RetroTink 4K All But Officially Announced ....

[ZellSF]

AMD also has a generic solution, that works in all DX11/12 games, not needing data from the game engine as Guspaz warns about, so AMD thinks it's viable for (3D) gaming. I don't have an AMD card so I can't say if I agree or not.

That said, there's a difference between what a expensive GPU and a FPGA can do and there is a difference between AMD and Mike Chi's research and development budgets.

[VEGETA]

AMD also has a generic solution, that works in all DX11/12 games, not needing data from the game engine as Guspaz warns about, so AMD thinks it's viable for (3D) gaming. I don't have an AMD card so I can't say if I agree or not.

That said, there's a difference between what a expensive GPU and a FPGA can do and there is a difference between AMD and Mike Chi's research and development budgets.

well, i previously wrote about changing frame rate while not using BFI. looks like this discussion here is the same.

i believe some scaler ICs (not generic fpgas) support such features, but maybe someone here can shed a light on this. gbs-c chip supports this but very limited.

[Josh128]

It should be cautioned that FSR3 and DLSS framegen are quite different than the motion interpolation done by TVs. Those algorithms work with a significant amount of supplemental information from the game engine that allow them to better predict the intermediate frames. This includes things like per-pixel motion vectors (the game engine knows the location and movement of every object anyway), the depth buffer, and an exclusion mask (to handle things like crosshairs and HUD elements). These basically serve as hints for the GPU's hardware optical flow engine, since things like motion vectors don't cover everything (like particles or reflections). All your TV has to work with is the current and previous one or more video frames, it's missing all those hints.

Thats why I mentioned both FSR 3 and Fluid Motion Frames as separate tech. FMF is different from FSR3,and as ZellSF mentioned, works at the GPU driver level , so it works with any game and is agnostic to most if not all of the supplemental game engine information used by DLSS3 and FSR3-- it works in a similar way to the motion interpolation in TVs.

In any case comparing modern 4K games that use 24 or 32 bit color depth with 240p games using 8 or 9 bit color depth is different. Im not sure how anyone here can declare with authority that it would look like crap. And yes, it would only be viable for 60fps games, for sure. Games that run at 10, 15, 20, or 30fps would not benefit from it, as they would work perfectly in standard vanilla 240p120 mode with no discernable difference from 480p60 + scanlines other than brightness and color pop. They already exhibit double strobe effect on a 60Hz CRT and that doesnt change with going to 120Hz.

As far as Mikes software development resources vs AMDs, of course, I know there is a vast difference!! The reason Ive been putting this out there is because Marqs himself was the first one to mention it in the OSSC Pro thread years ago. He spoke about it as if it would be something that could actually be implemented on his device. As far as the nuts and bolts of implementing it, maybe its too much work, I dont know. The thing is, once this is out there-- Mike will have time to explore niche functionality like this and if he chooses, he can implement it and thus add even more value to this expensive piece of equipment to entice those who might be on the fence about it. This is not like the 5X where he had to rob Peter to pay Paul due to lack of hardware resources.

If its proven that it doesnt work, fine, dont do it, but until then, I for one certainly wont be trying to shoot down novel ideas no matter how niche, that could add value to a product like this.

[fernan1234]

Im not sure how anyone here can declare with authority that it would look like crap.

That was me. No authority but to me it seems obvious that with all the 2D patterns and solid colors that make up so many frames of 60fps retro games it's inevitable that motion interpolation will lead to tons of very visible artifacts as it fails to mesh up frames on top of the solid colors and patterns of new frames, unlike the case of more detailed 3D graphics and video images where it becomes easier for artifacts to go unnoticed within more detailed pictures. Would be nice to be wrong about this just to have more non-CRT options though.

[orange808]

I would also like to see a high quality comb filter implementation in the future as well. The only genuinely good comb filter I own has far too much lag. That would be a natural companion for the composite input on the machine.

[Josh128]

I would also like to see a high quality comb filter implementation in the future as well. The only genuinely good comb filter I own has far too much lag. That would be a natural companion for the composite input on the machine.

+1. These are the types of things that can separate a Lamborghini from a Camaro.

