Linus tech tips on dashcams.

[Vortex Radar]

Regarding 30 / 60 FPS recording "VR" neglect to mention that if you go with 60 FPS, your bitrate are used to process 30 frames, so watered out VS 30 FPS recording.

Yeah I felt like that would open up a whole 'nother can of worms and I didn't wanna go down that path, lol.

The A119 v3, for example, seems to have better image quality at 30 fps in general. It records at a higher bitrate at 30fps than 60fps, plus all that bitrate goes to half the number of frames, so overall quality is better with 30... BUT sometimes 60 fps can do a better job at freezing plates, specifically, due to the minimum shutter speed bumping up from 1/30 to 1/60 and thus you'd get less motion blur.

I left it kinda nebulous in the video since 30 vs 60 can vary depending on how each dashcam implements things and so the answer is like "well it depends..."

 

[Screen name taken]

Another 2022 review video. This forum even gets a short mention.

 

[rgbdct]

Linus tech tips on dashcams.

dashcamtalk.com

 

[Lothar]

Other than Starvis 2 coming up, any updates coming to Novatek or anything major?

Seems like you have a lot of trash/fake dashcams and then the more reputable brands, but they're just adding feature after feature while still using the same parts for what seems like the last 2 - 4 years.

Novatek is a very secretive company.
There won't be anything ground breaking...just slight incremental improvements (increase in processor speed mhz, amount of RAM, # of cores which will allow an increase in the number of channels, more AI stuff which allows for more ADAS crap, etc...)

You will be waiting a long while for anything ground breaking (ex: AV1 codec, etc...).
After examining the state of things elsewhere (Phone, TV, etc...), I'm not optimistic that AV1 is "just around the corner".
I don't have my hopes up, but I would probably guess conservatively either 2025 or later before you see AV1. If it can coincide with my dreams of a 2K "Starvis 2" (hopefully front and rear) dashcam from a quality manufacturer in 2024, then great!

 

[Nigel]

Novatek is a very secretive company.
There won't be anything ground breaking...just slight incremental improvements (increase in processor speed mhz, amount of RAM, # of cores which will allow an increase in the number of channels, more AI stuff which allows for more ADAS crap, etc...)

You will be waiting a long while for anything ground breaking (ex: AV1 codec, etc...).
After examining the state of things elsewhere (Phone, TV, etc...), I'm not optimistic that AV1 is "just around the corner".
I don't have my hopes up, but I would probably guess conservatively either 2025 or later before you see AV1. If it can coincide with my dreams of a 2K "Starvis 2" (hopefully front and rear) dashcam from a quality manufacturer in 2024, then great!

Too pessimistic!

Starvis 2 will arrive in our dashcams in 2023 from leading manufacturers.
The first sample sensors arrived a while ago and are currently being tested, could be on sale this year, but unfortunately are not the ones we are waiting for.

Don't know about AV1, it is already in some fairly low cost devices, so it would be fairly easy for Novatek to add. Not sure they have a good reason to try at the moment though, and we want a decent version, not a low quality version designed for video calls, and although licensing of the codec is free, implementation will not be, so there may be licensing costs for the implementation they use, which may delay its arrival in dashcams until it is seen as an essential.

There is something significant included in Novatek's next processor, but like you say, they are very secretive, so what it is is not public knowledge!

 

[Lothar]

Too pessimistic!

Starvis 2 will arrive in our dashcams in 2023 from leading manufacturers.
The first sample sensors arrived a while ago and are currently being tested, could be on sale this year, but unfortunately are not the ones we are waiting for.

Don't know about AV1, it is already in some fairly low cost devices, so it would be fairly easy for Novatek to add. Not sure they have a good reason to try at the moment though, and we want a decent version, not a low quality version designed for video calls, and although licensing of the codec is free, implementation will not be, so there may be licensing costs for the implementation they use, which may delay its arrival in dashcams until it is seen as an essential.

There is something significant included in Novatek's next processor, but like you say, they are very secretive, so what it is is not public knowledge!

My estimate is based on an actual physical launch of a 2K Starvis 2 sensor dashcam (preferrably front & rear) and not when it arrives in the hand of beta testers or pre-release cameras.
Yes, I definitely am not looking for a 1080P Starvis 2 sensor. 2K is what I'm looking at.


