Apple dumps Intel and shows Final Cut Pro on new Mac with Apple CPU !

[joema]

...Is there any firm evidence about how delicate/how much stress our electronic babies can take?

It is shipped from the factory in the product case which is inside a brown cardboard shipping case. The brown case is molded to the angular product case, with about 1/2" or less foam between them. For factory product cases not intended for reuse, they are well designed. But it's unlikely they are individually air-shipped, rather are palletized in a more secure fashion. At the depot level they are probably de-palletized, then individually shipped, and UPS or whoever drops it at your door.

I wouldn't air-ship an iMac 27 in the original boxes, but I've used those boxes and crammed mine into a large truck full of A/V equipment many times, and it survived cross-country trips. I make sure it's not positioned on the bottom of an equipment stack.

I've heard of people air shipping an iMac 27 in the original boxes and they survived, also heard of people air-shipping them in the soft-sided Gator case and they survived. The Gator case is well padded but I wouldn't air ship mine in that -- that's why I have the hard-sided SKB case.

When you order an iMac 27 or iMac Pro for delivery to your house, they are not carried on a little satin pillow. I think they are fairly robust. OTOH if it arrives from Apple damaged, it's not your problem. If you ship it and it gets damaged (depending on your insurance) it might be your problem. The Gator case and especially the SKB case are very durable and protective.

 

[stewhem]

Good points, Joema. Thanks for your reply. Like the image of the satin pillow delivery! And there's the crux: if Apple deliver it damaged, no problem; if I cause it to be damaged, my expensive problem. I'll look into hard case options but will probablyl stick with taking the MBP on my travels. It's okay for downloading and quick assembly edits, which is all I usually have time for when working away from base.

 

[say_doyster]

Thanks jbregar! My eyes are not what they used to be - I often miss things in plain sight. Appreciate your info very much.

B

 

[markfpv]

I wouldn't air-ship an iMac 27 in the original boxes, but I've used those boxes and crammed mine into a large truck full of A/V equipment many times, and it survived cross-country trips. I make sure it's not positioned on the bottom of an equipment stack.
.

Totally concur with this - and have also taken mine with me to edit on location - like joema & PD. Save that box in comes in when you purchase it! So hard to go back to less screen real estate once you become accustomed to having 27" or more.

 

[Thomas Smet]

I used to haul my 27" around in a iLugger case. I Don't know if they make them anymore but it worked really well. Allowed me to have the better desktop to work on but also use it on location for live video streaming events.

 

[Thomas Smet]

https://www.ilugger.com/product/detail.cfm?nid=1460&pid=40029

Guess they are still around. Mine didn't have wheels so its even better now.

 

[Thomas Smet]

Haha I know. Based on this unknown, in your opinion is the m1 air too risky to use purely for high speed data transfers?

After having a PC micro super computer in an online shopping cart, I'm now ditching this plan and strongly leaning towards waiting for a 16" MBP as a main computer.

Would it be ridiculous having an air/13" for transfers/on set/quick work then a 16" for semi-permanent editing that can go anywhere if needed?

I'd hate to delay finishing a project mid-edit because I spilled my iced latte on the main computer when offloading footage.

Personally I wouldn't get a 13" and a 16". Just not enough difference to justify them both. I would rather get a 13" and a stronger/cheaper desktop like a iMac. The iMac will be faster and an overall better machine than the 16". In the field I don't think I would miss the performance as much. I don't tend to do full edit projects in the field. Mostly to transfer footage, check footage, organize media, rough cuts and maybe some quick conversions. The 16" would provide faster conversions if you really must convert ASAP to handoff or upload for an eager client.

The 13" is a fully capable machine in the field/hotel to even do a full edit as long as you are ok waiting a bit longer to render. Render overnight or while the crew is out to dinner/drinks. If you render overnight it likely will not matter much if you have the 13" or 16". Both would likely be done in the morning.

The 16" to me is if someone would only ever have one Mac. One computer to do it. Then they should get the 16" to also act as a desktop.

I wanted the 13" for years as a light portable machine to take on flights. Sometimes airline seats are a bit too cramped for even a 15" screen. Tried to sweet talk into a free business class upgrade to have more space or swap with someone that had an empty seat next to them but its still tighter than I would like. The 13" always sucked for just about everything until now.

Put it this way the M1 GPU is not desktop class but its faster than the 560X which until until 2019 was the highest GPU the 16" MBP could get. First there was the build to order Vega 20 that was super expensive and now the 5600m that is insanely expensive and most didn't do those options. The majority of 16" MBPs out there editing video from 2016 until now are actually slightly slower than the M1 GPU. So if people could edit in the field on those they definitely can on the M1. The M1 on the MBP is also within spitting distance of the 5300M which is the current base GPU on the 16". We also know the 5500M is only slightly faster than the 5300M so the M1 likely holds it own very well compared to almost every current 16" MBP except for the astronomical priced 5600M GPU option.

Also really depends on the software used. FCPX really kicks butt on the M1. Resolve kind of sucks for 6k Braw but 4k Braw seems to be ok. Definitely nowhere as good as 4k Braw is on my Vega 56 eGPU but usable. Thats what proxy is for however.

