Adobe Releases Lightroom for Apple M1 and Windows Arm, Adds Apple ProRAW Support

[Bombastinator]

1 hour ago, comander said:

also iPads and the iPads were getting close to 4C Intel parts in geekbench (flawed benchmark but it wasn't a blow out).

I’m not saying you’re wrong.  I’m just pointing out perceptions.  Tablet doesn’t really help with that a tablet processor isn’t considered generally to be at the level of a laptop processor whether it actually is or not.

[tech.guru]

7 hours ago, Bombastinator said:

I’m not saying you’re wrong.  I’m just pointing out perceptions.  Tablet doesn’t really help with that a tablet processor isn’t considered generally to be at the level of a laptop processor whether it actually is or not.

people perceptrons are skewed by the information they read.

everyone wants to belong to a "cult" nowadays.  some people in this "cult" the "true believers" we will call them take it upon themselves to scour the internet and share misinformation and distorted reality and convince others to join the same "cult".

 

I see so much misinformation and artificial narratives being constructed with an agenda to promote a product. Often conveniently,.down-playing or omitting product weaknesses entirely.

 

its so annoying to those who are actually care about technology and do not want to want to belong to a "cult".

 

think critically people!... buy the best product for your needs

[Blade of Grass]

14 hours ago, tech.guru said:

the m1 laptop doesnt even support 4k display, but  2560-by-1600 (ref https://support.apple.com/kb/SP824?locale=en_CA)

Quote

Video Support 

Simultaneously supports full native resolution on the built-in display at millions of colors and: 

• One external display with up to 6K resolution at 60Hz

?

[Spindel]

1 hour ago, Blade of Grass said:

?

When your brand hatred runs so deep you just make things up in support of your argument. 

[tech.guru]

4 hours ago, Blade of Grass said:

?

external monitor yes, but built in display is not that was my point when i quoted the resolution in the builtin display.   you did read the entire statement about tests being about what people will be using in real life scenario.

 

How many people have 5 grand to buy a 6K external monitor to play 4K youtube content? 

even better yet how many are going to play videos in AV1 on youtube when theres hardily any videos out there?

 

and lets use this scenario as a basis to judge the overall performance of a product. for an use case almost zeo percentage of people will ever use?

 

keep on stretching.... you missed the point

[Blade of Grass]

4 minutes ago, tech.guru said:

external monitor yes, but built in display is not that was my point when i quoted the resolution in the builtin display.   you did read the entire statement about tests being about what people will be using in real life scenario.

 

How many people have 5 grand to buy a 6K external monitor to play 4K youtube content? 

even better yet how many are going to play videos in AV1 on youtube when theres hardily any videos out there?

 

and lets use this scenario as a basis to judge the overall performance of a product. for a use case what percentage of people will ever use? lets keep on stretching.... you missed the point

Well let's be clear, that's not what you said, but perhaps it is what you meant. 

 

Who said they need $5k for a 6k monitor, 4k monitors--which you were speaking about--are much cheaper and widely available. 

Lots of people buy high-res monitors to connect to their laptops, lots of companies deploy laptops to users and give them high res monitors for them to connect to.

 

I'm not sure why you're so focused on YouTube here, you can watch 4k on Netflix using Safari on macOS 11, and I know they've been rolling out AV1 this year (but, I am not sure if the browser version uses it). People watch more than just YouTube on their computers. 

[tech.guru]

3 hours ago, Blade of Grass said:

Well let's be clear, that's not what you said, but perhaps it is what you meant. 

 

Who said they need $5k for a 6k monitor, 4k monitors--which you were speaking about--are much cheaper and widely available. 

Lots of people buy high-res monitors to connect to their laptops, lots of companies deploy laptops to users and give them high res monitors for them to connect to.

 

I'm not sure why you're so focused on YouTube here, you can watch 4k on Netflix using Safari on macOS 11, and I know they've been rolling out AV1 this year (but, I am not sure if the browser version uses it). People watch more than just YouTube on their computers. 

the benchmark is decoding 4k AV1 content on youtube....... he posted not me

im lucked out no one at my company bought me a 4k monitor to watch youtube at home.

[f22luke]

On 12/10/2020 at 1:31 PM, Lord Vile said:

POWER houses is the right word. The M1 runs 4 performance cores at less power than a Ryzen 3rd gen chip takes to run one. 

