King_Dodongo
Bob-omb
I will use next weekend just to play FEngage
Your talk about the 3060 and ray tracing reminds of of Nvidia GDC 2021 demo of Wolfenstein running with an rtx 3060 and MediaTek Kompanio 1200 Arm processor.Unrelated:
I thought I'd have a go at quantifying RT performance across arches in a way that would be useful for comparing, say, Drake to Series S. Often the idea that Drake can keep up with Series S in one realm or another comes up, and I want some more data for that. I had mixed results
Methodology:
Basically, my goal was to quantify the frame time cost for RT effects by running a game at fixed settings under a bench mark, then the same benchmark with only the RT settings changed. So for example:
RTX 3060, Control, High Settings, 1440p: 159fps, 6.3ms/frame
RTX 3060, Control, High Settings+RT, 1440p: 28fps, 35.7ms/frame
That's a cost of 29.4ms to generate the RT effects. There are a few problems here
The results
- "RT effects" are a bunch of different things. Comparing multiple games with wildly different implementations (shadows? reflections? GI?) will give wildly different results.
- RT isn't just augmenting, it's replacing rendering features. So the cost is actually higher than this number. If some aspect of the rasterized lighting path performed very poorly, then these numbers might make the RT cores seem unusually good. Conversely, excellent raster perf might penalize the RT numbers here.
- I don't have a benchmarking rig, so this requires pulling data from places that compare multiple cards under identical conditions, and have separate raster and RT benchmarks which otherwise match settings. That's basically Digital Foundry.
- You want to find a way to compare cards across arches fairly. The way I have done this in the past is to match for CU/SM, and then do some slight corrections for Clock speed. Essentially, this gives us comparing RT performance flop-per-flop across cards, with as little futzing with the numbers as possible
- These two things mean we have very few data points
When comparing across cards directly, the Ampere advantage is pretty consistent - 1.9-2.2x performance. When accounting for clocks, however, it starts to get wonkier. AMD keeps pushing the clocks further and further as it gets up the stack, but the RT performance doesn't improve as much, leaving a wider 2.2-2.7x perf range.
Series S vs Drake
Obviously, we can't compare Series S to Drake when not only do we not have the same games to test on both, one of the two pieces of hardware doesn't exist.
Secondly, these tests are running identical CPU/RAM/Mobo configs. We've isolated the performance of the GPU, but on console, the whole hardware has to be taken as a package. RT can be CPU intensive, and no one is debating Series S will have a tremendous CPU advantage. In that sense, this raw comparison heavily favors Drake
Third, we don't know Drake's clocks yet, which is a remaining piece of this whole dang puzzle.
Nor the target resolutions! RT+DLSS interact in non-obvious ways when it comes to reflections especially...
Spit it out man!
Drake's raw RT performance is ~2/3rds of Series S, depending on where the clocks are. Drake's target resolution is likely beneath Series S, which will help with what effects it will be reasonable to use RT on.
The idea that Drake will kick Series S's pants because of DLSS + RT seems unfounded, but this is one place where the meme of a "handheld series S" might actually be kinda true.
Your talk about the 3060 and ray tracing reminds of of Nvidia GDC 2021 demo of Wolfenstein running with an rtx 3060 and MediaTek Kompanio 1200 Arm processor.
I am curious on the level of RT that the NuSwitch can use, I can believe why people think that the NuSwitch can deliver a more robust level of Ray Tracing in comparison to the Series S but I am not going to say they can hit 4k 30fps with RT...maybe something more in the line of 1080p 30fps with RT + DLSS or maybe at it's absolute best 1440p with RT+ DLSS in Nintendo games...low balling because we don't know anything about the clocks or rt capabilities but it's exciting to know what is possible.
This guy does find some nuggets.Is there hope for Retro Studios to have an unannounced title close to completion besides Metroid Prime 4?
Their last new game was in 2014, which they ported to Switch in 2018. Nothing else. That seems like a very long time to be financing a studio with nothing to show for. Then they supposedly started MP4 in early 2019.
Could it be that they have another game in development that is too ambitious for the Switch that they retargeted to next-gen?
They could have a smaller team optimizing that game for the finalized hardware and Nintendo is just sitting on it.
Re: Whether 480p->720 or 480->1080 would look better:
Using the standard modes available in PC games I couldn't compare 480->720 and 480->1080, but 360->720 (Performance) and 360->1080 (Ultra Performance) was possible, so that's what I went with. Went through opening cutscenes of Edge of Eternity, mashing F12 at scene changes hoping to get some decent images very very close to each other for both. Then as with previous 720/1080 comparisons I've done, nearest neighbor scaled them up to 4K as the least common multiple. Some I split in half down the middle, some I cropped the same half and put side-by-side. I think the latter turned out much better. For both object edges and textures, I think 360->1080 is a winner over 360->720. FWIW, this is DLSS 2.2.11.0.