 

[filmguy123]

I'd go with the 16"... on the go productivity to me is also deeply tied to screen real estate and I find I work much better with a few extra inches of space and the ability to crank up the resolution without needing a magnifying glass as well. YMMV. The extra footprint of the 16" has in the past also allowed for higher specs; we'll see if the ARM situation changes that but I imagine a 16" will be capable of more. I hate cramped flights as much as anyone but I've always found a way to make it work with the larger screen.

I am still waiting for more substantial tests and benchmarks, and I think we'll have to wait until Resolve and Adobe Premiere are fully ready for ARM prime time to get meaningful comparisons. A lot of the comparisons I see are putting the M1 chips up against mobile Intel chips or iMacs. Which is completely 100% fair as these are mobile machines priced extremely competitively. We know that mobile against mobile, ARM is dominating and that's enough to make a purchase decision.

The academic curiosity, however, of how ARM is against real desktop workstations is something we may have to wait for. I really want to know how a 5950x will fair against an M1 in Premiere/Resolve once they are ARM optimized, and as well how the upcoming M1x will do. Next year will be very interesting. Right now, based on those first Puget system benchmarks (in conjunction with other benchmarks that show the M1 to be only about 25% slower in Rosetta 2), it looks like a very, very large gap to close. It's one thing to compare the M1 chips to another Intel laptop or an iMac, and another thing to compare it to a true modern desktop.

I am cautiously optimistic about the ARM desktop platform compared to your tower workstations. I would be surprised if the upcoming M1X can outclass desktop workstations. There's currently a lot of hype about the ARM chips being a miracle chip, but right now all we know is that it is a miracle chip in the mobile domain. But will it be a miracle in the workstation domain? Apple says it will complete the ARM transition in 2 years, and this includes the MacPro. So, popcorn out. Apple has delivered thus far. But where will their bar be for workstations? Beating the existing Xeon won't be hard, as it was outdated and underwhelming from day 1. I assume they will outclass the Xeon workstations, or, match them at a lower price. But, will it be able to beat Ryzen 6000 series, the next gen Threadripper, and Intel's long await foray into 10nm desktops, all coming by late 2021? No idea. These chips have always been geared as low power chips so I eagerly await to see what Apple can do. They have the talent, money, and motivation.

Regardless, Apple is going to win here. The question is, in what way? As it's been said, there are no bad products, just bad prices. Even if the workstation class Apple ARM chips can't take top spot, Apple has the supply chain down... so Apple will either "win" on pure performance, or price/performance. I think it's safe to say (given Apple's track record) that they won't win on both on the high end pro side of things, unless they take a change of strategy (and the backlash of outclassing a $50,000 MacPro a couple years later at half the price or less). We'll see.

 

[Thomas Smet]

I'd go with the 16"... on the go productivity to me is also deeply tied to screen real estate and I find I work much better with a few extra inches of space and the ability to crank up the resolution without needing a magnifying glass as well. YMMV. The extra footprint of the 16" has in the past also allowed for higher specs; we'll see if the ARM situation changes that but I imagine a 16" will be capable of more. I hate cramped flights as much as anyone but I've always found a way to make it work with the larger screen.

I am still waiting for more substantial tests and benchmarks, and I think we'll have to wait until Resolve and Adobe Premiere are fully ready for ARM prime time to get meaningful comparisons. A lot of the comparisons I see are putting the M1 chips up against mobile Intel chips or iMacs. Which is completely 100% fair as these are mobile machines priced extremely competitively. We know that mobile against mobile, ARM is dominating and that's enough to make a purchase decision.

The academic curiosity, however, of how ARM is against real desktop workstations is something we may have to wait for. I really want to know how a 5950x will fair against an M1 in Premiere/Resolve once they are ARM optimized, and as well how the upcoming M1x will do. Next year will be very interesting. Right now, based on those first Puget system benchmarks (in conjunction with other benchmarks that show the M1 to be only about 25% slower in Rosetta 2), it looks like a very, very large gap to close. It's one thing to compare the M1 chips to another Intel laptop or an iMac, and another thing to compare it to a true modern desktop.

I am cautiously optimistic about the ARM desktop platform compared to your tower workstations. I would be surprised if the upcoming M1X can outclass desktop workstations. There's currently a lot of hype about the ARM chips being a miracle chip, but right now all we know is that it is a miracle chip in the mobile domain. But will it be a miracle in the workstation domain? Apple says it will complete the ARM transition in 2 years, and this includes the MacPro. So, popcorn out. Apple has delivered thus far. But where will their bar be for workstations? Beating the existing Xeon won't be hard, as it was outdated and underwhelming from day 1. I assume they will outclass the Xeon workstations, or, match them at a lower price. But, will it be able to beat Ryzen 6000 series, the next gen Threadripper, and Intel's long await foray into 10nm desktops, all coming by late 2021? No idea. These chips have always been geared as low power chips so I eagerly await to see what Apple can do. They have the talent, money, and motivation.

