m1 Apple M1 = the rest of us are living in the stone age!?

[Lord Vile]

15 hours ago, GDRRiley said:

your understanding of engineering samples is so dam broken. its for 3 things, show how this chip is doing, program bios and other low level code, making sure your hardware design works.

wtf do you think apple gets when they chose to stick in a 1067G6 or whatever. they get samples of those chips 3-6 months beforehand.

Yeah 3-6 months not 18. 

[gal-m]

19 hours ago, Riccardo Cagnasso said:

 

 

ditto

Holy cow. I have to watch this..

[GDRRiley]

14 hours ago, Lord Vile said:

Yeah 3-6 months not 18. 

what the hell makes you think I mentioned anything or tried to say something about 18 months?

[Lord Vile]

12 hours ago, GDRRiley said:

what the hell makes you think I mentioned anything or tried to say something about 18 months?

 

Based on product timelines? 

[gal-m]

On 12/7/2020 at 9:03 PM, Riccardo Cagnasso said:

I won't be surprised to see a 256 core mac pro in the near future

I'm REALLY interested in seeing what's going to be happening with desktop class Apple computers in the future..

 

Does anyone have any idea about when we could be expecting new Apple Silicon equipped Mac Pros?

[Ar558a]

Apple could make a 1024 core 2048 thread 100GHz all core boost CPU for $1 and I wouldn't buy it....

[gal-m]

3 minutes ago, Ar558a said:

Apple could make a 1024 core 2048 thread 100GHz all core boost CPU for $1 and I wouldn't buy it

How come?

[Ar558a]

3 hours ago, gal-m said:

How come?

I'd rather die than give that evil company a single cent.

[Video Beagle]

On 1/18/2021 at 2:07 PM, Ar558a said:

Apple could make a 1024 core 2048 thread 100GHz all core boost CPU for $1 and I wouldn't buy it....

 

On 1/18/2021 at 5:15 PM, Ar558a said:

I'd rather die than give that evil company a single cent.

Kewl story, bro.

[LAwLz]

@GDRRiley @leadeater

I haven't read the thread but I thought this would be the right place to post in. I remember you both telling me M1 wasn't that great because AMD could match it with Zen3.

 

Well, yesterday Anandtech did a review of the Zen3 based 5980HS and wouldn't you know it... It is an improvement over the 4000 series but not enough to really change any conclusions.

AMD Ryzen 9 5980HS Cezanne Review: Ryzen 5000 Mobile Tested (anandtech.com)

 

AMD's highest end offering:

8 cores,  16 threads.

35 watt TDP.

Power consumption for the CPU alone = 35-42 watts (35 watts for sustained loads, 42 watt mode is capped at 5 minutes of sustained load).

 

Apple's lowest end offering:

4 big performance cores and 4 little efficiency cores.

22 watt TDP for the entire chip, including memory.

Power consumption for the entire computer when measured at the wall (including things like fans, memory, AC to DC conversion efficiency loss, storage etc)  = ~25 watts

 

 

Geekbench single-core:

M1 - 1745

5980HS - 1506

Difference: M1 is ~16% faster.

 

Geekbench multi-core:

M1 - 7715

5980HS - 8391

Difference: M1 is ~8% slower

 

SPEC2006 geomean total:

M1 - 79.4

5980HS - 61.3

Difference: M1 is ~30% faster

 

 

I hope Anandtech runs more benchmarks but all in all, it seems like AMD still can't compete on both fronts at the same time. If AMD wants to win at performance they have to use far more power. If AMD wants to win at power efficiency they will lose hard on performance.

 

 

In before people start calling Anandtech paid Apple shills, or start criticizing benchmarks and only wanting to look at benchmarks that shows the results they want to see.

It's worth pointing out that there isn't a 20% performance uplift going from zen2 to zen3 on mobile. The uplift varies greatly depending on application. Haven't looked very closely at it but it seems to be around 15% on average.

 

The 5980HS seems great for an x86 processor, but no match for Apple's M1 in terms of efficiency or single core performance.

[leadeater]

2 hours ago, LAwLz said:

I haven't read the thread but I thought this would be the right place to post in. I remember you both telling me M1 wasn't that great because AMD could match it with Zen3.