I think a goal for Nintendo moving forward, as far back as when Iwata was still alive, was eShop continuity (one of the favoured interpretations of Iwata's "family of systems" comments that fell away in an effort to drive the pro-revision agenda) and the only effective way to achieve that is through BC.What point is he making that’s all that good or worth debating? The merits of ensuring backwards compatibility?
Not something I really care to debate. Not having backwards compatibility is bad for consumers, and kills any confidence in investing in Nintendo as an ecosystem. It’s the literal opposite of a smooth transition and Furukawa’s goals:
Pray tell what lasting relationship has Nintendo built if they tell us that the successor system is a clean break from the Switch? My NSO icon collection?
I think a goal for Nintendo moving forward, as far back as when Iwata was still alive, was eShop continuity (one of the favoured interpretations of Iwata's "family of systems" comments that fell away in an effort to drive the pro-revision agenda) and the only effective way to achieve that is through BC.
We largely have a pretty good idea of what happened with Retro after 2014. They worked on a game for several years which Nintendo eventually canned. After which they were put on porting Tropical Freeze, and remastering/remaking Prime, and/or the entire Prime trilogy, while also assisting with MP4 when it was under development with Bamco. Eventually they took over MP4 entirely, and the project to remake all of the Prime games became just Metroid Prime, with the rest to come later. Allegedly Prime 1 has been "done" for some time now.This guy does find some nuggets.
I keep thinking that Retro Studios are very secretive and that they may have an ace up their sleeves.
(Yes, I'm quoting myself because I need validation to justify my existence, and I seem to be the only one willing to grant it) /s (or not)
Metroid Prime FIVE babyRetro had brought in 2 Writers on a contract back in July. I’ve seen them. Who is this person?
Retro hired an engineer programmer from Rockstar to be a gameplay engineer in July also
Agreed. My favorite has been the animal crossing one just because of how simplistic it isNow that I think about it. Lots of Switch Special editions have been ugly.
Never really thought about it until now.
Now that I think about it. Lots of Switch Special editions have been ugly.
Never really thought about it until now.
Unless they did something big with the Zelda remakes I doubt they'd save them for DrakeI wonder if Prime 1 remake/remaster is being saved as a Drake launch title. That and those alleged Zelda remakes that we heard about last year.
4 tf is the best possible clock speed scenario. It's almost certainly less than that.DF droped a video talking about performance of 4tf consoles today and I find it interesting because 4tf is the romored performance of T239 in docked mode.
This video provides a good reference as what we can expect for the next switch.
I just hope every Retro game will release for real, we need this studio back.
And I’m sure they could deliver us very good looking games on an hardware like Drake
Well if they were all 4K remasters then there's only one system I could see that being useful for. Maybe they'd do a regular 1080p version for the original Switch also.Unless they did something big with the Zelda remakes I doubt they'd save them for Drake
Them upscaled to 4k is a very small difference, you can see people doing it in videos. Now if they revamped the textures and lighting and it ran at 4k I could see it being for Drake. But just a 4k resolution bump probably isn't enough for Nintendo to make it a Drake releaseWell if they were all 4K remasters then there's only one system I could see that being useful for. Maybe they'd do a regular 1080p version for the original Switch also.
Not all Nintendo Switch special editions have had the Nintendo Switch logo, so I don't think that's proof that it's real or fake.* Hidden text: cannot be quoted. *
It doesn't seem like any videos with the TOTK OLED have been taken down by Nintendo. Was this the case for the Splatoon OLED when it leaked?Not all Nintendo Switch special editions have had the Nintendo Switch logo, so I don't think that's proof that it's real or fake.
Wasn't that leaked just prior the announcement?It doesn't seem like any videos with the TOTK OLED have been taken down by Nintendo. Was this the case for the Splatoon OLED when it leaked?
I think it was like a week or something I'm not sure to be honest I didn't really pay attention. I was wondering if people that were making videos on it if they were taken down. It might be an indication on wether the TOTK OLED is real or not maybe.Wasn't that leaked just prior the announcement?
I didn't want to go in with a version I hadn't yet tried out, but I'll give it a go now. I tried to recreate the last four images and added them to the bottom of the same gallery. To me it doesn't seem like a big change--sometimes an edge rougher in one version or the other, seems like luck of the frame.Are you able to use and test DLSS 2.5.1 (using DLSS Swapper or the .dll file swap)?
From the comparisons, I think the hair mostly resolves better in the latest DLSS version than compared with version 2.2.11.I didn't want to go in with a version I hadn't yet tried out, but I'll give it a go now. I tried to recreate the last four images and added them to the bottom of the same gallery. To me it doesn't seem like a big change--sometimes an edge rougher in one version or the other, seems like luck of the frame.
