info and then it's get vague after that. Nvidia confirms it's commitment to whatever they think will make them the most money, like you know their commitment to mining cards and such. As for AMD I don't think they care if it has a Nvidia gpu if it means selling a AMD cpu, at least they got some money out of it. They are quite content to be the second player for video cards for now there focus may shift due to the AI craze though. Nvidia has always been worried they were going to get squeezed out, thus why they wanted ARM so bad. Though I really doubt that will ever happen, otherwise Intel would have gave it a go some years ago. Instead they just have a love hate relationship.


So in summary maybe one day you will be right, but so many have said the desktop is dead and over and it's still here doing just fine. Now mobile gaming on a cell phone, well yeah that exploded and birthed micro transactions, god got to love that invention. Cell phones definitely sell at a much higher rate then computers these days but I don't see that kind of adoption for Laptops or me and the many people I know are all anomalies along with this forum. So will see what happens 6 years down the road but I don't want to derail this thread any farther then we both already have.
 
Gideon said:
Laptops will never become gaming devices for most, they lack the ability to have any upgrade path at all and the graphic cards are always weaker in laptop form. People are just not buying because prices are way up on computers in general so is everything else. People have reached spending fatigue and manufactures have dug in on not wanting to drop prices. Will se who wins the battle in the end.
Click to expand...
I think external gpu connectors will become a thing again
 
DukenukemX said:
I'm in the camp that I don't like TAA. I'd rather use SMAA
Click to expand...
Yeah man!!! This was my favorite version of AA before I experienced the wonder that is DLAA. I used to be a fan of FXAA despite its flaws, and I never really liked TAA. But SMAA is some good shiznit! I agree that not enough games have it as an option. I still see FXAA around sometimes, and TAA is always around, but SMAA I rarely see. That is a damn shame.

socK said:
MSAA will be expensive and does nothing to help aliasing that is coming from shading instead of geometry.
Click to expand...
Lakados said:
Problem is MSAA doesn’t work with modern rendering setups, MSAA does not work on textures only polygons, and it struggles with any objects that use opacity or programmable shaders. It also incurs a higher overhead as it causes each frame to essentially get rendered 1.5 times over.
Click to expand...
SMAA, not MSAA. As in "Subpixel Morphological Anti-Aliasing".
 
DrezKill said:
Yeah man!!! This was my favorite version of AA before I experienced the wonder that is DLAA. I used to be a fan of FXAA despite its flaws, and I never really liked TAA. But SMAA is some good shiznit! I agree that not enough games have it as an option. I still see FXAA around sometimes, and TAA is always around, but SMAA I rarely see. That is a damn shame.



SMAA, not MSAA. As in "Subpixel Morphological Anti-Aliasing".
Click to expand...
SMAA doesn't work on transparent textures, or any texture with an alpha value less than 1.0, and by extension any object behind it, it also tends to smear at high frame rates or during fast movement same problems as MSAA, but SMAA could add a temporal component hence the extra S2x, T2x, and 4x variants of it, but those get less stable during movement and generates a ghosting effect, so the higher the framerate the worse it gets. TAA is designed to fix jagged edges in motion so when moving things look better but at the cost of them looking worse when standing still.
 
You must log in or register to reply here.
Back
Top