 

 

On 12/10/2020 at 1:50 PM, leadeater said:

No it doesn't, that is factually not correct.

 

I'm surprised you guys haven't mentioned process nodes or the special memory configuration used on M1.  AMD will be on 5nm soon so they will then have parity with Apple in fact it's probably nice for AMD to now have Apple running pipe cleaners for them.  It always feels like nodes take a bit of work getting them tuned for high performance since they always get small mobile parts run through them first. Now that Apple is running performance parts through the node it just means probably less work for TSMC to get AMD running.

 

As far as memory goes, AMD is limited to DDR4L power draws and can't help that but it does sound like they are aware of it and have some creative solutions of their own coming.  They are already more power efficient than their main rival so incremental steps are the correct way to approach these new technologies.  AMD did beat the drum on HSA for a long time and they still have that tech and the realizable BAR is also and example of memory performance enhancements coming out from them so I'm sure we will see more from them in the coming architectures. 

 

The future is bright for both AMD x86-64 and ARM, and with chip-lets it would be neat to see a hybrid chip maybe a big little that could have a mixture of cores but maybe the software devs will v-toe that idea.  Still lets not argue about stuff like this and just enjoy it when it gets here.

 

 

[thepoorgamer]

 I think ARM powered pc sales will probably pass x86 ones in this decade. Personally, I am fine with that. As long as ARM chips have broad software support and better performance then x86 chips, I would have no qualms getting a ARM powered PC. This is why it is so important that developers release ARM optivised software, so the power of these chips can be utilised by these applications without emulation. I think it will take a few years, but once software is good for ARM machines, I imagine they will become much more popular. I think gamers will probably be the last x86 holdout (we don't care if our chips draw a lot of power) but I think the industry will eventually transition away from x86 chips and to ARM or RISK V.  

[Commodus]

21 minutes ago, thepoorgamer said:

 I think ARM powered pc sales will probably pass x86 ones in this decade. Personally, I am fine with that. As long as ARM chips have broad software support and better performance then x86 chips, I would have no qualms getting a ARM powered PC. This is why it is so important that developers release ARM optivised software, so the power of these chips can be utilised by these applications without emulation. I think it will take a few years, but once software is good for ARM machines, I imagine they will become much more popular. I think gamers will probably be the last x86 holdout (we don't care if our chips draw a lot of power) but I think the industry will eventually transition away from x86 chips and to ARM or RISK V.  

The question is whether or not Windows successfully makes that leap. Qualcomm's current chips don't come close to rivalling Apple's, and Microsoft hasn't had much success either fostering ARM apps or offering robust x86 emulation.

 

It still feels a bit like this is Microsoft's Steve-Ballmer-dismissing-the-iPhone moment for the '20s — that moment where the company doesn't do enough and misses out on a major transition. Like in 2007, the impact might not be noticeable right away, but the landscape could be very different in a few years.

[Lord Vile]

15 hours ago, f22luke said:

 

 

 

I'm surprised you guys haven't mentioned process nodes or the special memory configuration used on M1.  AMD will be on 5nm soon so they will then have parity with Apple in fact it's probably nice for AMD to now have Apple running pipe cleaners for them.  It always feels like nodes take a bit of work getting them tuned for high performance since they always get small mobile parts run through them first. Now that Apple is running performance parts through the node it just means probably less work for TSMC to get AMD running.

A die shrink may do very little.

[leadeater]

3 hours ago, Lord Vile said:

A die shrink may do very little.

A die shrink may or may not but the IPC increase and clock increase of Zen 3 when offered in a mobile product will bring the single thread performance in many applications to parity of the M1. After that it'll just be application and architectural differences that better fits as to which will be faster and why, for a lot of Apple first party applications and any that leverage in the same way application performance of the M1 will be higher due to the unified memory model and better integration with GPU acceleration (which we have already seen).

[GDRRiley]

On 12/10/2020 at 12:29 PM, Vitamanic said:

Well, Ryzen has lost its performance advantage now by quite a massive margin. When these crazy M1 chips start rolling out with 12, 16, 32 and 64 cores, Ryzen will be getting spanked in performance 5 times harder than AMD spanked Intel.

they didn't. We haven't seen ryzen APUs yet which if my math is right will be as fast as M1 in single core with 2x the big cores and 4x the threads.