[Guspaz]

AMD's Fluid Motion Frames technique does work on any game, but without all the game engine input, doesn't work as well as FSR3 or DLSS framegen. It particularly has issues with static elements due to the lack of the exclusion map: things like crosshairs jitter around like mad. That said, if given a sufficiently high input framerate (like doing 60 -> 120), there's less changes between each frame, resulting in less artifacts, and it's also harder for the human eye to notice artifacting, since the artifacts don't persist for long.

Keep in mind that all motion interpolation techniques (excluding the motion extrapolation techniques that BlurBusters has discussed) add a bunch of additional latency since they can't start processing an interpolated frame until the next frame has been fully received.

[orange808]

We're discussing machine learning now, right? That doesn't seem fair.

I seriously doubt the Retrotink 4k could be equipped to handle ML.

I don't even know how you would go about something like that on FPGA. I guess you'd deploy a series of RISC-V soft cores, but that would cost a lot of gates--and how would you keep them all fed? Where's your low latency RAM going to come from at a reasonable price?

Ultimately, it's still cheaper to physically cram static cores and caches on a physical board, right? And, you need bleeding edge performance to get the frames out in real time, right?

So, unless the Retrotink 4k has a GPU expansion coming (and the bandwidth to handle it), I'm not expecting any GPU ML video technology. We don't need that for an interpolation implementation, but I'm not expecting GPU ML processing.

[Josh128]

From Marqs almost 4 years ago:

Doubling the refresh rate induces both motion incoherency (2 strobes/scans per frame) and latency (half a frame at minimum) so it's not without tradeoffs. You can do motion interpolation to mitigate the former (remember the 100Hz CRT TVs) but that bumps the lag even more. That said, 240p 120+Hz output should be doable on the Pro model.

High power GPUs need not apply. If it is technically possible on the OSSC Pro, should be on the 4K as well.

[Guspaz]

From Marqs almost 4 years ago:

Doubling the refresh rate induces both motion incoherency (2 strobes/scans per frame) and latency (half a frame at minimum) so it's not without tradeoffs. You can do motion interpolation to mitigate the former (remember the 100Hz CRT TVs) but that bumps the lag even more. That said, 240p 120+Hz output should be doable on the Pro model.

High power GPUs need not apply. If it is technically possible on the OSSC Pro, should be on the 4K as well.

That is a statement that 240p120 output is possible on the OSSC Pro, not motion interpolation.

You can do motion interpolation without ML, some googling turned up some papers about doing it on what appear to be relatively modest FPGAs, but that doesn't guarantee that it'd be possible on retro scalers: the papers looked very memory bandwidth intensive.

[Josh128]

Perhaps you are right, but I find it strange that he would mention it as well as additional latency it would add if it wasnt possible. It'd be awesome if one of the new scalers could do it at some point.

[Hoagtech]

Does anyone know if the Tink 4k can output through Dsub VGA?

The 5x was capable of output through SCART and component when downscaling.

It would be nice to adapt analog inputs to VGA for PC and Pro CRT monitors to avoid the hit and miss DAC options the current market offers.

[SavagePencil]

The 5x was capable of output through SCART and component when downscaling.

I'm pretty sure the 5X outputs its downscaling through the HDMI and it's an exercise for the user to provide a DAC: https://consolemods.org/wiki/AV:RetroTI ... ownscaling

[Guspaz]

Does anyone know if the Tink 4k can output through Dsub VGA?

The 5x was capable of output through SCART and component when downscaling.

It would be nice to adapt analog inputs to VGA for PC and Pro CRT monitors to avoid the hit and miss DAC options the current market offers.

Neither the RT4K nor the RT5X supports analog output of any kind. Not sure where you got the idea that the RT5X does, it doesn't. If you want analog output from an RT5X/RT4K, you need to use a DAC.

[Hoagtech]

Does anyone know if the Tink 4k can output through Dsub VGA?

The 5x was capable of output through SCART and component when downscaling.

It would be nice to adapt analog inputs to VGA for PC and Pro CRT monitors to avoid the hit and miss DAC options the current market offers.

Neither the RT4K nor the RT5X supports analog output of any kind. Not sure where you got the idea that the RT5X does, it doesn't. If you want analog output from an RT5X/RT4K, you need to use a DAC.

Weird. My memory must have played a trick on me.

This DAC has been a moving target for CRT owners. Cross my fingers someone produces one in this community. Id buy Mikes in a second if he ever decides to.

I don’t use LCD for the majority. My OLED is basically my kids netflix and Switch box.