Neither of us said it wasn't that great, I know I for sure said the M1 chip just isn't actually as groundbreaking as people are saying or want to believe, bit of a difference. Far as what I said Zen3 would increase single thread performance to match the M1 and that is exactly what has happened. Are there tests which each one is faster than the other, of course.

 

2 hours ago, LAwLz said:

It's worth pointing out that there isn't a 20% performance uplift going from zen2 to zen3 on mobile. The uplift varies greatly depending on application. Haven't looked very closely at it but it seems to be around 15% on average.

The uplift is basically single thread only, as the power limits during multithread keep the performance the same between Zen2 and Zen3 at these lower power targets. I would hazard a guess the increased IPC is accompanied by higher power, which is not unexpected at the same process node, which is directly counteracting that performance gain which is why there is hardly a difference at all for multithread between Ryzen Mobile 4000 and 5000, maybe this is not the case on lower end lower core count CPUs but I've only seen the 5980HS reviewed myself.

[LAwLz]

1 hour ago, leadeater said:

Far as what I said Zen3 would increase single thread performance to match the M1 and that is exactly what has happened.

No it hasn't.

Zen3 loses quite a lot to the M1, even without factoring in power consumption. Once you factor that in the M1 is a lot better.

The M1 is ~20% ahead in terms of single core performance in all benchmarks we have right now. That's not "matching" it.

Like I said in the other threads, the only time Zen3 can match Firestorm in terms of performance is when it is running at way higher frequency, and at that point it consumes waaay more power, like in the desktop chips.

 

This is exactly why I kept comparing it to desktop processors before.

Zen3 just barely performs better than Firestorm if you run the Zen3 cores at around 5GHz. You kept saying that I shouldn't compare it to the desktop processors because they are running outside of their peak efficiency and therefore it was unfair.

But you can't have it both ways. Either you think Zen3 is as fast as Firestorm, but the only time that is true is when it is clocked really high and therefore not efficient. If you want to argue Zen3 can be just as efficient as Firestorm then you have to clock it very low and at that point it doesn't perform nearly as well as Firestorm.

 

You can't have it both ways, which was my entire point. With Zen3 you can either match performance, or you can go for efficiency. You can't have both. It's either equal in perform of performance but uses way more power, or it is slower but uses the same amount of power.

That goes for multicore store as well. Since Firestorm is so much ahead of Zen3 in terms of performance (at lower power targets), you need far more Zen3 cores to be able to match it for multithreaded workloads. As we can see in these benchmarks, what's essentially an Apple quad core is goes head to head with an octa core Zen3 processor.

 

How someone can say that isn't fantastic and groundbreaking is beyond me.

[Jeppes]

Just now, LAwLz said:

No it hasn't.

Zen3 loses quite a lot to the M1, even without factoring in power consumption. Once you factor that in the M1 is a lot better.

The M1 is ~20% ahead in terms of single core performance in all benchmarks we have right now. That's not "matching" it.

Like I said in the other threads, the only time Zen3 can match Firestorm in terms of performance is when it is running at way higher frequency, and at that point it consumes waaay more power, like in the desktop chips.

Until M1 is running at higher clocks and with more cores, it is what it is. Great for laptops if you dont need a gpu for anything real.

[RTX 3090]

On 12/2/2020 at 5:37 PM, Lord Vile said:

This is the same spec~ with no OS:

 

On 12/2/2020 at 5:37 PM, Lord Vile said:

RX 560

 

On 12/2/2020 at 5:37 PM, Lord Vile said:

£1912.01

You can get a significantly better PC for that price. 

On 12/2/2020 at 5:37 PM, Lord Vile said:

LG 31MU97Z-B 31.0" 4096x2160 60 Hz Monitor £1179.00

Why sped £1000 on a monitor tho. Most of the money is being wasted on the monitor, not PC

On 12/2/2020 at 5:51 PM, Lord Vile said:

legally

 

[Ankh Tech]

2 minutes ago, LAwLz said:

No it hasn't.