Regardless, Apple is going to win here. The question is, in what way? As it's been said, there are no bad products, just bad prices. Even if the workstation class Apple ARM chips can't take top spot, Apple has the supply chain down... so Apple will either "win" on pure performance, or price/performance. I think it's safe to say that they won't win on both, at least on the high end side of things.

Yeah but to be fair these are laptops and realistically should only be compared to other laptops. It isn't fair to really compare any laptop to a higher end desktop. Something gets compromised when shrinking down to portability and battery power.

Desktops may hold an advantage for some time but then again they are tethered to a desk, monitor, keyboard and mouse. They use an insane amount of electricity compared to a M1 Mac and they tend to make a ton of noise unless one uses liquid cooling.

Apple also tends to ignore the desktop segment. They jump from a iMac all the way up to a workstation class system like the Mac Pro with nothing in between. It was always unrealistic to compare Apple to desktop PCs. The iMac was typically underpowered compared to even a modest gaming or edit system and the Mac Pro was massive overkill for that segment.

The latest iMac with 8 cores and the 5700 XT is finally somewhat competitive but still not the same as a $1,500 PC with 8 cores and a 3070 GPU. Sure it may not include the 5k display but I'm perfectly fine with my 27" 4k 100% p3 color 10 bit factory calibrated LG monitor for $320.

Now if apple pulls off 4x the performance of the rumored 20 core (16 performance cores and 4 efficiency cores) then the iMac will be the desktop to beat. On the CPU side the M1 already beats pretty much every i3 and i5 desktop CPU and some i7's. Even if Apple only doubles the cores we are getting into to performance equal to 12 core Intel CPUs which is very competitive in the desktop market. If Apple does 4x the CPU speed thats getting very close to the top Threadripper territory.

What every rumor seems to completely ignore right now however is what the M1X or M2 means for GPU performance. The M1 is already insanely fast for a low power CPU. The 16" and iMac could use the M1 as is for CPU and they would still be impressive machines. Its the GPU that is lacking right now. Even doubling what the M1 GPU can do falls far short of desktop Nvidia GPUs or the 5700 XT. Apple really needs to 4x the GPU performance and if they do that with a tiny amount of power used that will be a game changer for sure. Imagine having a desktop with 3080 performance with a 200 watt power supply all in a small form factor or all in one like the iMac.

 

[filmguy123]

I agree Thomas. Hopefully I made that clear by calling that an academic curiosity, and calling the M1 a miracle chip in the mobile domain, and my comment "Which is completely 100% fair as these are mobile machines priced extremely competitively. We know that mobile against mobile, ARM is dominating and that's enough to make a purchase decision." Apologies if that was misconstured.

The question of how ARM will scale at the higher end of the power spectrum is of course extremely interesting to all of us, especially those contemplating desktop workstation purchases. And, there's been a fair amount of inductive reasoning as to how the chips may scale when juiced up.

All that said, despite just purchasing a rather expensive desktop, I would be personally excited to see a close to linear scaling here. While it would feel a bit flustering to have a 16" MBP perform the same as my new burly desktop, more logically speaking, it would be a huge win for me to have no real drawback to working mobile, and a huge win for all of us to see such substantial competition in the computer space from which we would all benefit. But, I am skeptical of that and would caution anyone riding the hype train to remember these ARM chips have always targeted low power thresholds and even Apple themselves has showed drastically diminishing returns on their power/performance curves (at least for the M1 chips - we'll see what happens with M1x, etc)

On the GPU front, though, I am actually pretty optimistic. Nvidia is ARM based, so there is no reason to think Apple can't produce an incredibly powerful ARM based GPU. I don't know if they can pummel nvidia based on performance per watt though, but it should be theoretically possible to produce an extremely competitive GPU for their desktop workstations. Especially since Nvidia has spent a lot of energy on gaming related features like DLSS and RayTracing, where Apple will be most focused on workstation based tasks.

As far as comparing Apple to desktop PCs being unfair, I suppose that is true based on Apple's goals and product structure, but from a user point of view, it's the only pragmatic thing to do. It's our responsibility as pros to find the best solutions for our business at various price points and trade-offs. At this juncture, I think it's pretty clear for me that my next mobile machine is going to be an ARM based Macbook Pro. The desktop workstation world is something I will watch with great curiosity.

 

[Tamerlin]

On the GPU front, though, I am actually pretty optimistic. Nvidia is ARM based, so there is no reason to think Apple can't produce an incredibly powerful ARM based GPU.

Since when is nVidia ARM based? Are you drawing that conclusion based on the fact that nVidia has an ARM license?

nVidia's GPUs have next to no realationship to ARM, other than being integrated into some SOCs that include ARM processors. Its GPU design inherits a lot more from SGI's than it does from ARM.

I don't know if they can pummel nvidia based on performance per watt though, but it should be theoretically possible to produce an extremely competitive GPU for their desktop workstations. Especially since Nvidia has spent a lot of energy on gaming related features like DLSS and RayTracing, where Apple will be most focused on workstation based tasks.