Zen3 is a match for M1. if they don't bring some SMT then the 12core m1 is going to be beat by a 8 core zen3. 16 will be beat by the 12 core zen3. 32 and 64 will likely require thread ripper to beat them.

[Lord Vile]

6 minutes ago, leadeater said:

A die shrink may or may not but the IPC increase and clock increase of Zen 3 when offered in a mobile product will bring the single thread performance in many applications to parity of the M1. After that it'll just be application and architectural differences that better fits as to which will be faster and why, for a lot of Apple first party applications and any that leverage in the same way application performance of the M1 will be higher due to the unified memory model and better integration with GPU acceleration (which we have already seen).

You can't really say that though. Zen 2 is a long way behind the M1 in ST though even the 4900HS is still around 20% behind the M1 and even using Zen 3's gains from the desktop side they'll still lag behind the M1

[leadeater]

2 minutes ago, Lord Vile said:

You can't really say that though. Zen 2 is a long way behind the M1 in ST though even the 4900HS is still around 20% behind the M1 and even using Zen 3's gains from the desktop side they'll still lag behind the M1

Yes I can, go to the single core benchmarks and apply the 20% uplift that we know to exist for Zen 3 and you'll see just how close it is to The Apple M1. Only geekbench has significantly higher single thread performance and we don't need to rehash why that is not particularly representative, neither is CB solely for that matter either.

 

The single thread performance on the mobile CPUs will be very similar to the desktop CPUs as the single core power required for that is under the short term power limit of the mobile CPUs.

[Sorenson]

On 12/10/2020 at 2:05 PM, leadeater said:

No it is not correct, been down this before. You don't have to pick the 5950X to do this. How about pick a laptop/mobile part with the same or even previous generation Zen 2 core. To say Zen 3 requires more power for a single core as does the entire M1 is not at all correct.

 

So no it is wrong, and no performance and power scaling is not linear and we already know the mobile parts at significantly lower per core power are very very close to the desktop high power parts for performance in both single and multicore performance.

 

But yes lets just pick the worst case possible to make an incorrect point, happens ALL THE TIME.

 

Edit:

And FYI Apple is more than capable of also increasing the the power targets on the Firestorm cores just the same making performance per watt worse. Can we apply just a little bit more critical thinking. I would also not be surprised to see Apple do this on even more performance targeted desktop class devices, there is no reason Apple has to limit themselves to the current single core performance they have right now and they can achieve higher if they want to.

 

Having to counter this kind of senseless rhetoric is exactly why people think and accuse me of not liking the M1, which is far from the truth. This goes back to my point about not acting like the M1 is the seconding coming of tech Christ. For what ever reason something simply being very good isn't enough for people and have to make wild claims and statements supported by fundamental falsehoods.

The comment was comparing zen3, which does not have a mobile part, to the m1. Also, you should not compare the Mac mini, a desktop, to an ultra low power variant of the zen 2 cpu. That is disingenuous to the original point that was trying to be made, which was, the M1 provides a huge amount of performance at a low TDP. You could downclock and under-volt the zen 3 chip, but then you would be sacrificing performance. 

[leadeater]

18 minutes ago, Sorenson said:

The comment was comparing zen3, which does not have a mobile part, to the m1

The issue is, and has been in the past, what is being tried to be implied with that statement and the continued bias towards M1 and the courtesy of that only shown towards that of the M1 and ignoring the design differences of a desktop CPU to a mobile CPU. As I explained in a later post. It does not matter if a mobile CPU does not exist now they will exist and when you state Zen 3, not a specific CPU model, you are talking about the Zen 3 architecture itself which can be implemented in to many different products in different ways. So it is entirely disingenuous to be make those kinds of statements or claims, we have historical data and evidence to disprove them anyway which I did.

 

18 minutes ago, Sorenson said:

Also, you should not compare the Mac mini, a desktop, to an ultra low power variant of the zen 2 cpu. That is disingenuous to the original point that was trying to be made, which was, the M1 provides a huge amount of performance at a low TDP

I was not comparing to desktop CPUs, he was. That is the problem itself and it is his point that is disingenuous not my rebuttal to his.

 

18 minutes ago, Sorenson said:

You could downclock and under-volt the zen 3 chip, but then you would be sacrificing performance. 