Reason that I'm a bit pessimistic about AV1 in dashcams is...

In addition to AV1 decode, wouldn't a dashcam need AV1 encode?
Yes...There are a lot of TV's and a few mobile processors (Google Tensor released in 2021, and Qualcomm's Snapdragon Gen 2 to be released in 2023) that have built-in AV1 decode capability(meaning you can watch AV1 encoded videos), but none of those devices have AV1 encode capability (meaning you cannot record AV1 videos or in AV1 format).

The essential element of dashcams is recording videos, not just playing videos so therefore a dashcam would need AV1 encode capability.

Once you look at this chart, I don't think I am being pessimistic at all...Only 3 companies have fixed-function hardware AV1 encode (Allegro, Chips&Media, and Intel). I'm not including Dwango because who wants to be using a 720P dashcam in this age

AV1 - Wikipedia

en.wikipedia.org

When I see the likes of Apple, Google, Qualcomm, Samsung, and others release a mobile processor that has fixed-function hardware encode, then I will be more optimistic that AV1 is around the corner for dashcams.
Since Novatek is very secretive, I can only make an estimate based on where their "indirect" competitors (Apple, Google, Qualcomm, Samsung) are in other places (Phone, TV) that don't relate to dashcams.

Sometimes it is better to be pessimistic than optimistic in case what you are planning or believing does not work out.

 

[Nigel]

The essential element of dashcams is recording videos, not just playing videos so therefore a dashcam would need AV1 encode capability.

Absolutely, no point in having a dashcam that can't encode 🤣

Chips&Media say:

The new encoder IP, WAVE627, is perfectly suitable for edge servers (e.g., video transcoding, video analytics, cloud data center) and consumer applications in the smartphone, tablet, action camera, automotive, surveillance cameras, AR/VR, and much more.

Link

I don't expect that is the one Novatek will use, but it is targeting SoC suppliers like Novatek. Most of the low level processor design companies have been working on AV1 since the base spec was released several years ago. They are all reaching the point where it can be included in things like action camera/dashcam SoCs. at reasonable cost, and I suspect there is enough competition to keep it reasonably cheap. Novatek just have to buy/license the block from whichever supplier they choose, to include on their silicon. Why would they delay, and maybe loose some customers or just customer license purchases for use of the block?

It will happen soon, don't know how soon though, and I don't want to guess, it has already taken longer than I expected!

 

[Lothar]

Absolutely, no point in having a dashcam that can't encode 🤣

Chips&Media say:


Link

I don't expect that is the one Novatek will use, but it is targeting SoC suppliers like Novatek. Most of the low level processor design companies have been working on AV1 since the base spec was released several years ago. They are all reaching the point where it can be included in things like action camera/dashcam SoCs. at reasonable cost, and I suspect there is enough competition to keep it reasonably cheap. Novatek just have to buy/license the block from whichever supplier they choose, to include on their silicon. Why would they delay, and maybe loose some customers or just customer license purchases for use of the block?

It will happen soon, don't know how soon though, and I don't want to guess, it has already taken longer than I expected!

Yes, but only 2(Allegro DVT, and Chips&Media) so far has emerged. Intel doesn't count since theirs is in their discrete GPU and Intel does not license it's IP to competitors (except AMD which they have an X86 legal agreement going back over 30 years ago).
Two companies alone are not enough to make it cheap.

Maybe they want to develop their own block rather than license? Yes, doing it this way costs more initially but it may save them more money in the long run if they are going to include this on all their future SOCs?
Maybe they are not in any rush to do it because their competitor (HiSilicon, and Ambarella) has not done so either?

 

[Nigel]

Maybe they want to develop their own block rather than license?

I think that is less likely than Viofo developing their own SoC!

I'm sure there are others working on the low level stuff. ARM was in from the start, haven't heard from them yet, but then we probably wouldn't unless someone like Novatek announces that they are using ARM codecs, like they have with processing cores...

 

[RedSeal]

The outtake from Linus's video is Dashcam video quality is very average, and massive improvements need to be found and made.

These developers like Novatek, need to bring some major changes to the SoC's.

The market could bare a considerable price hike if the dashcams could do their primary function properly.

