Discussion NV Re-Enter ARM PC market in Q2-2026!

[SpudLobby]

Modern MTK SOCs are leaps and bounds ahead of every Tegra ever made in history of humankind.

Widest doesn't mean the best.
The good part is that modern MTK SOCs are competent at worst; i.e. they're not tegras.

Agree with this on both parts. If you had to have NV IP implemented having MTK do it on their own > having NV do it on their own. MediaTek has come a long way.

 

[FlameTail]

Nvidia SoC would be very attractive to the kind of people who would buy a top end AMD APU for gaming.

 

[gdansk]

Nvidia SoC would be very attractive to the kind of people who would buy a top end AMD APU for gaming.

It depends, doesn't it?
How many Windows games will have an aarch64 build?

 

[FlameTail]

It depends, doesn't it?
How many Windows games will have an aarch64 build?

Few. But the thing is Nvidia making an SoC would incentives developers to port their games to ARM, more than anything else.

 

[gdansk]

Few. But the thing is Nvidia making an SoC would incentives developers to port their games to ARM, more than anything else.

I don't think Nvidia will be involved enough to be giving developers money for that.
But maybe I'm wrong.
Between Qualcomm and MTK/Nvidia maybe there will be enough users to pressure developers into it anyway.

 

[Tup3x]

I don't think Nvidia will be involved enough to be giving developers money for that.
But maybe I'm wrong.
Between Qualcomm and MTK/Nvidia maybe there will be enough users to pressure developers into it anyway.

The way it's meant to be played - Runs best on ARM™️. x86 version optimised for 486.

 

[Shivansps]

Few. But the thing is Nvidia making an SoC would incentives developers to port their games to ARM, more than anything else.

Its not that would be too difficult, it mostly depends on the engine and nvidia blobs for their propietary stuff.

But Nvidia needs to start by releasing ARM Windows drivers... That something we could use TODAY.

 

[soresu]

I don't think Nvidia will be involved enough to be giving developers money for that.

I don't think that they would make the effort to enter the market if they didn't intend to induce developers to support the cause through money or donated coders to port their game engines.

Luckily for them more and more games are being built on just 2 major game engines, Unreal and Unity.

With Unity making a good show of stuffing their feet in their mouths I can see it shifting even more to Unreal in the future.

So if UE5 supports ARM, then there will be plenty games out there for it going forward.

 

[Saylick]

Mediatek + Nvidia ARM SoC in late 2025?

As Qualcomm's exclusivity for Arm-powered processors for Windows PCs is reportedly coming to its end, other chipmakers are getting ready to offer their Arm-based system-on-chips for Windows computers. And, according to a new report from Reuters, MediaTek will be among the companies jumping into the Windows-on-Arm field, with plans to launch their first PC processor late next year.

https://www.anandtech.com/show/21442/report-mediatek-working-on-armbased-processor-for-windows-pcs

 

[FlameTail]

@igor_kavinski

I wonder how the Nvidia+Mediatek partnership is going to work.

We know that both companies are independently developing their own ARM PC SoCs.

Here's all we know as per rumours:

Nvidia's SoC
• 2025 release date
• TSMC 3nm
• LPDDR6
• High End part ($300 OEM price)
• Developed with the assistance of Mediatek

Mediatek SoC
• Late 2025 release date

We scarcely know anything about Mediatek's project.

I guess Nvidia is targeting the high-end segment, while Mediatek is targeting the mid-end/low-end segment. It would be funny if as partners, both Nvidia and Mediatek launch products of the same class that compete with each other.

 

[igor_kavinski]

I guess Nvidia is targeting the high-end segment, while Mediatek is targeting the mid-end/low-end segment. It would be funny if as partners, both Nvidia and Mediatek launch products of the same class that compete with each other.

It could be some sort of cross-licensing agreement. Nvidia gets access to their ARM core related IP to use and customize to their needs and vice versa for Mediatek with regards to some low powered GPU from Nvidia. No money changes hands which is what must've made the deal sweet for both of them.

 

[FlameTail]

Unlike Snapdragon X who is being sued, Nvidia-Mediatek's SoC has the full blessing from ARM;

 

[Thibsie]

Wait for Nvidia buying Mediatek, lol

 

[igor_kavinski]

Jensen, you buy now!

 

[Ghostsonplanets]

Nvidia is the biggest chance WARM has to be sucessful.

If Mediatek brings their design class to WARM with good pricing, I could see it being successful too.

 

[FlameTail]

Nvidia is the biggest chance WARM has to be sucessful.

If Mediatek brings their design class to WARM with good pricing, I could see it being successful too.

Is the Qcom dream busted?

 

[Thibsie]

Nvidia is the biggest chance WARM has to be sucessful.

If Mediatek brings their design class to WARM with good pricing, I could see it being successful too.

On terms on investment, R&D etc, Nvidia and Qualcomm are the biggest chances.
In terms of what SOCs will be round in the WARM PCs, Mediatek might become a lot more common IMO.

 

[Thibsie]

Is the Qcom dream busted?

The took too long and hyped it way too much.

 

[burninatortech4]

Is the Qcom dream busted?

What dream was that? Youtuber boot-licking?

 

[adroc_thurston]

Is the Qcom dream busted?

Gen2 part is pretty far off so for now, yes.
No winning for you.

 

[soresu]

I wonder what the likelihood is of ARM Ltd designing a CPU µArch with TSO.

 

[Doug S]

I wonder what the likelihood is of ARM Ltd designing a CPU µArch with TSO.

If their customers demand it, they probably will.

It isn't like it would hurt Android performance, because you don't have to turn it on when you aren't using it. I'm not really sure how much extra work/verification it is to do though. There's a reason ARM (and most ISAs) use a more lax memory ordering. Heck, the only reason Intel uses such strict memory ordering is historical reasons - x86 traces its lineage back to the 8/16 bit days when every ISA was like that.

The more lax ordering came along as part of the RISC philosophy of making the hardware simple and there were ways around those issues for when strong ordering really mattered. The reason Apple/Qualcomm need that mode for emulating x86 is that x86 software doesn't use those "ways around those issues" because it just assumes the strong ordering implicitly, because that makes programming a little easier and the developers didn't expect it would ever be run on anything other than x86.

 

[soresu]

If their customers demand it, they probably will.

They can demand all they like, but if the long term WoA market isn't there to justify it they won't put any R&D nehind it.

 

[DZero]

Wait for Nvidia buying Mediatek, lol

Lol, but how about a merge, MTK is now a very competent company and nVIDIA is the biggest GPU company in the sector.

What I want to see is Windows on other ARM processors like MTK ones.

 

[Doug S]

They can demand all they like, but if the long term WoA market isn't there to justify it they won't put any R&D nehind it.

True enough, Qualcomm and Microsoft seem to have completely bungled the much hyped launch. That isn't enough to kill WoA but it has dealt it a major blow. If Microsoft can't fix the issues on their end (like Prism) none of it will matter, because there isn't any incentive for developers (especially of lower level stuff like drivers or AV software) to port to ARM regardless of how many clueless fools want to claim that will be 50% of the Windows PC market in five years.

I think a much overlooked factor of Rosetta 2's success is how long Apple was preparing for it - and how much experience / institutional knowledge they had from the previous PPC to x86 migration (I don't count the 68K to PPC migration because that was so long ago, and it was before they were using their modern Mach/BSD based OS)