It was leaked 1 day before the announcement, if I remember correctly, it didn't make any sense to start the claims.It doesn't seem like any videos with the TOTK OLED have been taken down by Nintendo. Was this the case for the Splatoon OLED when it leaked?
Re: Whether 480p->720 or 480->1080 would look better:
Using the standard modes available in PC games I couldn't compare 480->720 and 480->1080, but 360->720 (Performance) and 360->1080 (Ultra Performance) was possible, so that's what I went with. Went through opening cutscenes of Edge of Eternity, mashing F12 at scene changes hoping to get some decent images very very close to each other for both. Then as with previous 720/1080 comparisons I've done, nearest neighbor scaled them up to 4K as the least common multiple. Some I split in half down the middle, some I cropped the same half and put side-by-side. I think the latter turned out much better. For both object edges and textures, I think 360->1080 is a winner over 360->720. FWIW, this is DLSS 2.2.11.0.
I believe Nintendo's new hardware equipped with Drake won't achieve 4 TF in terms of raw performance.DF droped a video talking about performance of 4tf consoles today and I find it interesting because 4tf is the romored performance of T239 in docked mode.
This video provides a good reference as what we can expect for the next switch.
Yeah, 4 TF is too much. PS4 level of graphical performance (which is already equivalent to a 750Ti!) would be enough. Memory bandwidth will be as much of an important metric to follow.I believe Nintendo's new hardware equipped with Drake won't achieve 4 TF in terms of raw performance.
Not sure if anyone posted this, the PS4 Pro vs Xbox Series S.
Kind of gives you an idea what could Nintendo's next gen hardware could be capable of. It didn't matter if the PS4 Pro had higher CU's or a faster clock speed, the Xbox Series S is capable at running games with better fidelity. It will be indeed interesting once we get our hands on with what Nintendo is cooking, and compare it to last gen and this gen
This should be a reminder to stop using GPU power as the only marker of system performance. That's why "7/8ths of a PS4" is always a frustrating way to talk about it to me.Kind of gives you an idea what could Nintendo's next gen hardware could be capable of. It didn't matter if the PS4 Pro had higher CU's or a faster clock speed, the Xbox Series S is capable at running games with better fidelity. It will be indeed interesting once we get our hands on with what Nintendo is cooking, and compare it to last gen and this gen
Don't kill my dream of a limited edition Zelda Switch. That is the one limited edition Switch that needs to happen!!!!!!!!!!* Hidden text: cannot be quoted. *
I think this is very very supported by the data we have, certainly.Disregarding teraflops, I anticipate Drake at a minimum will hit "PS4 Pro"-like visuals after DLSS, with DLSS doing a better job than checkerboard rendering.
I think Serif's look at it is pretty accurate. PS4 Pro will be old, but it was GIGANTIC. Being able to get PS4 level perf squeezed into a tablet is as much a leap as anything else. DLSS, however, could take it up to PS4 Pro-esque experiences. Different paths to a similar end.PS4 Pro will be very old tech by the time the next Nintendo console launches. I think as with anything, if the new device has >8GB of RAM and it most certainly will have a more mdoern architecture and features, it will outpace PS4 Pro performance.
It's actually pretty close to 1:1, but only when AMD is running the rest of the system optimally.Agree even docked, it's unlikely to hit for 4 nvidia Tflops, but I'm not as up to date on the covnersion with amd and nvidia flops, but even if its 1-1,
DLSS means that Drake just needs to modestly outperform the original PS4, to do PS4 Pro level stuff. That's a target it will very likely reach.i suspect the more modern featuresets will more than make up for the shortfall.
AMD mentioned that "The new RDNA [1] architecture is optimized for efficiency and programmability while offering backwards compatibility with the GCN architecture."How exactly do PS5/XBS achieve BC?
here's PS4PI can't find a geekbench 5 result for the PS4 Pro anywhere, only geekbench 4 results, which aren't comparable to 5 values.
@ILikeFeet @ReddDreadtheLead @oldpuck @LiC
correct me if I'm wrong in my line of thought here but:
If I'm not mistaken, drake would perform similarly in multicore compared to the steam deck but lose considerably in single core perf. under geekbench 5.
My point of reference is this benchmark for the AGX orin which I know, is a 12 core orin device instead of the 8 on drake and running on a 2GHz clock even. But afaik, clock speed doesn't scale linearly (far from it) on geekbench however, core count on multicore scores kinda does (again, afaik).
So I'm assuming at worse a ~33.3% reduction in multicore from the AGX Orin score(~4000ish points end result) followed by another drop (related to the clock speed drake would run at) which would put it around the same ballpark as the steam deck's CPU score (again, in multicore).