The incessant add-on's, functions, and features are great; but we need the primary function to be greatly improved.

 

[Google]

I despise this monetized sell-out, Linus Sebastian. If he told me water was water was wet, I would remain skeptical.

Moreover, the last place anyone should expect to credible information and facts is from YouTube or TikTok type sites. This is where conspiracy theorists and right wingers get their information from.

Linus (like Trump and Joe Rogan) have become very popular and is a testament in my opinion on how society as a whole is regressing.

 

[Google]

It made a good story, and that is what he video is, not a technical or scientific investigation/tutorial/review of dashcam technology. If people had to think and understand what he is saying in all his videos then he wouldn't be getting the 2 million views in 2 days!

He had a point to make, and he made it well, don't take the rest too seriously.

LS's point was to lure and keep viewers glued, and to subscribe or "like" first and foremost. That's how YT works and you don't need me to tell you something that you already know. However, gaslighting doesn't change this fact and that LS is a wealthy scumbag.

 

[Google]

The outtake from Linus's video is Dashcam video quality is very average, and massive improvements need to be found and made.

These developers like Novatek, need to bring some major changes to the SoC's.

The market could bare a considerable price hike if the dashcams could do their primary function properly.

The incessant add-on's, functions, and features are great; but we need the primary function to be greatly improved.

That's your perspective. Mine is that he expects 4K, 2160p or higher resolution to capture a license plate.

The market is already suffering from unprecedented inflation, greed, and gouging by corporations. So, hell no. The market could NOT bare a considerable price hike unless the market is wealthy, privileged, or elite like you.

While the tech behind dashcams hasn't changed much over the years (and it's lagging) depending where you are, LE won't usually take dashcam footage because they want their own officer to witness the event. In contrast, more often (then not) vehicle insurance companies do welcome dashcam footage but they sometimes go so far to dismiss the footage when their intent is blame both parties.

What we need is automakers to provide their own dashcam as standard equipment.

 

[Nigel]

LS's point was to lure and keep viewers glued, and to subscribe or "like" first and foremost. That's how YT works and you don't need me to tell you something that you already know.

That is a fairly major aim of all entertainment, and clearly Linus is providing entertainment, it may also be somewhat educational/informative, but it is primarily entertainment. If your videos don't provide entertainment then very few people will watch them, go to the other extreme and you get TopGear, originally an education/information program, but it chased viewer figures to the extreme! "The series made it into the Guinness World Record 2013 Edition book for being the world's most-watched factual TV program." YouTube isn't about entertainment, it is just a service, they simply want people to use their service and make YT a lot of money in the process.

However, gaslighting doesn't change this fact and that LS is a wealthy scumbag.

I don't think that is fair, nothing wrong with using Youtube to obtain an income, and the main message was a reasonable one to make, not entirely accurate but then no video sources provide 100% accurate information. I do think that he should have tested his Gopros before claiming that they make better dashcams than dashcams do, but a lot of people make the same mistake, same with believing the latest iPhone will make a far superior dashcam, when actually it is inferior.

While the tech behind dashcams hasn't changed much over the years

That's not true, just a few years ago we were using 1080 resolution as the sweet spot, now nearly all dashcams have moved to 2K, most using the same sensor, because that is where the sweet spot is this year, next year the Sony IMX335 will lose that position and dashcams will move on to a new range of sensors, including 4K within the sweet spot. Just a few years we were using mainly 1 channel dashcams, now we have 3 channel mainstream dashcams. It is true that we are not seeing developments month by month, but we never have.

LE won't usually take dashcam footage because they want their own officer to witness the event.

That may be the case in your homeland, but here it is the other way around!

What we need is automakers to provide their own dashcam as standard equipment.

A bit like automakers should provide their own mobile phones as standard equipment?

Much better if they provided a standard powered dashcam mount, with video networking to other viewpoints included as standard.