We already know just how well Ryzen Mobile 4000 U and H/HS series performs and in the beloved single thread performance M1 supporters have latched on to this is the closest to the desktop CPU in performance. 4800U (15W/25W) ~475/~480, 4900HS CB 1T is 490-495 and 3900X CB 1T is 500-505, Zen 3 mobile CPUs will not be any different.

 

So it doesn't matter that it will lower than the desktop CPUs, the single thread performance is not as significant as people want it to be and also isn't the reason alone for why the M1 is good.

 

My only wish is that people stop beating the drum of irrelevance.

[Lord Vile]

30 minutes ago, leadeater said:

Yes I can, go to the single core benchmarks and apply the 20% uplift that we know to exist for Zen 3 and you'll see just how close it is to The Apple M1. Only geekbench has significantly higher single thread performance and we don't need to rehash why that is not particularly representative, neither is CB solely for that matter either.

Go on then, show the benchmarks 

30 minutes ago, leadeater said:

 

The single thread performance on the mobile CPUs will be very similar to the desktop CPUs as the single core power required for that is under the short term power limit of the mobile CPUs.

Again show it.

[leadeater]

30 minutes ago, Lord Vile said:

Go on then, show the benchmarks 

Again show it.

I already have, In this topic, in that post, in other posts in this topic and in other topics. Your problem is you just keep ignoring them so what point is there to post them yet again to be ignored yet again?

 

But to extend a single olive branch I'll post again what I already have. Cinebench R23 1T scores:

• 3900X: 1342
• 4900H: 1284
• 4900HS: 1246
• 4900U: 1279
• 5600X: 1572
• 4900HS + 20%: 1495 (HS used as U is abnormally high)
• M1: 1514

[Blade of Grass]

41 minutes ago, leadeater said:

We already know just how well Ryzen Mobile 4000 U and H/HS series performs and in the beloved single thread performance M1 supporters have latched on to this is the closest to the desktop CPU in performance. 4800U (15W/25W) ~475/~480, 4900HS CB 1T is 490-495 and 3900X CB 1T is 500-505, Zen 3 mobile CPUs will not be any different.

 

So it doesn't matter that it will lower than the desktop CPUs, the single thread performance is not as significant as people want it to be and also isn't the reason alone for why the M1 is good.

 

My only wish is that people stop beating the drum of irrelevance.

Just want to go on a bit of a tangent though, but I think the dialog of single vs. multi threaded performance is generally pretty bad, with people on both sides making... less than optimal statements about it. Realistically, people need to stop trying to paint something as broad as all of the computing tasks into two categories that seem to be either 1) only single threaded and 2) massively parallel workloads. Obviously those are easy to talk about a help the masses make decisions, but they really muddy the waters. 

 

Benchmarking should be purpose-driven. If your purpose is to figure out how fast it is for normal user-done tasks, then it should benchmark doing that in a representative way. There is no one-size-fits-all, but for some reason the public dialog always tries to make it sound like there is. A system that's fast at thing A really might not be fast at thing B, even if they seem similar (and a corollary to that is, the system might be fast at both things, but without benchmarking how can you tell )

 

Funnily enough, I found a benchmark recently that covers a lot of stuff relevant to me (programming tasks), and it reinforces this point well. The M1 and the 3900x trade blows on tasks, sometimes one system is significantly faster, other times the other is, and sometimes they're close and trade places (like SQL read vs. SQL seq read). 

 

https://tech.ssut.me/apple-m1-chip-benchmarks-focused-on-the-real-world-programming/

 

Anyway, my point is, lots of workloads are not as simple as "single threaded" or "multi-threaded", nor do they necessarily scale well if they're multi-threaded. Some tasks have quirks and see significant benefits from things another task may not, and only the user's purpose can deduce that. (an example we had at work was that AMD EPYC CPUs were slower at doing large RandomForrest inferencing, because the size of our Forrests were too large to fit into L1 cache, so the Intel CPUs were faster)

Conclusions should be drawn from relevant data, extrapolation of performance generally is bad. 

 

Also this isn't specifically at you lead eater, you just said something relevant about performance

[Lord Vile]

11 minutes ago, leadeater said:

I already have, In this topic, in that post, in other posts in this topic and in other topics. Your problem is you just keep ignoring them so what point is there to post them yet again to be ignored yet again?

