Will we get a Desktop class ARM device from anyone besides Apple?

With all the excitement over the new Apple M1 SoC and its performance/Watt advantages is anyone else going to deliver a 1st class ARM device that isn’t Android based?

P.S. I don’t consider Raspberry Pi to be Desktop class performance yet…

3 Likes

At some point we will, its just a matter of when.

Qualcomm is about ~3 years behind Apple when it comes to CPU performance, so we would expect Qualcomm ARM chip to reach performance parity with x86 in ~2023

Note that Apple themselves have only reached performance parity with x86 last year.

Of course, Intel (or any other x86 chipmaker for that matter) will try to extend the domination for as long as they can.

Lakefield is a good example, it’s the effort to bring x86 cores to an SoC, the form factor that’s commonly used with ARM cores. Unlike regular desktop chips that contain only CPU and occasionally GPU cores, an SoC could contain RAM, chipset, various controllers, etc. so you could save as much power and physical board space as possible, which is critical on portable devices like phones and tablets.

It also brings similar technology to ARM’s big.LITTLE, combining low-power cores that you would normally find in an Intel Atom product, and high-performance cores that you would normally find in an Intel Core product.

2 Likes

Would you say that is a true CPU performance gap of ~3 years? All the benchmarks I’m aware of are to some extent also comparing the software stack and I think its fair to say the Apple/iOS software stack is more efficient than Android or Windows.

The M1 MacBook running ARM Windows in a VM on top of macOS is still much faster than a Surface Pro X running Windows natively. That says a lot.

Here’s an interesting graph


Unlike a regular single-threaded benchmark, this graph has the x86 CPU running two threads (on one core), thanks to hyperthreading, while an ARM core is already saturated just with a single thread.

Basically, if the cores are saturated, the M1 isn’t really that fast compared to x86 cores. In fact, a 14nm core (9980HK) still manage to slightly beat the 5nm M1, albeit at higher power consumption. However, it is near impossible for a single thread on x86 to fully utilize a core due to how complex it is, hence the performance penalty

5 Likes

I’m now thinking this might happen within the next 18 months…

Leaked new ARM based laptop…

This is ignoring AMD’s shakeup into the market, as well. There’s a lot of leaked information going around that suggests they might have a 20-25% IPC increase for multiple generations to come - Zen 3 was just the start. I figure Intel is pretty much dead in the water for the next while; it’s going to be a heat between AMD and Apple for the CPU performance crown.

2 Likes

Maybe Intel isn’t quite dead in the water yet…
Looks like a whole new architecture with potential for lots more power efficiency?

Snapdragon SC82820 on the way…

I’m not sure your interpretation of this chart is fair.
The M1 chip is an ultra low power class chip (10/13w) where as the i9 is a 45w chip, so you are not comparing chips of similar TDP.
Lets see where the Mx chips sit once they are available in similar power draw brackets.

2 Likes

They are artificially limited to one core though, so the tdp doesn’t matter as much, maybe the bigger chip has an advantage because of the bigger surface to dissipate heath.

Why is there no Zen 3 Cpu in that chart?

Also it remains to be seen if Apple can scale that power up to a desktop sized chip.

I guess because its benchmarked before Zen 3 was widely available to the public? Im not sure.

More updates on Qualcomm chips…
https://winfuture.de/news,121737.html

I’ll be honest that website was so horrible. I closed it before I actually read anything.

1 Like

We’ve got brand new ARM chips on the way…

2 Likes