 

[Screen name taken]

opps

 

[kamkar]

Well i certainly do not like big companies, no matter what their product are, or in the case of your user name what kind of product they make me.
In general i always think when it get big, it also get evil.
I will gladly admit i am a Apple hater, but it go back to the MP3 player days, not so much the Apple computers, and dont get me started on their phones.
Needless to say i would not buy a google phone, but there have been android on all my smart phones except my venture with a windows phone, and i am using windows even if i also have issues with Microsoft.
I am also still using youtube, but all other google "services" i am done with, and it is pretty much also just burner accounts i have.
I do also wish there was a way i could prevent people from following my channels, even if that was possible they could just bookmark my channel and "stalk" me that way.
I would of course never turn on " suck on the google tit" even if a channel of mine should get several 100 million stalkers / addicts.

 

[Lothar]

That's your perspective. Mine is that he expects 4K, 2160p or higher resolution to capture a license plate.

The market is already suffering from unprecedented inflation, greed, and gouging by corporations. So, hell no. The market could NOT bare a considerable price hike unless the market is wealthy, privileged, or elite like you.

While the tech behind dashcams hasn't changed much over the years (and it's lagging) depending where you are, LE won't usually take dashcam footage because they want their own officer to witness the event. In contrast, more often (then not) vehicle insurance companies do welcome dashcam footage but they sometimes go so far to dismiss the footage when their intent is blame both parties.

What we need is automakers to provide their own dashcam as standard equipment.

Similar to what BMW is doing now with their heated seats?

BMW starts selling heated seat subscriptions for $18 a month

Now being sold in South Korea, the UK, and elsewhere

www.theverge.com

BMW Bets the Fuss Over Its Heated-Seat Subscriptions Will Pass

The carmaker still has high hopes for selling digital services to drivers.

www.bloomberg.com

Most automakers already offer an integrated dashcam as a package.
The typical offer from the automakers is $350-450 for a 1080p FHD, front only dashcam with 8GB SDcard and no features (WiFi, low bitrate parking mode, buffered parking protection, etc...)
My 2CH camera costs much less than that, has a higher video quality(2K), and has more features...

Bad enough that they already overcharge for their inferior dashcam product. You want to open the microtransactions path for them too?
I fail to see how requiring automakers to provide their own dashcams as standard equipment will increase quality of the dashcam market, or lower costs for the consumers.

 

[DAP]

Subscription fees make sense only for things that require continuous support. For example the Tesla Premium Connectivity. Renting built in hardware features like heated seats is insane.

Many cars already have built in cameras. In the US, all new cars are required to have a back up camera. Many have an equivalent parking camera in the front. Some adaptive cruise control systems use a stereo pair of cameras on either side of the center rear view mirror. For these cars it would not take much to add a dashcam capability.

Tesla already does this. Any dash cam is better than no dash cam, but you will see from my profile that I find Tesla's dashcam inadequate, and have supplemented it with two Viofo 129 pro cameras. The shortcomings of Tesla's dashcams are as follows:
No sound recording. you can't read out license plates to cover for when the camera can't capture the plate.
Low resolution video and extreme compression. These cameras were designed for AI driving, not capturing license plates.
Low priority task in system. During an accident, the computers can get so busy that they run out of processing power and drop low priority tasks, so the critical seconds of video can be missed.
Unreliable. Tesla's computers can be picky about the memory they will talk to. Sometimes they will decide to stop talking to memory that was working fine in the past. Sometimes the car just fails to record for no apparent reason.
Still, I'd rather have the teslacam running than not.

I do think car manufacturers should build dashcams into their cars. I don't expect them to be good dash cams, but again any dashcam is better than no dashcam.

 

[Dashmellow]

Having to pay $18.00 a month for heated seats you ordered when you purchased your car sounds an awful lot like the Mafia practice of PIZZO. Pizzo (Italian: [ˈpittso]) is protection money paid to the Mafia usually in the form of a forced ongoing transfer of money resulting from extortion! 😎

 

[Nigel]

Having to pay $18.00 a month for heated seats you ordered when you purchased your car sounds an awful lot like the Mafia practice of PIZZO. Pizzo (Italian: [ˈpittso]) is protection money paid to the Mafia usually in the form of a forced ongoing transfer of money resulting from extortion! 😎

$18 should get you a reasonable quality seat heater, but not the installation.

Of course, if it is fitted during seat manufacture then $18 fitted is a lot more than manufacturing costs, decent profit for the manufacturer.

So what does the $18 per month pay for, do they supply the fuel?