[tongshadow]

So this is the power of Marketing...
https://twitter.com/WobblingP/status/17 ... 3592652960
Seriously though, I was expecting a smaller gap between the Tink and Pro.

[SavagePencil]

So this is the power of Marketing...
https://twitter.com/WobblingP/status/17 ... 3592652960
Seriously though, I was expecting a smaller gap between the Tink and Pro.

What’s your point by posting an internet poll that only X users can vote on?

[tongshadow]

So this is the power of Marketing...
https://twitter.com/WobblingP/status/17 ... 3592652960
Seriously though, I was expecting a smaller gap between the Tink and Pro.

What’s your point by posting an internet poll that only X users can vote on?

What do you mean by X users? It's a public platform. It would be a different story if it was a poll from some discord server or closed group.

[fernan1234]

It's not just marketing though, it's actually showing what your product can do.

[SavagePencil]

So this is the power of Marketing...
https://twitter.com/WobblingP/status/17 ... 3592652960
Seriously though, I was expecting a smaller gap between the Tink and Pro.

What’s your point by posting an internet poll that only X users can vote on?

What do you mean by X users? It's a public platform. It would be a different story if it was a poll from some discord server or closed group.

You have to have an X account to vote. It’s irrelevant because it doesn’t actually affect the conversation.

[tongshadow]

What’s your point by posting an internet poll that only X users can vote on?

What do you mean by X users? It's a public platform. It would be a different story if it was a poll from some discord server or closed group.

You have to have an X account to vote. It’s irrelevant because it doesn’t actually affect the conversation.

Twitter is too popular, you'd have a harder time finding someone who DOESNT have an account than the opposite. Furthermore, WobblingPixels is a retrogaming account, so the votes came from users who at least know about such devices. The sample size is fairly considerable too, nearly 2500 voters.
If your reasoning for the Tink4K's dominance is because the Morph and OSSC Pro voters didnt have a Twitter account, that's a very shaky foundation to stand on.

[SGGG2]

As Kanye West once said, “I do NOT have a Twitter!”, so I’m unable to see the results — but it’s important to note this is a poll of engaged retro technology enthusiasts. As RetroRGB Bob often points out, there are wayyyyy more casuals out there. To add to that, Morph hasn’t really had a showing yet…

[spmbx]

I can't ser the results, which is fine because i don't see what the use of such a poll is anyway. other than more drama. Of which the scaler crowd has too much already.

[tongshadow]

i don't see what the use of such a poll is anyway. other than more drama. Of which the scaler crowd has too much already.

Not my intention, we're all grown ups here after all. I just want to discuss why the Tink4K had overwhelming numbers compared to the others. In my opinion, many are unaware of the potential from the other products, and that's partially due to the Tink4K's presence.

[marus]

I just want to discuss why the Tink4K had overwhelming numbers compared to the others. In my opinion, many are unaware of the potential from the other products, and that's partially due to the Tink4K's presence.

I think you hit the nail on the head there. There's been enough footage and feedback on the Retrotink4k at this point that any potential customer can understand all of its features and be reasonably confident that they'll have a solid experience with it. I'm not personally in the market for one as it's just too expensive and I don't really care about 4k scaling. However, considering how much the retro gaming market has relied on FOMO and hype to sell stuff lately, it's been refreshing to see someone market their product so transparently and explain clearly who it is and isn't for.

I could potentially see myself picking up a digital-only Morph or the OSSC Pro in the future, as I have a use for a scalar with digital inputs and they're both more in my price range. In fact, the Morph at $350 right now is tempting, but I just don't feel comfortable pre-ordering one before other people have had a chance to test it and iron out the kinks. Based on all of PixelFX's other products I'm sure it'll turn out fine, but the fact is I haven't actually seen one running yet. I wouldn't be surprised if there's a lot of other people taking a wait-and-see approach as well.

[SavagePencil]

If your reasoning for the Tink4K's dominance is because the Morph and OSSC Pro voters didnt have a Twitter account, that's a very shaky foundation to stand on.

Who said that? It’s irrelevant and you’re trying to turn this into a scaler war. This thread is to talk about the RT4K and you’ve shit up all these threads, including dismissing people who might want S-Video/composite.

[Josh128]

i don't see what the use of such a poll is anyway. other than more drama. Of which the scaler crowd has too much already.