I guess you've forgotten about the Quadro line, and the A100 monsters... and the fact that nVidia debuted its tensor hardware in Volta, of which there were no consumer variants.

As far as comparing Apple to desktop PCs being unfair, I suppose that is true based on Apple's goals and product structure, but from a user point of view, it's the only pragmatic thing to do. It's our responsibility as pros to find the best solutions for our business at various price points and trade-offs. At this juncture, I think it's pretty clear for me that my next mobile machine is going to be an ARM based Macbook Pro. The desktop workstation world is something I will watch with great curiosity.

Agreed. I'm probably going to end up with a hybrid setup, using an M1 or successor for mobility and a Zen based system for heavy lifting.

 

[rob norton]

Personally I wouldn't get a 13" and a 16". Just not enough difference to justify them both. I would rather get a 13" and a stronger/cheaper desktop like a iMac. The iMac will be faster and an overall better machine than the 16". In the field I don't think I would miss the performance as much. I don't tend to do full edit projects in the field. Mostly to transfer footage, check footage, organize media, rough cuts and maybe some quick conversions. The 16" would provide faster conversions if you really must convert ASAP to handoff or upload for an eager client.

The 13" is a fully capable machine in the field/hotel to even do a full edit as long as you are ok waiting a bit longer to render. Render overnight or while the crew is out to dinner/drinks. If you render overnight it likely will not matter much if you have the 13" or 16". Both would likely be done in the morning.

The 16" to me is if someone would only ever have one Mac. One computer to do it. Then they should get the 16" to also act as a desktop.

I wanted the 13" for years as a light portable machine to take on flights. Sometimes airline seats are a bit too cramped for even a 15" screen. Tried to sweet talk into a free business class upgrade to have more space or swap with someone that had an empty seat next to them but its still tighter than I would like. The 13" always sucked for just about everything until now.

Put it this way the M1 GPU is not desktop class but its faster than the 560X which until until 2019 was the highest GPU the 16" MBP could get. First there was the build to order Vega 20 that was super expensive and now the 5600m that is insanely expensive and most didn't do those options. The majority of 16" MBPs out there editing video from 2016 until now are actually slightly slower than the M1 GPU. So if people could edit in the field on those they definitely can on the M1. The M1 on the MBP is also within spitting distance of the 5300M which is the current base GPU on the 16". We also know the 5500M is only slightly faster than the 5300M so the M1 likely holds it own very well compared to almost every current 16" MBP except for the astronomical priced 5600M GPU option.

Also really depends on the software used. FCPX really kicks butt on the M1. Resolve kind of sucks for 6k Braw but 4k Braw seems to be ok. Definitely nowhere as good as 4k Braw is on my Vega 56 eGPU but usable. Thats what proxy is for however.

Thanks Thomas. What are your thoughts on an air/16" combo?

 

[rob norton]

My laptop is a 27" iMac if the place I have to edit is within driving distance. It's surprisingly easy to bungie a couple of towels around the body and lay it in the trunk to
go anywhere in Southern California. Editing on a single 16" screen, ugh. Editing on a 27" 5k Retina, nice. Of course if I had to fly somwhere to edit, the 16" all the way.
Luckily I havent had to do that for years. I have one NYC agency who always hires me to edit on site at trade shows like E3, back when we had trade shows. Just edited
an EPK package for Macys and Make a Wish Foundation in Orange County. Was a breeze to take my 27 iMac with me and even my 16TB RAID.

Thanks Dan. I think I'd be like Clermond and go with a separate monitor into the 16".

I'd like to get an atomos sumo at some point so maybe that can perform double duty on set/editing.

 

[filmguy123]

Since when is nVidia ARM based? Are you drawing that conclusion based on the fact that nVidia has an ARM license?

I didn't mean to imply that the GTX/RTX series were ARM based, rather that Nvidia itself as an entity has taken an ever increasing interest in ARM for its future plans, so it clearly sees the potential in platform and seems to be basing a sizeable part its future on ARM.

https://www.zdnet.com/article/nvidia-arms-itself-to-terminate-intels-processor-dominance/
https://nvidianews.nvidia.com/news/...-enabling-new-path-to-exascale-supercomputing
https://www.theverge.com/2020/9/13/21435507/nvidia-acquiring-arm-40-billion-chips-ai-deal

I guess you've forgotten about the Quadro line, and the A100 monsters... and the fact that nVidia debuted its tensor hardware in Volta, of which there were no consumer variants.

No, how so? My statement was based entirely around whether or not Apple could pummel nvidia based on performance per watt, which I think is unlikely. But, I do think it is possible because Nvidia's big focus in recent years has been on Raytracing and DLSS, with far more modest advancements seen in rasterization and workstation performance. This can be clearly demonstrated in benchmarks, where the new Radeon cards are actually superior to Nvidia in pure rasterization when you remove RayTracing performance and DLSS (both gaming features). As well, the Radeons beat Nvidia in workstation tasks that are not optimized for CUDA. Unfortunately for AMD, most things are optimized for CUDA. Point is, NVIDIA has not invested as much effort or seen as much improvement in its workstation advancements in the last couple years. The Quadro and Titan cards are now outclassed by an RTX 3090 in all but just a few particularly niche use cases; and the RTX 3090 is outclassed by Radeon if you removed RayTracing, DLSS, and the innate advantage of industry optimization for CUDA over OpenCL - at least on the PC side (Apple platform is optimized around OpenCL and Metal however).