But considering this is what the current switch performs like on geekbench 5, and the arguments made ITT in the past, I expect at the very least, a 3x increase on that single core score under drake.
But going back to the whole PS4 Pro vs drake conversation (CPU-wise), I still think drake wouldn't lose by much versus the PS4 Pro's CPU.
The GPU talk is kinda pointless when it comes down to TFLOPs considering drake has half the bandwidth to work with compared to the PS4 Pro and series S.
Honestly, I think the reason Sony and MS moved up to $70 for new games is because they discount their games fairly early alongside 3rd-parties. With Nintendo, it takes far longer, if ever, for them to discount their games outside of general holiday sales and whatnot. Barely any permanent price drops, and they are making bank on that. They could stick to the $60 price tag for Switch 2 games, and they could likely still earn more profit than the others that start off high that then drop off later.What happens to Nintendo first party software prices?
My take:
Switch 2 is back compat and charges $70 for Switch 2 games. Switch 1 games stick at $60 or go down to $40(or lower).
The shader situation is well understood, and the Lapsus$ hack heavily suggests that they are both right about the shader situation.It’s been said by some that we may be missing parts of this whole Drake puzzle. Who’s to say that ScriesM and/or MVG aren’t also forgetting some small aspects regarding BC or shaders. I mean, these are singular persons making valid claims, but that doesn’t compare to the entire T239 team at Nvidia.
This would be a book. Seriously, this is a get your PhD in compsci level question. Those systems use multiple BC technologies. But the one you probably care about is how are they compatible with PS4 shaders, and the answer is that they didn't break backwards compatibility when they designed the newer GPUs. They kept the hardware from the PS4 era to execute old shader instructions that modern games don't need anymore.I’m going to ask this because I feel like it’s never been truly answered. How exactly do PS5/XBS achieve BC?
Rosetta 2 is a binary translator which monkey patches API calls. Here are two key points when talking about PC backwards compat:How specifically does Rosetta 2 and Apple’s other translation methods work on Apple silicon Mac systems?
Rosetta runs on the CPU. It reads the code from an old program, dynamically retranslating as it goes, and then feeds the retranslated instructions back out to the CPU. In essence, the CPU doesn't even see the emulated program, it only sees Rosetta. Meanwhile, Rosetta only has to read tiny chunks of the emulated program at a time, very quickly recompile it, and send it on in tiny, fast bursts. Lots of CPU code is repetitive, going around in loops, so many times, Rosetta doesn't even have to do anything, because it has already recompiled the code.What, if anything, would prevent such methods or something else being used to emulate the Switch GPU?
Sorta? But it doesn't matter much, only the shaders are a tricky problem.And regarding those Maxwell GPU snags, what is the current level of BC of the Maxwell GPU in relation to Ampere? I’ve heard the discussion on the pre-compiled shaders, of course, but what about the rest of the GPU? Is the entire Switch GPU incompatible with Drake’s in every facet?
No, but again, only the shaders are tricky.And for emulation/translation: is it feasible for Drake to do so solely for the pre-compiled shaders “portion”? Yes or no?
You can't really partially emulate anything the way you mean, but you can essentially "pass through" the parts the are identical. Your emulator always has to sit in the middle of the whole thing, and you pay a performance cost for that. But when two systems are very very similar, your emulator only has to work hard at the differences.And if not, why not? Are they forced to emulate the entire TX1 GPU? Is partial GPU emulation a preexisting concept? Has it been attempted before?
SM3DAS could do all these recompilation steps in advance, on a developer's powerful workstation. Drake has to do it in real time, using nothing but a fancy tablet.And Nintendo has already done GPU emulation with SM3DAS, so what’s the difference with Switch?
What's up with all these Switch 2 videos on the same day from the Spawncast?
Not really? Because their contention isn't merely that a new GPU architecture can't natively/directly run the shaders in current Switch games -- which is true, and not something you need the leak to tell you -- it's that the solutions to this problem are intractable or that Nintendo won't pursue them for some reason. That's the stupidity of the arguments that they make (which the leak has no bearing on).The shader situation is well understood, and the Lapsus$ hack heavily suggests that they are both right about the shader situation.
I think people are too obsessed with 100% backwards compatibility, hence the argument that Nintendo needs to have the Tegra X1 installed on the DLSS model*'s motherboard for backwards compatibility, which I think is a ridiculous argument to begin with.Not really? Because their contention isn't merely that a new GPU architecture can't natively/directly run the shaders in current Switch games -- which is true, and not something you need the leak to tell you -- it's that the solutions to this problem are intractable or that Nintendo won't pursue them for some reason. That's the stupidity of the arguments that they make (which the leak has no bearing on).