Nvidia to make Arm-based PC chips in major new challenge to Intel (opens in new tab)

The Tegra series is still ongoing, its just marketed at self-driving vehicles.

You can get dev boards, but they are hilariously expensive, and I'd guess the idle/low load performance pales in comparison to other vendors these days.

I had tegra 2 tablet. Couldn't play video my year older phone could. Also didn't have NEON so was stuck using old versions of everything.

> gaming will likely never move to arm.

Oh, it can if Nvidia, AMD and Microsoft push it.

Most of the back catalog can probably run fine emulated, though you may want to stick to x86 for those older CPU bound sim games that aren't going to get a recompile.

> Do they have an answer to rosetta so the transition can be mostly seamless (for everyone except developers if the M series is any indication...).

Microsoft had an AoT x86->ARM (and now, amd64->aarch64) binary translation layer before Apple's Rosetta had that capability (at least, publicly).

> I am curious if this is actually from an increase in Mac sales or a decrease in PC sales and Mac has just been stable?

Once the M1 came out everybody HAD to upgrade, so Apple got a huge boost for a couple years as people cycled out their x86 stuff.

Latest data shows Apple's computer sales slumping much harder than the PC industry in general.

You stopped hearing about it because the chips Qualcomm made were dreadful.

Most gaming is on ARM (mobile) now I think. Most of the revenue too.

From the stats I could gather I really doubt macOS market share increased substantially in usage. What I think happened is that they doubled their usual sales figures. Which is unsurprising if you look at the buildup of not-so-great hardware they release post-2015 before AS. Their machines had annoyances and were very uncompetitive with comparable offering because of soldered RAM/SSD pricing.

Then they released AS and the first round of hardware looked competitive because at least they seemed to have something different, a real advantage worth paying more. At least that was the marketing. I believe it got many to update very old hardware that was kept running because Apple offerings seemed so out of touch; then some others got interested to "complete" their iOS devices and even traditional PC users got sucked in for novelty factor or battery life argument. In practice if the second-hand market is to be looked at, many went back to other machines and the market is inundated by underpowered, overpriced, close to entry level machines (people figure out the hard way that 8GB of RAM is very tight for a post 2020 computer no matter how good your software optimisation is...).

Now the second release of AS was disappointing to say the least (pretty bad considering the price hikes) and I think many are holding to see what they can do with M3. So the sales have dropped a lot, at least as much as all other OEM if not more (especially in comparison to previous years).

So I think what they call the market share is actually the sales number, that doesn't mean much. If you account for all the hardware that got retired plus all the hardware that sit unused (waiting to be sold or else) macOS market share has been slightly slopping upward at best. Mostly stable in practice.

When you look at the sales numbers, it is almost 80% laptops. They barely sell any of what you would call a "PC". It makes sense; since laptops is the only place where AS has any advantages and it is also the only way to use your extra expensive "computer" as a status symbol. This is also what most employers are going to buy for their staff because unless you need real power where AS is almost disqualified from the get go it makes everything easier. They used to sell a lot of iMacs (especially the 27" version) because it was very convenient, but they don't have those anymore so...

If anything, macOS is becoming less relevant as a computing platform by the day and more of a luxury alternative brand. So its market is becoming less relevant by the day too. I think Apple is on the path to become to computing what Campagnolo is to cycling...

Traditional PCs are not going anywhere and for way more reasons than just upgradability (your car is not technically upgradeable but is made with components from many different competing suppliers).

We can get that with x86.

We just don't, because it's cheaper to clock the chips higher and burn more power, and to keep large graphics seperate from the CPU.

Intel's/AMD's attempts at addressing this (Broadwell/Skylake with eDRAM, Vega M, AMD's Van Gogh and Dragon Crest) were universally shot down by laptop OEMs. I don't know why, but they were probably just being cheap.

Let's not exaggerate here. I've got the M2 Pro Mac Mini, and it is nearly silent, but depending on your workload you still generate nontrivial heat and the fans do need to run to keep it cool. Maybe I notice more because I build a lot of software, but this is not some magic bullet that avoids the need for traditional heat management.

We've had them direct from Intel since at least 2013: https://en.wikipedia.org/wiki/Next_Unit_of_Computing

> Advanced Micro Devices (AMD.O) also plans to make chips for PCs with Arm technology

Technically, the Platform Security Processor that's integrated in most (all?) Zen CPUs is already an ARM chip, though I guess that they mean actual general purpose CPUs.

>Why only Apple is able to pull it off? Why this lack of care from everybody else?

Apple got burnt, multiple times, by both Intel and Nvidia. That set them on a complete war path to move to arm where they control the chips. As part of that war path, their goal was to drop x86 entirely so they needed the transition layer.