This is not a knock on nvidia's workstation performance; I purchased the RTX 3090 for my new edit rig. It is a statement that Nvidia's recent efforts have clearly been focused heavily on new technologies like RayTracing and machine learning, and not on improving pure compute numbers either for gaming cards or for workstation cards. They have stated this themselves. Last week, nvidia sent out a letter to them complaining to a review site that they were focusing too much on Nvidia's pure compute horsepower rather new features. This complaint was leveled after this site clearly showed how the RTX 3000 series was in some instances only a mere 35% faster than Pascal from 2 generations back and 3+ years ago. Beyond that, Nvidias new cards are also extremely power hungry.

My statement was merely to say I think there is an opportunity here for Apple to produce a more competitive OpenCL based workstation card and beat Nvidia on both performance AND on performance per watt, at least vs anything that Nvidia has available right now on market. However, I doubt it would "pummel" them as nvidia is till quite competitive, but there is no reason to think an ARM based GPU win for Apple vs Nvidia is unfeasible given that Apple's efforts will be entirely dedicated to the workstation side of things and not gaming features or performance.

 

[Tamerlin]

I didn't mean to imply that the GTX/RTX series were ARM based, rather that Nvidia itself as an entity has taken an ever increasing interest in ARM for its future plans, so it clearly sees the potential in platform and seems to be basing a sizeable part its future on ARM.

nVidia's interest in ARM is most likely a result of not having a horse of consequence in the CPU race, which also means that it's largely left out of the SOC market, and it's buying AMD processors for its A100.

Meanwhile AMD has no reason to purchase nVidia GPUs for its supercomputer nodes, and now that Intel has Xe, it also has no reason to incorporate nVidia GPUs into products like Ponte Veccio.

No, how so? My statement was based entirely around whether or not Apple could pummel nvidia based on performance per watt, which I think is unlikely. But, I do think it is possible because Nvidia's big focus in recent years has been on Raytracing and DLSS, with far more modest advancements seen in rasterization and workstation performance.

Both have been very helpful for workstations, actually -- DLSS is a library that uses nVidia's tensor hardware, and CAD and 3D visualization, not to mention virtual production, remain nVidia dominated, partly because of the hardware ray tracing.

This can be clearly demonstrated in benchmarks, where the new Radeon cards are actually superior to Nvidia in pure rasterization when you remove RayTracing performance and DLSS (both gaming features).

That also actually demonstrates that AMD's Navi2 line are more gaming focused than the Ampere line is. nVidia placed a major emphasis on GPU compute along side ray tracing and tensor hardware.

I also suspect that nVidia, like pretty much the entire industry, underestimated where AMD would land in terms of performance. AMD's leaks implied far less performance than delivered, most likely by design. It does seem like the clock speeds that AMD is achieving with its GPUs is higher than even AMD expected, though.

As well, the Radeons beat Nvidia in workstation tasks that are not optimized for CUDA. Unfortunately for AMD, most things are optimized for CUDA. Point is, NVIDIA has not invested as much effort or seen as much improvement in its workstation advancements in the last couple years. The Quadro and Titan cards are now outclassed by an RTX 3090 in all but just a few particularly niche use cases;

Since the Ampere versions of the Quadro and Titan haven't launched yet, that's neither surprising nor unexpected, given how large the performance jump has been in Ampere from Turing.

This is not a knock on nvidia's workstation performance; I purchased the RTX 3090 for my new edit rig. It is a statement that Nvidia's recent efforts have clearly been focused heavily on new technologies like RayTracing and machine learning, and not on improving pure compute numbers either for gaming cards or for workstation cards.

Have you seen the compute resources nVidia built into Ampere, or its updated architecture? The pure compute throughput on Ampere is huge. What nVidia didn't emphasize is straight rasterization, because its theory is that by providing high levesl of GPU compute, ray tracing, and tensor performance it can deliver higher overall image quality in realtime rendering, but for that to happen requires optimizing for them, so the jury's still out on whether or not it's going to pan out.

Either way, AMD being competitive is a very good thing all around.

They have stated this themselves. Last week, nvidia sent out a letter to them complaining to a review site that they were focusing too much on Nvidia's pure compute horsepower rather new features. This complaint was leveled after this site clearly showed how the RTX 3000 series was in some instances only a mere 35% faster than Pascal from 2 generations back and 3+ years ago. Beyond that, Nvidias new cards are also extremely power hungry.

That's always been one of the challenges in using new features to boost performance rather than souping up existing features like AMD did. If AMD hadn't stepped up so well it wouldn't have made a difference.

