NVidia and MediaTek to collaborate on ARM gaming laptops

I'd be more interested in seeing if they extend this to desktops also. If they are going to focus on laptops only, and ARM is the future of gaming... well I'm not looking forward to regressing to the point where a 17" monitor is considered large and you "upgrade" by tossing your entire computer (and monitor) in the trash every few years, buying a new one instead. That's laptop gaming.
 
I'd be more interested in seeing if they extend this to desktops also. If they are going to focus on laptops only, and ARM is the future of gaming... well I'm not looking forward to regressing to the point where a 17" monitor is considered large and you "upgrade" by tossing your entire computer (and monitor) in the trash every few years, buying a new one instead. That's laptop gaming.
Normally I would agree, but the laptop market is bigger, and the low power envelope of mobile plays to all of ARM's strengths and can take advantage of all of x86's weaknesses. It will result in a product that seems much better in comparison, and that would be far harder to do in a Desktop space where the higher power envelopes give x86 a larger advantage. There is also another side of this, by doing this NVidia is essentially developing a console, and by working with one of the major existing ARM providers they are displaying that they aren't rocking the boat and they don't intend to make big changes to the existing ARM models which can help with their arguments on why owning ARM isn't anti-competitive. Starting with a mobile spec also greatly simplifies the design process, there are expectations from a Desktop about future upgradability and part compatibility, and reliance on existing standards, none of those really exist in the mobile space. They decide that an 8v setup is most efficient. they can do that, wherein a desktop they would have to stick to the existing ATX standards, PCIE specifications, and blah blah blah blah, in Mobile they can design it from the ground up as proprietary as they need to to achieve maximum results.
 
and by working with one of the major existing ARM providers they are displaying that they aren't rocking the boat
Eeeeeeeeeeh. Yea that's probably gonna be seen differently by judges. Aren't they making their own server silicon now as well?
 
Eeeeeeeeeeh. Yea that's probably gonna be seen differently by judges. Aren't they making their own server silicon now as well?
Yes but through their purchased ARM license, NVidia has whole existing product stacks of ARM CPU's that are already for sale and installed in automotive fleets and heavy machinery around the world they have been selling their own ARM chips for almost 10 years.
 
Your only hearing about this because someone leaked.

Nvidia is working with everyone not just Mediatek. :) And they are also not just coming for laptops.

Expect nForce chipset powered motherboard announcements before the end of the year. When Nvidia announces socketed versions of grace... then the fight will be on for real. Jensen has been waiting to do a leather jacket wearing happy dance on a stage announcing Intels death for a decade now.
 
Your only hearing about this because someone leaked.

Nvidia is working with everyone not just Mediatek. :) And they are also not just coming for laptops.

Expect nForce chipset powered motherboard announcements before the end of the year. When Nvidia announces socketed versions of grace... then the fight will be on for real. Jensen has been waiting to do a leather jacket wearing happy dance on a stage announcing Intels death for a decade now.
I don’t think we’re going to see sockets, the slot setups they are using for their dev kits work awesome and let’s them do complete SOC. But it’s all speculation so I’m probably wrong but a slot design plays better with others and NVIdia needs 3’rd party players if they are going to pull this off.
 
  • Like
Reactions: ChadD
like this
Eeeeeeeeeeh. Yea that's probably gonna be seen differently by judges. Aren't they making their own server silicon now as well?
Nvidia is one of the handful of companies that have a ARM architecture licence. Qualcomm, Apple, Samsung, ARM, AMD, Fujitsu and a few others. Mediatek is a standard license holder meaning they make SOC with ARM designed cores rather then designing their own cores.

Nvidia working directly with pleb license holders probably will play well. But ya Nvidia isn't anything special they're are many other players building ARM chips for server/mobile/IOT markets ect. Nvidia doesn't need to own ARM to do anything they are doing... but it would give them some authority if they want to try and standardize platform bits... chipsets... and perhaps if everything goes well socket standards.
 