Not my intention, we're all grown ups here after all. I just want to discuss why the Tink4K had overwhelming numbers compared to the others. In my opinion, many are unaware of the potential from the other products, and that's partially due to the Tink4K's presence.

Being brutally honest, its easy to see understand why the 4K has overwhelming numbers, even with its very high price-- Mike completely raised the bar with the 5X with features and FW updates, has distributed something like 30+ 4K units to the community for bug testing and recommendations, and its now known that the hardware Mike has chosen is on another level from the other 2 scalers. People want "the best", even if it really isnt the best bang for the buck for their use case. With Mikes software + the ridiculous hardware, theres little doubt the 4K is going to be hands down the best unit at launch, and the sky is the limit for updates.

Thats not to say the other units arent/wont be great - but they have a very tall order to match the polish and love Mike has poured into the 4K.

[tongshadow]

If your reasoning for the Tink4K's dominance is because the Morph and OSSC Pro voters didnt have a Twitter account, that's a very shaky foundation to stand on.

Who said that? It’s irrelevant and you’re trying to turn this into a scaler war. This thread is to talk about the RT4K and you’ve shit up all these threads, including dismissing people who might want S-Video/composite.

This is a discussion forum, you're free to disagree and present your arguments. If you dont like seeing opposing opinions, just go to a Discord echo-chamber.

[SavagePencil]

If your reasoning for the Tink4K's dominance is because the Morph and OSSC Pro voters didnt have a Twitter account, that's a very shaky foundation to stand on.

Who said that? It’s irrelevant and you’re trying to turn this into a scaler war. This thread is to talk about the RT4K and you’ve shit up all these threads, including dismissing people who might want S-Video/composite.

This is a discussion forum, you're free to disagree and present your arguments. If you dont like seeing opposing opinions, just go to a Discord echo-chamber.

It is a discussion forum. And your posts do not foster discussion. You posted a worthless poll and insinuated that the RT4K was somehow unfairly manipulating the market through nebulous “marketing.” Now I’m a hypocrite for calling out how useless that was.

[bobrocks95]

Who said that? It’s irrelevant and you’re trying to turn this into a scaler war. This thread is to talk about the RT4K and you’ve shit up all these threads, including dismissing people who might want S-Video/composite.

This is a discussion forum, you're free to disagree and present your arguments. If you dont like seeing opposing opinions, just go to a Discord echo-chamber.

It is a discussion forum. And your posts do not foster discussion. You posted a worthless poll and insinuated that the RT4K was somehow unfairly manipulating the market through nebulous “marketing.” Now I’m a hypocrite for calling out how useless that was.

You're overreacting, the dude was not trying to insinuate anything about the RT4K. Having early units out in the wild for people to test and show off IS a form of marketing, on top of beta testing. Most of the scaler war vitriol garbage I have seen is people *thinking* someone is "trashing" one of the products and blowing everything said out of proportion and turning it into a thing when it's really not.

and its now known that the hardware Mike has chosen is on another level from the other 2 scalers.

Apparently comparing "Logic Elements" or "Logic Cells" across brands gets immediately tricky since they all build and measure things differently: https://www.reddit.com/r/FPGA/comments/ ... ification/

Woozle posits they're more similar than they may appear on paper (and yes I fully understand it's in his interest to do so, but if you already think the guy is a blatant liar or idiot or something for whatever reason, it's not like you're considering buying a Morph anyways)

https://imgur.com/WIRLIxa
https://imgur.com/CjPQ3zr
https://cdn.discordapp.com/attachments/ ... da95aba49&
https://cdn.discordapp.com/attachments/ ... 25ed6cd62&

Is the Cyclone hardware better? Probably, and Altera's compiler is probably more mature than Trion's. Is it ~34% better because Altera measures it at 150K logic elements? Probably not, but I'm no FPGA engineer. Doesn't seem as cut and dry as you might think though.


In the broader scheme of things I don't think PixelFX or Mike will run into FPGA space issues. Mike said the 4K's chip was what, 20x bigger than the Tink 5X? And that thing kept getting features he thought would never fit on it. Firmware and pricing will win the day here, Mike is ahead on the former and PixelFX is ahead on the latter.

Like you said Josh some people just want the best and will get the most expensive option they can- I am hoping most people will be civil about things and decide for themselves what feature-set vs price balance is right for them since I think they can both fit nicely in their pricing segments.