Zen3 loses quite a lot to the M1, even without factoring in power consumption. Once you factor that in the M1 is a lot better.

The M1 is ~20% ahead in terms of single core performance in all benchmarks we have right now. That's not "matching" it.

Like I said in the other threads, the only time Zen3 can match Firestorm in terms of performance is when it is running at way higher frequency, and at that point it consumes waaay more power, like in the desktop chips.

Because it was optimised for low power only, it's performance gains on higher power aren't linear, the perf/watt decreases the higher the power, the entire archotecture, or arm, is made for low power operation, if zen 3 was entirely based on low power, it would crush the m1

[Loote]

36 minutes ago, LAwLz said:

Zen3 loses quite a lot to the M1, even without factoring in power consumption. Once you factor that in the M1 is a lot better.

The M1 is ~20% ahead in terms of single core performance in all benchmarks we have right now. That's not "matching" it.(...)

I am also not sure how much better Apple's design is than Zen3.
There is no doubt M1 is way better than even newest AMD product in laptops, but...
1. It seems designed exactly for this purpose, Zen seems to be affected by being made with 8/16 core desktop designs, with plenty of IO etc. That's why double cores for AMD are nowhere near double energy cost etc.
My point is that I am not sure how well will Apple scale it, is it going to be just as impressive, or a bit less, I don't doubt it's going to be great anyway.
2. Some of the difference can be attributed to process node, just a quick search returns TSMC's 5nm technology is 15% faster with 30% lower power than 7nm. Though this is a moot point because I fully expect Apple to be a node or two in front for the foreseeable future.

We'll see, although I'd prefer Apple to use less of TSMC's capacity, M1 is a revolutionary thing and if others copy some of their improvements, it should be positive for everyone in the long term. The improvements for a new architecture should be big too.

[LAwLz]

1 hour ago, Jeppes said:

Until M1 is running at higher clocks and with more cores, it is what it is. Great for laptops if you dont need a gpu for anything real.

Not sure what you mean by "anything real" but yeah, the M1 is most certainly a laptop chip. I am very much looking forward to see how Apple will scale it up to the desktop.

 

 

 

1 hour ago, Ankh Tech said:

Because it was optimised for low power only

What was optimized for low power exactly?

 

1 hour ago, Ankh Tech said:

it's performance gains on higher power aren't linear, the perf/watt decreases the higher the power,

Of course, but how it scales remains to be seen. So far we have seen it scale really well from the A14 to the M1.

But I am not sure what that has to do with anything. I am comparing two laptop chips here.

 

 

1 hour ago, Ankh Tech said:

the entire archotecture, or arm, is made for low power operation

Ehm... Where did you get this idea from? It is completely wrong.

ARM is not designed for low power operations. The CPU architectures ARM licenses out have traditionally been designed for low power cores, but Firestorm, which is Apple's own architecture, is designed for high performance. Like I said earlier, it matches a desktop Zen3 processor in terms of performance, per core, on average. 

Something like Phoenix, which is also ARM compatible, is expected to land around 40% higher IPC than Zen2 and is most certainly not designed for low power operations either. Amazon's Graviton2 chips most likely consume around 100 watts of power, and Ampere's N1 chip consumes around 200 watts.

The idea that ARM is for low power consumption, low performance CPUs is a myth that needs to die. It's just not true anymore, at all.

 

 

1 hour ago, Ankh Tech said:

if zen 3 was entirely based on low power, it would crush the m1

Ehm, what? 

1) Crush it how exactly? 

2) What do you base that massive assumption on?

3) Why does it even matter? You can't just make up an imaginary product and go "well if AMD did make this then it would be better". What kind of argument is that?

[LAwLz]

1 hour ago, Loote said:

It seems designed exactly for this purpose, Zen seems to be affected by being made with 8/16 core desktop designs, with plenty of IO etc. That's why double cores for AMD are nowhere near double energy cost etc.

If you are trying to make the argument that "AMD has more IO and that's why it uses more power" then I would like to inform you that that's a massive assumption @leadeater made and has 0 evidence to support it right now.

 

 

1 hour ago, Loote said:

My point is that I am not sure how well will Apple scale it, is it going to be just as impressive, or a bit less, I don't doubt it's going to be great anyway.