I don’t think we’re going to see sockets, the slot setups they are using for their dev kits work awesome and let’s them do complete SOC. But it’s all speculation so I’m probably wrong but a slot design plays better with others and NVIdia needs 3’rd party players if they are going to pull this off.
Yes they will have to talk the ASUS Gigabytes of the world to build parts... if they get the big OEMs like the Dells and HPs selling ARM laptop parts it will be all that much easier to get the MOBO builders on board.

Sockets aren't required for ARM to do well. If they want to go after all those markets as fast as possible however it makes things a lot easier. I am quite sure Grace server parts will be some form of proprietary socket... it just makes it easier to scale production and increase profits. Easier to produce one blade board that can slot a bunch of CPUs even if your going to sell a bunch of those boards with only half its sockets filled with chips.
 
Yes they will have to talk the ASUS Gigabytes of the world to build parts... if they get the big OEMs like the Dells and HPs selling ARM laptop parts it will be all that much easier to get the MOBO builders on board.

Sockets aren't required for ARM to do well. If they want to go after all those markets as fast as possible however it makes things a lot easier. I am quite sure Grace server parts will be some form of proprietary socket... it just makes it easier to scale production and increase profits. Easier to produce one blade board that can slot a bunch of CPUs even if your going to sell a bunch of those boards with only half its sockets filled with chips.
This is what their Jetson Arm chips look like. The pin-out is standard across the whole lineup so far. You can drop them into dev boards or slot them right into customized hardware good to go.
1618523216753.png


Their more complicated AGX chips use a connector like this on the back, which again is pretty well documented and slots down onto their development board or any custom solution thrown together. Grace likely uses something like this to connect to the master board.
1618523623222.png
 
Last edited:
This is what their Jetson Arm chips look like. The pin-out is standard across the whole lineup so far. You can drop them into dev boards or slot them right into customized hardware good to go.
View attachment 348077

Their more complicated AGX chips use a connector like this on the back, which again is pretty well documented and slots down onto their development board or any custom solution thrown together. Grace likely uses something like this to connect to the master board.
View attachment 348081
Good points... your probably right they might just stick with what there using for grace. I am still thinkin they are going to spin some version of grace consumer parts... where a socket might make sense if they can get board support from third parties. However I guess nothing is stopping them from offering the silicon in different versions its not like they can't just change the packaging anyway.

Will be interesting to see more details over the next year.
 
Good points... your probably right they might just stick with what there using for grace. I am still thinkin they are going to spin some version of grace consumer parts... where a socket might make sense if they can get board support from third parties. However I guess nothing is stopping them from offering the silicon in different versions its not like they can't just change the packaging anyway.

Will be interesting to see more details over the next year.
It also lets them leverage a full SOC design with active cooling on both sides if needed it gives them some options a socket doesn’t.
 
Good points... your probably right they might just stick with what there using for grace. I am still thinkin they are going to spin some version of grace consumer parts... where a socket might make sense if they can get board support from third parties. However I guess nothing is stopping them from offering the silicon in different versions its not like they can't just change the packaging anyway.

Will be interesting to see more details over the next year.
I could easily see a consumer version, and that second one I have posted there is well the prosumer version, it's their AI dev kit for autonomous vehicles and motion tracking and all that fun razmataz.

I have a handful of the Jetson AGX Xavier's on order for an AI lab one of the teachers here wants to do. Should be fun so I will have those hopefully soon.
https://www.nvidia.com/en-us/autonomous-machines/embedded-systems/jetson-agx-xavier/

Which you see is offered as just the module, or as the kit, and the kit is just the module and a daughter board it slots down onto that contains all the IO ports.
 
Nvidia is one of the handful of companies that have a ARM architecture licence. Qualcomm, Apple, Samsung, ARM, AMD, Fujitsu and a few others. Mediatek is a standard license holder meaning they make SOC with ARM designed cores rather then designing their own cores.

Nvidia working directly with pleb license holders probably will play well. But ya Nvidia isn't anything special they're are many other players building ARM chips for server/mobile/IOT markets ect. Nvidia doesn't need to own ARM to do anything they are doing... but it would give them some authority if they want to try and standardize platform bits... chipsets... and perhaps if everything goes well socket standards.
I just don't think they are gonna get approval to buy it. Based off them making server chips now, i think they assume the same.

