How A Raspberry Pi 4 Performs Against Intel's Latest Celeron, Pentium CPUs

[CommunityMember]

Apple will have fast ARM SoC but they will never publish it or even allow anybody else to use it.

Apple Silicon is a captured ARM solution, it is extremely unlikely it will ever be seen outside of the Apple eco-system.

That does not say that competitors cannot produce equivalent SoCs, but they have some catch-up to accomplish, and it will require a lot of investment.

The danger (for some) is that if only the usual suspects (Alphabet, Apple, Amazon, Facebook, Microsoft) and their partners have the resources to throw at captured solutions you may never see how good ARM could be on the desktop (just good, not great) outside of products from those companies.

[wizard69]

No general disagreements with you - but I for one have been really disappointed with rpi4's low performance (including for my own benchmarks).

Given that Apple's a closed shop, I think that for an ARM revolution to take place, we need a great chip design that is not Apple's. I'm hoping that ARM themselves do that, but with so much ownership changes ... I'm not too hopeful. Of course, this is just me speculating.

If Apples processor is as good as I think it can be it will force changes in the industry. If nothing else that is something good that Apple is doing. By the way Apples move here isn't really about ARM, even though I suspect they will have industry leading performance. Rather I see special function units being the big performance driver in their chips, with Neural Engine getting a big boost in the next round of chips. Once the hardware is in place I expect a huge move towards AI/ML techniques in their software.

So what many will be seeing as great performance in Apples new machines will not always be because of Apples ARM processors.

[wizard69]

Apple knows moving to ARM is a journey (similar to the journey from PowerPC to x86), and and it will take some time to get to the destination, but they have chosen a path. The day when Apple announces the Mac Pro has moved to ARM is the day when Apple, themselves, actually believes they have actually delivered on the promise. Until then, it is just marketing (and Apple is nothing if not good at marketing).

Actually I think a lot more is going on here than Marketing. Apple has a vision for its computing hardware and ARM is only a small partner here. I'm pretty much ocnvinced after viewing a few WWDC videos this year that ARM is just an convenience to them on their journey to far more interesting platforms.

[herman]

If Apple does their ARM well I can see it forcing a major change to the industry. this is one reason why I waiting patiently for Apple to deliver their performance processors. I'm expecting to see the A14 in a number of Apple products this year but I don't see that as the performance solution that Apple needs for the laptops and iMacs coming this year.

According to the rumors, the laptop processors will be "based" off of the A14. Whereas the A13 currently uses 2 big cores and 4 little, one of the new processors is rumored to have 8 big and 4 little.

[tuxd3v]

I agree. In my own personal (and very subjective) assessment, Apple is 3-4 years ahead of the competition much like it was with the iPhone.

So much ahead that even a change in the date, had rendered that iphone useless..

[oiaohm]

The RPi4 chip (Cortex A72) doesn't have a proper Load Store Queue, to allow it to issue loads out-of-order with respect to earlier stores (so all prior stores need to complete before issuing a load). This would require machinery to detect and correct potential misspeculations, of course.

Heres a big question is a ARM core a Load Store design.

Historic Cambridge designs for CPU are a Store Store Load design. Not load store design from Berkeley and others this is important.

The Cambridge design gets interesting.

So you have the cpu core doing out of order stores to the L1/L2. But you also have the L2 doing store of values from the L2 into L1 of its own accord. This gives the horrible possibility if prediction in the L2 is working well that load store operations come store store operations and no loads at all are performed.

So of course Cortex A72 does not have a Load Store Queue its Cambridge it has store queues and cpu performing a load is that the L2 unit shoving data into cpu core L1 has failed to shove what would be need into L1.

Yes while the cpu is completing stores out to L1 the L2 is completing it stores into L1. This is a different beast the Berkeley designs are commonly design that the top down is controlling what the cpu L1 has. As in cpu has to do a load to trigger items to move from L2 to L1. Where Cambridge based designs like arm this is not the case. Yes a Load Store Queue is required in a Berkeley based design. Cambridge based design Load Store Queue is incorrect idea instead its make the L2 controller logic for pushed data into L1 smarter so cpu core is avoiding issuing loads out.

Something you missed vladpetric is that L2 can in a Arm/Cambridge design put data into L1 that the cpu did not request this radically changes how you have todo speculation. There are of course security risks to the Cambridge design particularly the early one where L1 cache space on inactive cores was used to expand L2 cache space for active cores.

Sometimes the Cambridge design gives a hell of a speed boost being able to skip the load stage almost completely "Coremark v1.0CoreMark Size 666 - Iterations Per Second" really shows this. Now remember the PI is only running at 1.5 Ghz and the slowest of these Intel chips is over 3Ghz and not all the benchmarks here X86 winning by enough to allow for their higher clock speed and way more cooling than the PI.

Yes the Cambridge design does take hits with misspeculations at times worse than the Berkeley style Load Store Queues. Its about time you stop applying Berkeley logic to Cambridge based designs.

[drakonas777]

It does not make much sense to compare RPI4 with entry level desktop CPUs. Well, I guess it does, if the only thing you want to know is how RPI4 holds against cheapest desktop as an alternative to it, but besides that we are talking massive difference in power consumption, SKU price, form factor and software ecosystem here. My personal believe is that comparing RPI4 to Intel low power architecture based boards (atom-derived CPUs) would be more useful from practical point of view, because it would show if there is any meaningful reason to invest in somewhat similar (a lot closer at least) priced Intel based SBCs, like a UP board. That's my 2 cents on this article.

[ldesnogu]

Apple processors only use the instruction set - they designed their cpus from scratch, and didn't use the toy reference designs from ARM.

A76 and up have changed the direction ARM took. They now seriously target performance... at last! And this explains why companies such as Samsung and Qualcomm decided to switch to ARM designs rather making their own ARM CPU. But yeah Apple is still miles ahead (at least 1/1.5 year in advance).

[Slartifartblast]

Unfortunately the microarchitecture of the RPi4 is still a bit of a joke.

I've had a lengthy (and peppered with the occasional insults, of course) discussion earlier on phoronix forums.

It's not just the clock speed. The IPC of the RPi4 is mediocre.

The RPi4 chip (Cortex A72) doesn't have a proper Load Store Queue, to allow it to issue loads out-of-order with respect to earlier stores (so all prior stores need to complete before issuing a load). This would require machinery to detect and correct potential misspeculations, of course.

It can only decode 3 instructions per cycle, and issue 5.

Actually the venerable Alpha 21264 from 1996 had better microarchitecture. Granted, that was a trailblazer chip, but still ... 1996!

So, this is not about nm. It's about a toy microarchitecture.

Let's be realistic here, we are paying toy prices. It's built for a price and not for sheer performance. You want the fastest ARM then you are more than welcome to pay Apple prices and good luck breaking out of their garden.

[carewolf]

Is it really?
ARM is an architecture that is supposed to scale up into the high-performance too.

The Raspi and embedded devices in general are designed to be efficient, but there are high-end ARM CPUs that really need a real heatsink.

Yeah, old ARM32 has some limitations and made it less than ideal, but no worse than the crap i386 was hampered with, but AArch64 is better designed. And instruction sets doesn't matter that much in the end with high end processors with decoding, queuing and out-of-order execution.