Efficiency wise, Ampere is actually a big improvement over Turing, but nVidia is targeting performance, so instead of focusing lower power designs it focused on higher performance... and while 350 watts is a lot of power, when you consider that 350 watts is for a 24GB memory, over 10,000 CUDA cores + the ray tracing and tensor cores whose counts I don't remember off the top of my head, and the external interfaces, that's proportionally not such a big draw.

My statement was merely to say I think there is an opportunity here for Apple to produce a more competitive OpenCL based workstation card and beat Nvidia on both performance AND on performance per watt, at least vs anything that Nvidia has available right now on market. However, I doubt it would "pummel" them as nvidia is till quite competitive, but there is no reason to think an ARM based GPU win for Apple vs Nvidia is unfeasible given that Apple's efforts will be entirely dedicated to the workstation side of things and not gaming features or performance.

An ARM based GPU win is unfeasible. It wouldn't be able to scale. The fact that the M1's GPU is built into the SOC and that the CPU is an ARM doesn't imply that the GPU is also ARM based. Its predecessors are PowerVR based, though I'm not sure whether the M1 GPU is also PowerVR based or not. It might be based on whatever Apple's been working on internally, which I think is more likely.

Intel tried that approach, in fact... it tanked -- the cores in GPUs are far smaller, and their instruction sets more streamlined. In terms of raw floating point horsepower Intel's approach ended up working well enough to get it some supercomputer design wins, but it was a one trick pony, so it flopped elsewhere -- and got pummeled by nVidia's GPUs, then later by AMD's. And now also by Intel's own GPUs.

 

[filmguy123]

Yes, nvidia's interest in ARM is explicity stated as SOC, CPU, internet of things, mobile, and a few others... so far. Everything is changing rapidly, and I think we are all curious to see where this goes.

Nvidia's has supported hardware RayTracing for a long time now; it's recent efforts have not been into hardware raytracing - that's old - but rather into *real-time* hardware raytracing, with the clear focus on gaming. Yes, this will have collateral benefit in live previewing of raytracing as well... but that has not been their motivating fuel for development here at all. I also imagine it's not very important to those on this forum.

DLSS is machine learning to super sample lower resolution images to offset the performance loss of utilizing raytracing and is a gaming focused development. That's not making a statement on their Tensor cores, which are clearly more multipurposed than DLSS.

AMD's cards are pretty clearly gaming focused - just like Nvidia's RTX ampere cards. The takeaway is that AMD has focused more heavily on pure compute power than Ampere and has seen more impressive gains. That doesn't mean Ampere's gains or efficiency are bad, it just means AMDs are better. Nvidia has by their own admission been far more focused on DLSS and realtime raytracing. I think its a great call for gamers, and that nvidia is the easy pick this generation for either gaming or workstation given the value of DLSS/RT Raytracing, and industry CUDA support.

As for how this pertains to Quadro, it's not really about seeing an Ampere version of the quadro; it's about assessing the Ampere vs Turing vs Pascal compute performance increase in order to project what ampere based quadro performance will look like. We can do that reasonably well right now. In any case, the quadro is increasingly becoming relegated to extreme niche users as the scenarios in which it offers any meaningful benefit over RTX series for mainstream content creation tasks has increasingly diminished. 10-bit support, studio drivers, high VRAM have all come down to RTX now, which really only leaves reliability and compataibility certifications for unique applications - something which none of us here are going to be interested in. Yes, I am sure Quadro will see a bump to CUDA cores which would make them a bit faster for content creation, as well as more VRAM, but the price/performance increase will most certainly be the definiton of diminishing returns, and the pragmatic benefits for folks like us will be even more limited.

Point is, due to nvidia's heavy focus elsewhere on its current generations, and based on current assesments of Ampere performance capabilities per watt, and based on correlation of past GTX/RTX releases and Quadro performance, we already roughly know what to expect. As for the Titan, it's likely this card is now dead as of this generation due to how Titan-esque the RTX 3090 already is.

As for A100... errm, that's hardly of interest to mere mortals like us, isn't that $200k?

As far as an ARM based GPU not being able to scale... I've pretty consistenly expressed the same concern not just for the GPU but for the CPU. As recently as today. So, I'm with you there. And yet, we'll see, because it's all conjecture and like you, I don't know what I don't know. I don't think its unfeasible, we are also talking about mobile platforms here like the Macbook Pro. Apple is going all in, and a couple months ago no one expected Apple to pull off what it did with the M1 CPUs. Apple also has an advantage in controlling their OS and some key pro apps, so they very well may be able to realize gpu performance gains focused for content creation tasks that simply wouldn't transpose over to a Windows platform, and will outclass Nvidia's Ampere.

We'll see.

 

[Tamerlin]

Nvidia's has supported hardware RayTracing for a long time now; it's recent efforts have not been into hardware raytracing - that's old - but rather into *real-time* hardware raytracing, with the clear focus on gaming.

You're still wrong. nVidia's been working on making its ray tracing interactive for years, but using its general purpose compute cores. What changed in Turing is that nVidia developed hardware dedicated to ray tracing, so it's now possible to execute ray tracing functionality without taking up any of the general purpose compute cores.