This is all just my opinion of course.
 
I just don't think they are gonna get approval to buy it. Based off them making server chips now, i think they assume the same.

This is all just my opinion of course.
They where already making ARM AI chips... ARM self driving systems... powering the switch. I'm not sure a server chip that won't be ready for almost 3 years that so far has only been confirmed to be using stock ARM server cores disqualifies them.

Swapping ARM chips in to replace the AMD chips they where already using in that market if anything strengthens their proposal imo. Nvidia already sells into this market their current systems simply use Epyc chips. Really they are simply planning to pull a Apple and replace their outsourced parts with in house parts.
 
It appears they are only working on developing a standard, but that's how things get started.
Nvidia already did this and failed with the Shield products. Sounds like they want to emulate x86 like how Apple did with the M1, but that'll fail. Nvidia has a lot of work ahead of them if they want ARM to be able to play games. They need to get Steam to have an ARM client. They need to push developers to port their games to ARM. You aren't going to do that with some overpriced Nvidia laptops. If people had to choose between a $2k Nvidia ARM laptop or a $2k x86 laptop then they'll go with the x86. With AMD releasing their new 5000G series, there's little reason to go with ARM.

Nvidia should absolutely open up the ARM platform to other hardware manufacturers as well. While I love emulation and it has its place, you won't find many people spending $2k on a Nvidia ARM laptop to get worse performance than a equally priced x86 laptop.
 
Nvidia already did this and failed with the Shield products. Sounds like they want to emulate x86 like how Apple did with the M1, but that'll fail. Nvidia has a lot of work ahead of them if they want ARM to be able to play games. They need to get Steam to have an ARM client. They need to push developers to port their games to ARM. You aren't going to do that with some overpriced Nvidia laptops. If people had to choose between a $2k Nvidia ARM laptop or a $2k x86 laptop then they'll go with the x86. With AMD releasing their new 5000G series, there's little reason to go with ARM.

Nvidia should absolutely open up the ARM platform to other hardware manufacturers as well. While I love emulation and it has its place, you won't find many people spending $2k on a Nvidia ARM laptop to get worse performance than a equally priced x86 laptop.
It may not happen this year or next but it is coming, the development studios can cross-compile for either ARM or x86, Vulkan supports it, and so do most of the major engines. NVidia and the people they partner with don't need something today that can emulate x86 to the degree of a top-end CPU can they just need it to emulate it to the degree that it isn't the bottleneck at the desired resolution for the system say 1440p. The M1 has shown them that it can be done, and that's first-gen what is the M2 or M3 going to look like, that's what they have to compete against. If they were to partner with Valve on a few things and really drive home the SteamOS side of things they could have a viable system by 2023, Microsoft is also working towards many ARM goals so I wouldn't count out their efforts on this either. Lots of things are up in the air right now, but given MediaTek's existing portfolio I wouldn't expect to be seeing them roll out $2K machines out the get-go, but gaming-focused $600, or $800 machines for MOBA's and other popular esports and MMO titles, they could blast those out all over Asia and sell them as fast as they could get them out of the factory. And once there are enough of them in the field the other developers cant ignore them.
 
Shield TV is great for enthusiasts and keep in mind the switch is basically the shield console. They've done well with it.
Taking the Shield and using it to play emulators is not Nvidia doing well with it. Letting Nintendo use the Tegra hardware to make a successful console is not Nvidia doing well with it. That's Nintendo doing well with it. When was the last time Nvidia released a Shield product? It was last released in 2017. There's a reason for that. I remember looking at Nvidia's games on the Shield devices and I saw games like Half Life 2 and Borderlands 2. Games you can't play on any other ARM based Android device.
 