Microsoft has no need to drop x86 entirely, in fact x86 will continue to remain a good part of a market for the foreseeable future. Who knows, maybe RISCV suddenly takes off because even Qualcomm has begun pushing a cool billion dollars towards RISCV development due to ARM/Softbank attempting to make them destroy their IP. Heck, ARM/Softbank is basically trying to destroy the ARM market for profit by terminating Qualcomm's licenses by 2025.

> Microsoft didn't do any of that, and still doesn't, their leadership is clueless and dangerous

It did with .NET but most software isn't .NET I think.

The problem with the current Qualcomm cores is that they are cheap and low power, but the new Nuvia cores should alleviate this.

...But be careful what you wish for. There have been some promising ARM core designs (Samsung's Mongoose series, Nvidia's Denver/Carmel) that all ended being worse than tweaked ARM-designed cores.

Others (Marvell's SMT ThunderX series, Fujitsu's HPC A64FX) were too niche, and ultimately discontinued.

Also, based on the M2's rather limited gains, some are suggesting that the M1 was an anomalously good design, and that Apple can't necessarily keep that massive edge.

Everyone keeps going on about ARM, but isn’t the difference mostly memory latency and throughput? Less waiting for memory means less wasted clock cycles. And probably also hardcore engineering in choosing the right tradeoffs? The x86 "penalty" seems to be mostly a few precent from what I’ve read. Yes x86 wastes some transistors on legacy instructions, but it doesn’t fully explain the year long difference between Apple and Intel and AMD. (Although AMD seems to catch up now and Intel also a bit with their HBM2 chips.) It’s also TSMC and Apple working together on designing a great chip.

First step is convincing Windows development community that Windows on ARM actually matters.

Until then, whatever Apple has over 80% of the worldwide desktop market hardly matters.

And after how UWP was managed, there aren't that many that care.

I suspect, based on almost nothing, a vibe perhaps, that this will be an iteration of the Denver core. Previous versions have not been amazing on power iirc, I don’t think they got major design wins with it. I think at one point (2010 ish?) they considered an x86 processor based on this core as well. How much it remains a dynamic recompilation based “software isa” that could be x86 or arm or risc-v I have no idea.

A linux kernel on Arm is by far the widest deployed pairing in history.

I think ATX style machines are very niche. Gamers love them and 3D artsits, but outside of that I think no one wants them anymore. I think that ARM-compatibility for games in the near term is a hard sell (although Blender, Maya, Photoshop, Final Cut Pro have been ported to ARM for running on Apple chips.)

I think the market that NVIDIA should be chasing with the ARM CPUs + good GPUs is business machines. Our company is filled with very poor performing Window Surface devices - outside of those who insist on Macs this is what people get. Companies are spending a lot on these. And they desperately need better performance while also being cool with long battery life.

You can have your cake and eat it with a wide LPDDR5X bus. This is what Apple does, and its fine. Other specialized chips (like Tenstorrent's accelerators) do this too.

I think an AMD 7900-like approach where the memory controllers/cache are on tiny external chips is particularly practical. Its efficient and economical. I hope AMD (and others) repeat this with laptop CPUS.

GDDRX is not a good fit for laptops anyway because its so power hungry. GPUs and the Xbox/Playstation use it just because its the absolute cheapest bandwidth/$, at the cost of everything else.

HBM is very expensive and being hoovered up by the AI market. I wouldn't count on seeing it in consumer stuff again.

It's more of a low latency (DDR/LPDDR) vs. high bandwidth (GDDR/HBM) tradeoff.

People are misunderstanding the source of Apple's advantage.

A Grace Hopper workstation sounds cool but I would expect Nvidia PC chips to be an evolution of Orin.

Because the best chip designs in the world are made by Apple, Intel, and AMD. ARM's chip designs are years behind those three. Apple has shown ARM can have excellent performance, but they are the only ones to do it so far. Android users have shown that they don't care about buying new phones with CPUs that are slower than 3-year-old iPhones.

Everyone but apple uses ARM-designed cores. Why are ARM-designed cores behind? Go check what the comp is for a CPU designer at apple, and then go check what it is at ARM. Suddenly, you'll be overcome with a feeling of clarity.

It takes four years or more to design a CPU. The M1 came out three years ago.

Yeah, Intel got xScale (née StrongArm) from Digital Equipment Corp as part of a settlement after the Alpha lawsuit for infringing on Pentium patents. Intel bought StrongArm which maybe meant Intel didn't really want it and sold it to Marvell. Intel didn't want Alpha either which was sold to Compaq, the last resort and buyer of all things great and small.