Yes, this will have collateral benefit in live previewing of raytracing as well... but that has not been their motivating fuel for development here at all. I also imagine it's not very important to those on this forum.

The hardware accelerated Redcode decoding certainly does... and anyone who starts delving into virtual production will certainly be interested. There's bound to be some overlap.

DLSS is machine learning to super sample lower resolution images to offset the performance loss of utilizing raytracing and is a gaming focused development. That's not making a statement on their Tensor cores, which are clearly more multipurposed than DLSS.

Once again, DLSS is built on the tensor cores. Black Magic's Neural Engine uses the tensore cores when they're available also. And yet again, the whole point of having dedicated tensor cores is that software can use them without taking up compute resources. Without dedicated tensor cores, most ML implementations fall back to general purpose GPU compute and then CPU resources, if it can't use the GPU compute.

AMD's cards are pretty clearly gaming focused - just like Nvidia's RTX ampere cards. The takeaway is that AMD has focused more heavily on pure compute power than Ampere and has seen more impressive gains. That doesn't mean Ampere's gains or efficiency are bad, it just means AMDs are better. Nvidia has by their own admission been far more focused on DLSS and realtime raytracing. I think its a great call for gamers, and that nvidia is the easy pick this generation for either gaming or workstation given the value of DLSS/RT Raytracing, and industry CUDA support.

Years ago ATI and nVidia did this to each other quite a bit. One would start adding new features (usually nVidia), and the other would focus on some aspect (or aspects) of performance, and they'd leapfrog each other. It's nice to see that happening again.

Of course AMD's efficiency gains were better than nVidia's... nVidia went from good to great, AMD's went from mediocre to great.

As for how this pertains to Quadro, it's not really about seeing an Ampere version of the quadro; it's about assessing the Ampere vs Turing vs Pascal compute performance increase in order to project what ampere based quadro performance will look like. We can do that reasonably well right now.

And that would imply anywhere between 30% and 200% depending on what features the software uses.

In any case, the quadro is increasingly becoming relegated to extreme niche users as the scenarios in which it offers any meaningful benefit over RTX series for mainstream content creation tasks has increasingly diminished. 10-bit support, studio drivers, high VRAM have all come down to RTX now, which really only leaves reliability and compataibility certifications for unique applications - something which none of us here are going to be interested in. Yes, I am sure Quadro will see a bump to CUDA cores which would make them a bit faster for content creation, as well as more VRAM, but the price/performance increase will most certainly be the definiton of diminishing returns, and the pragmatic benefits for folks like us will be even more limited.

We know for sure that the bump from the Turing versions to the Ampere versions will be huge. How nVidia plans on differentiating them from the GeForce line is unclear.

Point is, due to nvidia's heavy focus elsewhere on its current generations, and based on current assesments of Ampere performance capabilities per watt, and based on correlation of past GTX/RTX releases and Quadro performance, we already roughly know what to expect. As for the Titan, it's likely this card is now dead as of this generation due to how Titan-esque the RTX 3090 already is.

It might be. Your guess is as good as mine.

As for A100... errm, that's hardly of interest to mere mortals like us, isn't that $200k?

I guess you missed the decades of history where technology pioneered in ultra high end products made its way to consumer products...

As far as an ARM based GPU not being able to scale... I've pretty consistenly expressed the same concern not just for the GPU but for the CPU.
As recently as today. So, I'm with you there. And yet, we'll see, because it's all conjecture and like you, I don't know what I don't know. I don't think its unfeasible, we are also talking about mobile platforms here like the Macbook Pro. Apple is going all in, and a couple months ago no one expected Apple to pull off what it did with the M1 CPUs.

I don't think that anyone has actually made an ARM based GPU. The one in the M1 most likely is not. Stuffing a GPU into a box with an ARM CPU doesn't magically turn it into an ARM based GPU. I suspect that it's based on PowerVR like its predecessors. It's clearly not a powerhouse, but apparently has some dedicated hardware codecs that are rather useful. Dedicated hardware of course sacrifices flexibility, but it allows CODEC processing without eating compute resources and consumes less power than a more flexible solution would.

Apple also has an advantage in controlling their OS and some key pro apps, so they very well may be able to realize gpu performance gains focused for content creation tasks that simply wouldn't transpose over to a Windows platform, and will outclass Nvidia's Ampere.

Apple's biggest advantages right now is having access to TSMC's most advanced node and not having to care about customers for its processors. (By which I mean 3rd parties building it into products.) Apple can therefore get away with making such an inflexible solution with its 8/16 GB of embedded memory... it's not even accurate to say that it's soldered on the motherboard, because it's not; it's essentially stuffed into the same box as the CPU and GPU and hard wired to each. That cuts out a lot of external I/O, which in turns reduces system power consumption, and also manufacturing cost. And obviously, lower cost is the best way to raise profits, so...

 

[filmguy123]

Lol, Tamerlin, what are you on about? There's a list of things here in which you're pushing back on that I've never said a thing about, and most of them are besides the point. This has gotten terribly derailed and off topic and I'm failing to see the reason?