It may not happen this year or next but it is coming, the development studios can cross-compile for either ARM or x86, Vulkan supports it, and so do most of the major engines. NVidia and the people they partner with don't need something today that can emulate x86 to the degree of a top-end CPU can they just need it to emulate it to the degree that it isn't the bottleneck at the desired resolution for the system say 1440p. The M1 has shown them that it can be done, and that's first-gen what is the M2 or M3 going to look like, that's what they have to compete against. If they were to partner with Valve on a few things and really drive home the SteamOS side of things they could have a viable system by 2023, Microsoft is also working towards many ARM goals so I wouldn't count out their efforts on this either. Lots of things are up in the air right now, but given MediaTek's existing portfolio I wouldn't expect to be seeing them roll out $2K machines out the get-go, but gaming-focused $600, or $800 machines for MOBA's and other popular esports and MMO titles, they could blast those out all over Asia and sell them as fast as they could get them out of the factory. And once there are enough of them in the field the other developers cant ignore them.
Easier said than done. The main thing here is support, as in who's going to support a new platform with new problems? This is why Witcher 3 never came to Linux because CD Project Red got a bad reputation from poor Witcher 2 support. It's easy to compile a binary for Linux or ARM but that doesn't end there. Also, chances are games made for Nvidia's ARM laptops will only work on Nvidia's ARM laptops. Other people do make ARM based SOC's and most developers aren't going to waste their time porting games to support a very narrow and limited market. Which means we're going back to emulation of x86 which will be slower than x86 guaranteed. The Apple M1 has proven that Apple users are willing to deal with x86 emulation that's 25% to 50% slower than native, not that x86 emulation is viable for gaming. Even at $600 I don't see people buying these like crazy.

Nvidia needs to do a lot of things to move forward with ARM gaming.
#1 Open source drivers in Linux.
#2 Convince Valve to make an ARM Steam client.
#3 Start selling motherboards with sockets for enthusiasts.

Also, isn't MediaTek the SoC maker who floods the market with cheap crappy SoC's? Usually found in Kodi boxes? There's so many Chinese ARM based SoC manufacturers I can never tell.
 
Taking the Shield and using it to play emulators is not Nvidia doing well with it. Letting Nintendo use the Tegra hardware to make a successful console is not Nvidia doing well with it. That's Nintendo doing well with it. When was the last time Nvidia released a Shield product? It was last released in 2017. There's a reason for that. I remember looking at Nvidia's games on the Shield devices and I saw games like Half Life 2 and Borderlands 2. Games you can't play on any other ARM based Android device.
I'm not exactly sure what the Shield was supposed to be, and in the end, I don't think NVidia did either, but their ARM cores are progressing at a good pace. The Xavier platform is well rounded and Orin should come out the gate swinging, Atlan (Grace) should be a solid show stopper for its intended jobs. It may or may not end up leading to anything consumer at the moment, but until somebody builds it and gives it a try we'll never know.
 
I'm not exactly sure what the Shield was supposed to be, and in the end, I don't think NVidia did either, but their ARM cores are progressing at a good pace. The Xavier platform is well rounded and Orin should come out the gate swinging, Atlan (Grace) should be a solid show stopper for its intended jobs. It may or may not end up leading to anything consumer at the moment, but until somebody builds it and gives it a try we'll never know.
The purpose of Shield was to give Nvidia a console platform to compete with Sony and Microsoft. Do you think Nvidia is happy by not getting involved in those console sales? Same goes with Nvidia's Geforce Now. What Nvidia wants is a platform where the sub $300 graphic card market would instead go to make Nvidia money. It was a failure but Nintendo was able to come along and take all that Tegra hardware that nobody wanted and made use of it for the Switch.

Whenever a new platform is introduced, either by OS or CPU architecture, then a lot of work is needed to convince consumers to jump onto it, because the reality is that your new platform is going to have problems for consumers. Either the software isn't there, or the backwards compatibility isn't there, or the performance isn't there. This is why Windows XP was popular for a very long time, as well as Windows 7, because consumers want to avoid those problems. The shift to ARM Linux or ARM Windows is going to take a lot of effort and a lot of time. Apple's M1 isn't going to do as well as Apple hoped, due to these reasons. No amount of Geekbench scores and battery life is going to be worth losing x86 performance and compatibility.