 

But to extend a single olive branch I'll post again what I already have. Cinebench R23 1T scores:

• 3900X: 1342
• 4900H: 1284
• 4900U: 1246
• 5600X: 1572
• 4900U + 20%: 1495
• M1: 1514

But where are you getting 20% from? The 3900X to the 5600X which are the only 2 chips that are on the same platform (Desktop) and are a generation apart and the difference between them is 17% not 20. So you're theoretical 4900U score would be 1458 at 150% the power of the M1.

[leadeater]

15 minutes ago, Blade of Grass said:

Funnily enough, I found a benchmark recently that covers a lot of stuff relevant to me (programming tasks), and it reinforces this point well. The M1 and the 3900x trade blows on tasks, sometimes one system is significantly faster, other times the other is, and sometimes they're close and trade places (like SQL read vs. SQL seq read). 

 

https://tech.ssut.me/apple-m1-chip-benchmarks-focused-on-the-real-world-programming/

Oooo nice, that an impressive amount of tests in there. Looks like we should move our Redis instances to M1 lol.

[Blade of Grass]

6 minutes ago, leadeater said:

Oooo nice, that an impressive amount of tests in there. Looks like we should move our Redis instances to M1 lol.

Heh, I think the reason why (for that one) is because Redis is memory bound, and the M1 has really fast memory. 

 

I did think about a little Mac mini DC though, as a joke... power efficiency tho...

[leadeater]

30 minutes ago, Lord Vile said:

But where are you getting 20% from? The 3900X to the 5600X which are the only 2 chips that are on the same platform (Desktop) and are a generation apart and the difference between them is 17% not 20. So you're theoretical 4900U score would be 1458 at 150% the power of the M1.

From the average performance increases seen during the reviews and the stated 19% IPC uplift and slight clock increases. Also it would be better to use the same CPU model in the lineup rather than picking different ones, I did for that 5600X on purpose as it was the slowest and closest to the M1.

 

[edit] As an aside the 5900X is 1622 which is 21% faster [/edit]

 

It's not like the 1%-3% differences we are talking about is going to make any appreciable difference at all and as I explained is not the reason why the M1 is good. For application performance like in Final Cut Pro that is a direct result of Apples persistent efforts of software optimization and utilization of the unified memory model in the M1 and superior GPU acceleration integration, and that is true for the other render applications and other software that can make use of the GPU in the M1. The M1 also has very fast memory and low latency with very high usable per core bandwidth so applications and workloads that make use of memory performance do much better than Zen 3 as comparatively it has a much weaker memory subsystem.

 

30 minutes ago, Lord Vile said:

So you're theoretical 4900U score would be 1458 at 150% the power of the M1.

Yes and that is fine, I only dispute your specific claims when you imply it is greatly higher than this or say that the mobile CPUs use a minimum of 15W etc which are incorrect.

 

What I see is you thinking that in me correcting things said about competing products that I am saying the M1 isn't all the things it is, that is an entirely different thing all together.

[Lord Vile]

13 minutes ago, leadeater said:

From the average performance increases seen during the reviews and the stated 19% IPC uplift and slight clock increases. Also it would be better to use the same CPU model in the lineup rather than picking different ones, I did for that 5600X on purpose as it was the slowest and closest to the M1.

 

[edit] As an aside the 5900X is 1622 which is 21% faster [/edit]

According to reviews I've seen the difference between the 3950X and 5950X is around 15%

13 minutes ago, leadeater said:

 

It's not like the 1%-3% differences we are talking about is going to make any appreciable difference at all and as I explained is not the reason why the M1 is good. For application performance like in Final Cut Pro that is a direct result of Apples persistent efforts of software optimization and utilization of the unified memory model in the M1 and superior GPU acceleration integration, and that is true for the other render applications and other software that can make use of the GPU in the M1. The M1 also has very fast memory and low latency with very high usable per core bandwidth so applications and workloads that make use of memory performance do much better than Zen 3 as comparatively it has a much weaker memory subsystem.

 

Yes and that is fine, I only dispute your specific claims when you imply it is greatly higher than this or say that the mobile CPUs use a minimum of 15W etc which are incorrect.

You're trying to dispute a Zen 3 chip that is a 105W chip uses less power per core than the M1 uses for the entire SoC in which you are still wrong. 

13 minutes ago, leadeater said:

 

What I see is you thinking that in me correcting things said about competing products that I am saying the M1 isn't all the things it is, that is an entirely different thing all together.

No you're just trying to make an argument by putting words in peoples mouths.