Qualcomm ‘on track’ for Apple M-level Nuvia chips in late 2023

Qualcomm has vowed to unleash a generation of laptop processors that will rival Apple's M-series chips, though they won't arrive until late 2023. That's the latest word from Qualcomm CEO Cristiano Amon, who said development is "on track" for the Arm-compatible, Windows-friendly system-on-chips during his company's second- …

[aerogems]

Cautiously Optimistic

I will be eagerly awaiting some reviewers getting their hands on some of the final product. If they can get within even striking distance of x86, and without taking some of the same shortcuts Apple did like integrating RAM and GPU, they could have a real winner on their hands.

[Anonymous Coward]

Re: Cautiously Optimistic

> without taking some of the same shortcuts Apple did like integrating RAM and GPU

The whole point of these chips is that they have integrated components. Every time you want to leave one chip, cross the motherboard and enter another chip is take a shit load of time. For most workloads the CPU spends most of its life waiting for data to process, cutting the access time by bringing the RAM onboard leads to big performance wins. So it is precisely the fact that they brought the stuff on board which allows these things to run at the speed they do.

It is unlikely that we will see processors with external RAM running at the same speed as chips with integrated RAM unless someone either finds a totally new way to make chips work or they invent an entirely new sort of physics. The same happened years back with cache, you used to be able to make L1 cache external to the CPU and still get a 1 cycle access time but as clock speeds increased this stopped being possible and it had to move to being on board and we no longer see multi MB L1 caches.

Integrated RAM is obviously capacity limited. My hunch is that we'll see systems needing more RAM basically becoming tightly clustered groups of CPUs with onboard memory and that you'll add more RAM by adding more CPUs. Either that or we'll see traditional RAM being used as fast swap space.

[DS999]

Re: Cautiously Optimistic

Apple's use of lots of LPDDR channels was primarily dictated by the desire to scale to very high bandwidth to reach discrete GPU performance levels with their integrated GPU.

Since Qualcomm is unlikely to target desktops at all, let alone workstations like a Mac Pro, they don't need to aim as high for memory bandwidth. But I wouldn't expect to see them support discrete GPUs, they'll target an integrated GPU that's good enough for the broad home/business laptop market and not care about the gamers or CAD users on the high end. It isn't like there are any Windows/ARM games anyway, and porting demanding graphics stuff like CAD applications to it is a sort of chicken and egg problem.

Maybe they might think about supporting discrete GPUs after a few years if they think there can be a high end market for Windows/ARM, but for now they just have to worry about creating a Windows/ARM market AT ALL. Qualcomm's current underwhelming line is not good enough to make that happen.

[IGotOut]

I'm curious of the market.

So who is this aimed at?

They keep comparing to Apple, but that's a closed eco system, then saying its Arm but Windows compatible (emulation?).

So are they trying to take on Intel and AMD in the laptop arena or to create some kind of new device?

[druck]

Re: I'm curious of the market.

Lets think of a name for this new device, how about a Linux box.

[Peter D]

A backward step

At this rate within a couple of Christmases people will have to return to heating their home offices the traditional way.

[Anonymous Coward]

Re: A backward step

Ha!

I've been known to fire up another test server when the office is a bit cold. It uses less power than the heater but often seemed to provide just enough heat :-)