 
The purpose of Shield was to give Nvidia a console platform to compete with Sony and Microsoft. Do you think Nvidia is happy by not getting involved in those console sales? Same goes with Nvidia's Geforce Now. What Nvidia wants is a platform where the sub $300 graphic card market would instead go to make Nvidia money. It was a failure but Nintendo was able to come along and take all that Tegra hardware that nobody wanted and made use of it for the Switch.

Whenever a new platform is introduced, either by OS or CPU architecture, then a lot of work is needed to convince consumers to jump onto it, because the reality is that your new platform is going to have problems for consumers. Either the software isn't there, or the backwards compatibility isn't there, or the performance isn't there. This is why Windows XP was popular for a very long time, as well as Windows 7, because consumers want to avoid those problems. The shift to ARM Linux or ARM Windows is going to take a lot of effort and a lot of time. Apple's M1 isn't going to do as well as Apple hoped, due to these reasons. No amount of Geekbench scores and battery life is going to be worth losing x86 performance and compatibility.


Gaming looks half decent there, way better than what most Intel IGP’s are capable of. Apples M1 is going to do fine, in every scenario a 2021 M1 outperforms the 2020 Intel model so it’s an upgrade.
 
Taking the Shield and using it to play emulators is not Nvidia doing well with it. Letting Nintendo use the Tegra hardware to make a successful console is not Nvidia doing well with it. That's Nintendo doing well with it. When was the last time Nvidia released a Shield product? It was last released in 2017. There's a reason for that. I remember looking at Nvidia's games on the Shield devices and I saw games like Half Life 2 and Borderlands 2. Games you can't play on any other ARM based Android device.
Letting nintendo use old hardware from a failed tablet launch to sell 78 MILLION units is doing pretty ****ing solid :)
 
Gaming looks half decent there, way better than what most Intel IGP’s are capable of. Apples M1 is going to do fine, in every scenario a 2021 M1 outperforms the 2020 Intel model so it’s an upgrade.
Not really. It really depends on what you do but for the most part you can do more on an Intel than a Apple M1. If you use Apple software and nothing more than yes the Apple M1 is an upgrade. If you play games and use some Windows apps then it's a downgrade. The ARM Mac OSX isn't about to grow as a ecosystem either. They did port Metro Exodus onto the Apple M1 and it doesn't look good. Lots of stutter in the footage. Compared to the Intel HD 630, I would consider it more playable on the Intel.



 
Letting nintendo use old hardware from a failed tablet launch to sell 78 MILLION units is doing pretty ****ing solid :)
Exactly how many Tablets and Smart phones you see with Nvidia hardware in it? It's nearly all Qualcomm based now, with some exceptions.
 
Exactly how many Tablets and Smart phones you see with Nvidia hardware in it? It's nearly all Qualcomm based now, with some exceptions.
That 78 million is Switches. Sure, outside of that, you're right, but 78 million's nothing to sneer at.
 
That 78 million is Switches. Sure, outside of that, you're right, but 78 million's nothing to sneer at.
Yes but it doesn't mean that Nvidia has a chance to do it alone. Nintendo has a lot of unique games that gives them the edge. Nintendo could swap out the SoC for another brand and you wouldn't know unless told. Nvidia has like... really good drivers.
 
Yes but it doesn't mean that Nvidia has a chance to do it alone. Nintendo has a lot of unique games that gives them the edge. Nintendo could swap out the SoC for another brand and you wouldn't know unless told. Nvidia has like... really good drivers.

Consoles get straight to the metal access, so while you only ever see a portion hardware's potential on a PC, you can get all of it on a console. That plus being able to rip out all the bloat and nonsense from a full PC OS, in addition to having only one or two hardware specs to target straight translates into something that's more efficient and more able to use what it has.

The problem with a "gaming" laptop in this case is there is no way of knowing if they mean "a gaming console, with all of it's advantages, slapped into a laptop", or "a gaming laptop, with all of the shit that hurts them for gaming, powered by console guts". And that's sort of been the issue with all these sorts of things in the past be the tablet or handheld PCs for "gaming". They don't really know what they want to do or be. So they end up being a cluster fuck that doesn't really do the computer or the gaming side of things well.

The advantage Nintendo has, outside of it's own titles to leverage, is they know exactly what they want to do and be.
 
Back
Top