Recently created Python 3.9.2 (M1 optimised) + MySQL test and run on two machines, M1 Mac mini and 2020 iMac i7-10700K. Task was to compare 2 table data, 18.5M records to 10M records.
1 core: Mac mini - 140h, iMac - 170h
4 cores: Mac mini - 36h, iMac - didn't test it
8 cores: Mac mini - 27h, iMac - 26h
16 threads: iMac - 27h
iMac's Intel i7-10700K hit max temp in first seconds of test even it was showing only 105W/50% cpu usage . In 16 thread test cpu 105W/99% usage, but no performance gains at all. On the other hand, M1 performed without a whisper, silently and cool. Apple silicons has an unbelievable potential.

This is idiotic. The AMD 5980 (anything in the 59 series) is specifically designed for multicore processing and has intentionally slower cores to manage the tdp.

Why didn't you compare the M1 to the 5800? Because it would have wiped the floor. So dishonest and the comments reflect that.

Quote from: Sumedh on March 05, 2021, 15:24:10
2) As per cpubenchmark.net, both Ryzen 5800U and 4800U have higher power efficiency than Apple M1. Go figure.

That's no mystery. They use what they call Max TDP to gauge efficiency. Confusingly, as it's not the maximum TDP that the processor can be configured to. Not to mention that using TDP is just stupid in the first place. And it happens to be 15 W for both 4800U and 5800U and 15.1 W for M1. It's a useless chart. A typical 4800U is not going to be consuming 15 W when running at full tilt.

Quote from: Astar on March 05, 2021, 22:54:58
Absolutely hilarious how CrApple Sheeple like you lot can't understand all the "native" Geekbench crap results that CrApple and lackey websites like notebookchat keep using. Really how stupid can you Sheep be? Pathetic!

I recall Anandtech defending GeekBench, despite them using SPEC in their analysis. It placed similarly. And CB R23 results are where I would expect them based on that article.

It is remarkable that Apple can play with the best x86 cores. But the truly astonishing feat is the pace they have been keeping. And it doesn't look like their architecture is running out of breath.

Quote from: nicholai on March 05, 2021, 17:50:35
These are just single-thread performance benchmarks too, right?
So it's half an AMD core vs a full M1 core and AMD still comes out on top.

You can't say it's half a core. The core supports two threads, but they share resources. That one thread can still make use of practically everything in the core. What SMT does is improve utilization of resources (a thread is not using everything all the time). The benefit depends on workload. It's like with two humans sharing a workspace. Sometimes, you get along fine. And sometimes, you constantly wait on each other, needing the same tools. You'll never get double the performance. Even 30 % is a very good result in the real world. As I wrote, enabling SMT can actually slow some programs down. But it's cheap to implement.

Dumb benchmarking. I love how most ppl miss the complete fact that when u benchmark hardware using an app running on different OSs you completely skewed the results right there. 

Simple rule of benchmarking: the variable can only be one. Otherwise you'll get variance. So if you're benching a GPU...the runs have to be identical, u have to use identical cases, identical drivers (or at least best drivers for competing brands which in themselves can cause variance), identical API (DX 11 vs DX12 vs Vulkan etc), same mobo and RAM and Windows version even and you even should for most accurate results use the same PC case. 


You're benching the M1 no doubt on the super lightweight incapable OS known as MacOS and comparing it to Intel and AMD CPUs benched on Windows? Lmao completely incomparable results right there. Put the Intel or AMD CPU in the MacBook, it'll rip the M1 to shreds. Or put the M1 in a Windows PC, poor thing will choke and die.

SMH

Quote from: StephenJ on March 05, 2021, 21:48:28
Absolutely love the hate apple is getting on these articles. Look at these internet heroes trying to call out apple. True fanboys who circle jerk anything outside of apple. Pathetic and hilarious to read  ;D

Absolutely hilarious how CrApple Sheeple like you lot can't understand all the "native" Geekbench crap results that CrApple and lackey websites like notebookchat keep using. Really how stupid can you Sheep be? Pathetic!

Absolutely love the hate apple is getting on these articles. Look at these internet heroes trying to call out apple. True fanboys who circle jerk anything outside of apple. Pathetic and hilarious to read  ;D

What a idiot is author o this text. M1 beats ryzen in single core for 7%. Wow! But looses for 64% in multi core. Missleading title of this article.