As far as RT Raytracing and DLSS, I'm not sure what you mean by "I'm wrong" as there are numerous, clear statements from nvidia on their goals with RT raytracing and DLSS. I've never said or implied anything about their work on Tensor cores which power these functions not being good for workstation tasks and AI machine learning tasks. They are. As more developers adopt them - which will take time - they will be yet another great form of acceleration. While they are still in their infancy, I believe in them - it's why I just paid $1800 for an RTX 3090.

As far as Ampere Quadro performance, I am saying we already know what to expect from Ampere Quadro cards vs Ampere RTX based on how past architectures compare, and not to expect to see much more performance from them than what we've already seen today (and especially not at prices that make sense to any of us for the apps we use). Buying an Ampere Quadro card instead of an RTX card for anyone here would be largely pointless unless nvidia makes some unexpected moves that break from tradition.

Ampere is a great architecture. Sure, there are fringe cases where we see 200%, but there are also many cases where we see in real world use only 15%-25% better performance vs Turing (such as Adobe apps). For most content creators we might expect to see up to a 30%-50% boost vs Turing (i.e. Davinci Resolve). While that is a very solid performance boost, it could have been more if nvidia made different choices - though I am not implying that they should have. I think long term the investment in my new 5950x and RTX 3090 will pay off as developers better code for more cores/threads and take advantage of Tensor cores. Perhaps in the future this performance will increase with software updates. Yet I also think its realistic that by the time the Tensor cores are truly being utilized, we'll need a newer GPU to really harness what they can do. It's always a gamble to buy tomorrow's technology for performance today.

Yet, all of this is a rather large detour from where we started, which was -

there is no reason to think Apple can't produce an incredibly powerful ARM based GPU... I don't know if they can pummel nvidia based on performance per watt though, but it should be theoretically possible to produce an extremely competitive GPU for their desktop workstations....

I fully stand by my theorizing here. As for apple's gpu being based on PowerVR, I thought that stopped around the A10 or A11 and it was undisclosed after that. Everything I've heard them mention since this time has been under the umbrella of their ARM based SOC. Perhaps it is still PowerVR based, they aren't saying, though the point remains that there is no reason to think Apple cannot deliver on the GPU front in a way that is competitve with nvidia, and it may not be by playing by the traditional rules due to their hardware/software integration. Apple has already been playing Nvidia's Tensor core game under a different name long enough now - and Nvidia very clearly sees the power of ARM to deliver on its AI machine learning goals. Which is why they just spent $40 billion to acquire ARM.

Keep in mind that this also started with me saying that, like you, I am skeptical of Apple beating nvidia or AMD/Intel in a meaningful way on the high end desktop workstation anytime soon, which is precisely why I just invested in a new desktop PC workstation.

 

[Thomas Smet]

Thanks Thomas. What are your thoughts on an air/16" combo?

Same as before. The MBA and 13" MBP are more or less the same. Thats why most of the suggestions are to get the MBA. Its only rendering after ten minutes and longer that the MBA starts to slow down a bit due t the heating. Even then at most it may take a 20% hit. Depending on environment temperature as well of course. I live in AZ and if I tried to use a MBA outside in summer I bet it would slow down even more. Use it in February outside in WI and it may not slow down at all.

Most tasks I listed above like importing material, organizing media and doing a rough cut will not tax the system at all. Really only rendering for longer videos or playing heavier games that will start to slow the system down. If you tend to work on smaller projects where your exports take a few minutes you will likely never notice a speed difference between the MBA and 13" MBP.

I would personally get a MBA or 13" MBP and wait for the new iMac. You could get the Intel iMac which is still an amazing Mac with the 5700 XT and will continue to be one heck of a 4k video editing machine for many years to come. Plus desktops don't really need to worry about cooling and battery life the way laptops do. The M1 makes complete sense for why we use laptops. Unplugged portability. Apple may have to give the Apple Silicon iMac 4x the processing power to make the case for it over Intel and AMD. If its about the same then it will be a harder sell for a lot of users. Thats why it might be interesting to wait for the Apple Silicon iMac and try to get by with the M1 portable Mac for now.

When I say just 13" I also mean either the MBA or MBP. Both have a 13" screen. In fact identical screens other than the MBP having an extra 100 nits of brightness which apparently isn't a huge amount when viewing them side by side at full brightness. it may help a tad outdoors in bright sunlight but realistically if the sun makes viewing a 400 nit display impossible its likely not going to make a 500 nit any better. Both will suffer equally in harsh light. Might help a little bit if one insists on sitting in front of a window while they work but again its such a small visual difference I doubt one would be usable if the other is not.

The only reason I might get a MBP is because the MBA is next to impossible to find in stock now. 90% of reviewers suggest getting the MBA over the MBP and rightly so because it just makes sense for the majority of users. That makes it next to impossible to find and forget doing a custom order right now. Looking at end of January to get one. I can still sometimes find a M1 MBP in stock however although I may have to settle for a base 8GB ram version. I really wish Apple had a 16 GB option in stores so it wouldn't be outside of their normal supply chain. The wait times to get a 16 GB right now are insane.

 

[Thomas Smet]