Absolutely. Completely agree with that. It remains to be seen how well it will scale to higher end chips. I don't think there will be much of an issue scaling it up to be honest. I don't see why it would be.

 

 

1 hour ago, Loote said:

2. Some of the difference can be attributed to process node, just a quick search returns TSMC's 5nm technology is 15% faster with 30% lower power than 7nm.

I think you misread that. N5 offers 15% speed improvements OR 30% power reduction, not both at once.

So if the 5980HS was made on N5 instead of N7 then it would either use 30% less power or AMD could have made it 15% faster (assuming the TSMC marketing numbers are actually correct).

Even if we were to assume the real world benefits are exactly as good as TSMC describes them, and imagine a 5nm 5980HS, the AMD chip would still consume more power alone than an entire Mac Mini, and perform about ~15% worse.

 

 

1 hour ago, Loote said:

We'll see, although I'd prefer Apple to use less of TSMC's capacity, M1 is a revolutionary thing and if others copy some of their improvements, it should be positive for everyone in the long term. The improvements for a new architecture should be big too.

I'd love to have an ARM laptop (and in the future, an ARM desktop). But it seems like Windows and Qualcomm have way too many issues right now and I don't think they will succeed in fixing them.

My guess is that x86 is on its death bed, but it will be a very slow and drawn out death.

The server market will slowly but steadily shift towards ARM, especially as companies like AWS and Azure invests into it. Legacy stuff like terminal servers will remain x86 until they die and gets replaced.

Desktop and laptops will become more and more rare as people switch over to smartphones as their primary device. As more things become web applications I think we will move towards "browser based thin clients" like ChromeOS and Windows 10X.

[GDRRiley]

3 hours ago, Loote said:

I am also not sure how much better Apple's design is than Zen3.

2. Some of the difference can be attributed to process node, just a quick search returns TSMC's 5nm technology is 15% faster with 30% lower power than 7nm. Though this is a moot point because I fully expect Apple to be a node or two in front for the foreseeable future.

it is worse and you just prove it right there.
M1 being on 5nm is whats keeping it in the race. if it was on 7nm or someone other than TSMCs node zen3 would be beating it. Zen would easily be ahead by around 10% and m1 would draw more power.
it would have been so good for capacity if Zen3 APUs were on 5nm right now.

[Lord Bloobus]

New 5nm process optimized for SFF sips power and crams more performance than a 7nm in the same chassis, not sure why this comes as a surprise. You're just buying something more expensive now that will be behind in a few months when others get their 5nm silicon batches to play with.

[NunoLava1998]

Honestly, regardless of if it's really much better than other chips in terms of performance, the M1 chips still do very well when it comes to performance/watt (Zen 3/4 chips will probably take years to get to the M1's point when it comes to perf/watt), it's iGPU does very well (1050~1050 Ti performance is absolutely insane for an iGPU; I'm not sure why so many are overlooking that), and it moves on from x86 to ARM. I think it's overall a step in the right direction for Apple.

[leadeater]

6 hours ago, LAwLz said:

Zen3 loses quite a lot to the M1, even without factoring in power consumption. Once you factor that in the M1 is a lot better.

The M1 is ~20% ahead in terms of single core performance in all benchmarks we have right now. That's not "matching" it.

I've said it before and I'll say it again, Geekbench is the most garbage "benchmark suite" that has always and continues to extremely badly translate across to actual real application performance. If you use the actual few application tests to compare them there is very little between the two for single thread performance.

 

Is this a problem with Mac OS platform itself, sure and that's not your fault, it is however yours to actually take any meaningful credit of Geekbench.

 

And no I'm not discrediting it because the M1 is a decent bit faster in it, I'm discrediting it because it's had this problem well before M1 was a thing. Was crap back then and it's still crap now.

 

And can we please stop acting like single thread performance is the only thing the matters, zero people buying a 5980HS or any other mobile or non mobile high core count CPU is not buying it for single thread performance.