These are just single-thread performance benchmarks too, right?
So it's half an AMD core vs a full M1 core and AMD still comes out on top.

Pure clickbait. Headline contradicts is own findings. Sigh, how low can notebookcheck fall. Sure M1 is a very impressive CPU but does it really require such lies to market itself?

To all those tauting the power efficiency, two points:

1) Headline and content should have focused on that rather than these lies

2) As per cpubenchmark.net, both Ryzen 5800U and 4800U have higher power efficiency than Apple M1. Go figure.

Quote from: Don Donaldson on March 05, 2021, 00:03:18
With the M1 having onboard RAM, the TPD isn't a fair comparison unless one also takes into account the RAM, right?

You should never compare TDP between manufacturers. Because it means different things to different manufacturers. Each determines it in a different way. It can have nothing to do with power consumption whatsoever. Even though by laws of physics as we know them, what come in as electricity leaves as heat (or sound or light). Really, it's a pretty useless number. Not even cooler manufacturers like it. And it was originally intended to facilitate cooler selection.

Quote from: Lin Baden on March 04, 2021, 21:19:12
A cpu's performance isn't measured just by the score results you see in benchmarks, but also from the "results" of power draw, that's also an aspect of the performance.
...
Most of the time you use your laptop (not for gaming or heavy-work stuff), only 1 or a couple of cores will be active, that's why single-core performance is super important.

Energy consumption factors into efficiency. It certainly doesn't factor into absolute performance. They're actually opposed to each other as there are diminishing returns at play. You can increase performance at the cost of efficiency. Of course, mobile chips should in general run in the efficient range.

That's nonsense. Single core performance is important because a lot of software is limited by the performance of a single core. You can add as many cores as you want and it won't run any faster. If you want it to run faster, you need a faster core. If you're doing something light, it doesn't matter because you're not fully utilizing even that one core. And since there is no one heavy thread, you can easily spread the work among many cores. The reason not to spread light load around too much is so that as many cores can sleep as possible which is going to be more power efficient. Up to a point - you don't want to "overload" a core as efficiency is frequency dependant. The sweet spot for efficiency is typically far from maximum performance point.

I am a big AMD fan but you have to give it to Apple when you compare performance per watt. MTBF has to come into play at some point. If AMD is diving tens of watts to get decent performance, the die temp HAS to go up (throttling protects the silicon from destructive temps). With the M1 having onboard RAM, the TPD isn't a fair comparison unless one also takes into account the RAM, right? I think what Apple has done is impressive given the comparison to two large processor companies that have been designing and manufacturing silicon for decades longer.

TDP (putting aside that it's not the same as power limit) is typically not a limiting factor when it comes to single core performance. Even just 10 W is quite a lot for a single core. Of course, there is more than just a core running.

QuoteOn the other hand, M1 is able to outclass the Core i7-1185G7 by 20% despite featuring only four high performance cores.

And how many cores does the 1185 have? The M1 has four additional cores. The impressive part is that it can produce those scores while being passively cooled. That's truly impressive. Managing to post competitive single threaded scores is also remarkable for an ARM processor. Scoring 20 % more when you have four extra cores (meaning twice as much), albeit of lower performance, isn't in itself impressive. Icestorm cores aren't that weak. So, M1 performs like an exceptional four core processor. It can't bear comparison with good eight core processors. Except for single core/ thread tests which eliminate the difference. But then you shouldn't be surprised it does well.

That brings me to another detail. Single core tests are often single thread tests. Meaning that cores supporting SMT can't take any advantage of that. That's why MP ratios can be higher than core count. Firestorm cores don't support SMT. It's fair if you want to see single threaded performance. But you need to be careful if you want to compare performance of cores. If a core supports multiple hardware threads, you might need multiple threads to reach peak performance, to realize the full potential of that core. In something like Cinebench, SMT can easily make up 30 %.

This also explains why is M1 "only" 20 % ahead of a 1185. I believe Icestorm cores should perform at around 50 % of Firestorm cores. So, M1 should roughly correspond to a 6 core processor with no SMT. If you have a 4 core processor with SMT and you count with 30 % gain from SMT, it's going to be equivalent to 5.2 cores. +20 % and you're at 6.2. The right ballpark. A processor like the 1185 can also be though of as having a kind of 4+4 configuration. The problem with SMT is that the result is very much workload dependent. The impact can even be negative.