Nvidia closing in on deal to buy ARM *spawn*

Mali IP is known to be pretty cheap AFAIK. If nvidiarm happens and Mali gets shelved, nvidia won't charge the same for Geforce as ARM did for Mali, and that might cost the viability of the smaller SoC makers (AmLogic, Rockchip, etc.).

 

I don't know what NV's plans are, but for the time being it's times cheaper to keep the Mali team developing for ULP SoCs then to develop a ULP GeForce GPU.

 

Exactly what I said 3 pages ago :
https://www.thestreet.com/investing/4-key-things-nvidia-just-shared-about-its-plans-for-arm

Only difference is that Nvidia plans to continue MALI effort for customers only interested by graphic acceleration.

the important piece of information is obviously:

 

ARM being owned by an American company also means that the architecture could be "weaponized" against China, since a lot of cheap mobile/embedded/IoT processors are made there, by Chinese companies and using ARM licensed cores. What happened to Huaweii could happen again to any other company, thus making business with ARM may become a liability now; alternatives like RISC-V seems like a safer bet due to their 'open' nature.

 

Everyone: Ugh, what terrible thing will 2020 bring next?
NV: *cough* I got 'dis

 

Tech owned by any nation (other than China, obvs) can be weaponised against China by the US. It's fairly standard operating procedure for the US, when it imposes sanctions on a country, to turn around and threaten all third-party countries that unless they follow suit they will also be the subject of sanctions. ARM being owned in the UK specifically would be no guarantee of anything, given the sub-dom nature of the UK-US relationship over the past couple of decades.

 

Is Mali business model different than ARM? Are ARM licensees allowed to modify it? Or is it sold as is? If the former I can't see Nvidia using GeForce like that at all. Regardless it's not like Mali represents a substantial part of the market anyway, much less the high margins tier. Really doubt Nvidia would start selling GeForce tech to third parties. AFAIR they've tried that before and didn't work very well. It's not like suddenly the bottom feeders of the Mobile Market would have the budget to pay for it.

 

AFAIK Mali are IP blocks and except for deciding the number of "cores" and clock speeds there's no further customization possible. nVidia will want Geforce in everything as controlling the ecossystem is what they'll try to do in the long run.
They successfully did that with CUDA on the GPGPU market by not contributing to open standards while pressing their proprietary ones, then physics in games by forcing single-core x87 paths on CPU PhysX so that only Geforce GPUs could run them at decent performance, and later on the CUDA AI acceleration, and they keep imposing custom code / features on PC games aimed specifically at reducing performance on the competition (like sub-pixel triangles in Hairworks and tesselated invisible seas in the Kepler/Maxwell era).

This is the exact kind of practices we should expect from nvidia (e.g. block ISA evolution and optimization from their competitors and use that leverage to expand into new markets and crush the competition).
This is not good for anyone, industry or consumer. nVidia is obviously not going to use ARM just to boost their server offerings in the long run.

I don't even know how handling the keys to the castle to a single SoC maker can ever be considered a good choice for the consumers, the industry and even the technological advancement in society. It's just a bad event for everyone but nVidia.


I really hope regulators take nvidia's known past of corrupted practices into consideration. And let me be very clear that my opinion on this has nothing to do with the obviously enormous talent of nvidia's engineers. These are just the directors'/executives' trends of choices in the company, nothing else.

Mali is in all current SoCs from Samsung, HiSilicon, Amlogic, Rockchip and Xilinx, and save for a few with PowerVR GPUs it's in most SoCs from Mediatek and Unisoc.
If there aren't more Mali GPUs than Adrenos out there, then it's probably a very close second.

What didn't work for nvidia was 4G modem integration after they purchased Icera. After that failed, they saw that competing against Qualcomm on high-margin SoCs would be nearly impossible.
Their ULP GPUs were never considered a failure AFAIK. If the current rumors and inside info are anything to go by, they're about to release a new SoC for an upgraded Switch.

 

Beijing’s concerns loom over Nvidia’s $40bn Arm deal

 

Yup just as I said 2 days ago. The deal is far from being done.

 

In terms of high tier he mentioned, a very high percentage goes to Apple/IMG GPU IP, the closest second should go to Qualcomm Adreno (as Samsung also uses QCOM SoCs for it's high end smartphones) and Mali coming in third. Outside of the mobile consumer markets IMG has still the lions share in units sold, irrelevant if they don't make any profit from them (like automotive f.e.).

It wasn't ever a success either; neither in volumes nor in turnover until the high end automotive market started moving for them and they departed from the mobile market. A handheld console design win or two doesn't really suggest that Tegras flooded the smartphone/tablet market at any stage, otherwise they wouldn't had left the market in the first place.

 

Regardless, the Switch's success effectively made the ULV Geforce a success.
That accounts for ~65 million ULV Geforces out there.

 

50/50 Chance is so much higher than I thought.

 

By the time the Switch kicked in they probably already made the first profits from automotive, for which the Switch got more like leftovers from them moving out the mobile consumer market than anything else. 65Mio units isn't going to give you even a break even point for designing, developing and supporting an entire SoC Project like Tegra.

 

Nvidia’s Deal for Arm is a Game Changer
September 18, 2020

https://www.eetimes.com/nvidias-deal-for-arm-is-a-game-changer/#