 

The M1 is the best lower power CPU there is currently, problem is simply being good or the best for it's target use case isn't enough for some people 

 

6 hours ago, LAwLz said:

How someone can say that isn't fantastic and groundbreaking is beyond me.

Simple because it isn't. It's exactly no better than what I would expect from a low power ARM design on TSMC 5nm using a large and wide core architecture that targets single thread performance.

 

Because that's just the thing, if Intel or AMD wanted to greatly widen their cores to achieve the same thing they could, they don't because it doesn't lend very well when considering high core count scaling and some of the draw backs of x86 itself.

 

6 hours ago, LAwLz said:

But you can't have it both ways. Either you think Zen3 is as fast as Firestorm, but the only time that is true is when it is clocked really high and therefore not efficient. If you want to argue Zen3 can be just as efficient as Firestorm then you have to clock it very low and at that point it doesn't perform nearly as well as Firestorm.

And nobody has said otherwise. That's the benefit both Intel and AMD have, the fact that they have dynamic clocks and even more so for AMD the power management on Zen allows it to cater for a very broad range of workloads giving the best performance it can. The fact that as core count and core utilization goes up the clocks goes down doesn't matter as this reduction is non linear and as we have already seen and still see multithread performance is higher.

[leadeater]

3 hours ago, LAwLz said:

If you are trying to make the argument that "AMD has more IO and that's why it uses more power" then I would like to inform you that that's a massive assumption @leadeater made and has 0 evidence to support it right now.

False, I gave you the evidence, seems you just ignored it because you simply didn't want to believe it. You can go back to Anandtech and look at the per core power scaling across clocks again if you like, if you still don't believe it then your denying reality. They even have the IOD power nicely separated for you.

 

Then you can cross reference those figures with APUs and mobile parts that use a monolithic die and contain less IO and don't have to maintain IF link between chiplets and see that in this application power is indeed much lower.

[Donut417]

On 11/26/2020 at 9:07 AM, gal-m said:

I've really been impressed with the Apple M1 chip, though it seemed too good to be true at first, but after researching it extensively it looks like the claims actually do hold up.

 

My question is, what's next? Can Apple Silicon wipe everything else out? I've recently been thinking about building a new high end gaming system, but now it just seems like Apple might speed past anything in the years to come (in reference to single thread performance at least). The whole thing is making me extremely discouraged to invest in any type of Windows machine at the moment... - Yes I DO know that Windows and Macs, don't compare directly depending on a persons specific workload, but I think you get what I'm trying to say.. it's just making x86 CPUs feel a bit old ...

 

EDIT: I am NOT saying I want to play games on a Mac. I am just simply stating that the potential of Apple's new chips to wipe every other Intel or AMD off the face of the planet might move manufacturers like Intel or AMD to start looking at developing an ARM based chip as well and the potential that would have in a Windows machine..

 

Anyways, what do you guys think?

Well firstly its highly unlikely Apple will sell its chips to third parties. So all we can hope for is that this pushes Qualcomm, Samsung, Nvidia, etc to put more money in to R&D and make a chip that is just as good or better. The other side of the coin is Microsoft. One of the things that has helped make Apple's transition successful is Rossetta 2, while Microsoft can emulate X86 apps on ARM its not as good as what Apple has achieved, from what I have read. Microsoft needs to step up its game. 

 

On a second note, we dont know what innovations AMD and Intel have for us in the future. That being said, I think that if Apple can be successful and other ARM chip makers see this, it could have potential to directly compete with Intel and AMD. Assuming Microsoft can do better. To me where ARM would shine in the PC space is Laptops. Due to its power efficiency and the fact companies like Qualcomm could potentially build 4G and 5G radios in to the device. 

 

I think once we get to see benchmarks of Apples desktop line up its going to give us an idea of what we can expect in the future as well. As some have mentioned, ARM could very well be the future. But that day isn't here yet and I think it will be years before we really see a push. 

[gal-m]

4 hours ago, Donut417 said:

I think once we get to see benchmarks of Apples desktop line up its going to give us an idea of what we can expect in the future as well.

I'm really looking forward to that day @Donut417!

Really interested in Microsofts response if Apple Desktop Silicon turns out to actually be good, which I believe it will.