Apple quietly ditches AMD

[Commodus]

On 7/10/2020 at 12:56 AM, atxcyclist said:

But if the task is to go 0-60, then buying something that degrades that performance for something subjective or aesthetic is pointless. In workstation computing power is more important than anything, and the Apple ecosystem is only beneficial over Windows if you do sound/music or movie production; Apple will only retain those markets if they can build processors and graphics cards that are fast enough for those professionals, otherwise they'll go to Windows/Linux x86/AMD64 machines that have that power available. 

 

 

I'm not talking about subjectives or aesthetics, although those can be considerations in non-critical tasks.  It's things like workflow, integrations with other apps and devices, that sort of thing. For example, I was taking calls and texts directly on my Mac years before there was a native option for it in Windows 10; that's helpful for my work.  Photoshop is clearly available cross-platform, but there's a collection of smaller tools I use (both apps and OS-level features) that simply work better on the Mac.  There wouldn't be much point to Photoshop rendering faster if many of the ancillary tasks were slower.

 

And while that is image editing, that's not music or video production... and importantly, it's not the only element of my work.

 

You're right in that Apple is gambling that it can provide CPUs and GPUs fast enough to reel in creators, but that's part of the point of the transition.  It doesn't want to be chained to someone else's architectures.  It also has a good precedent when the cheapest iPhone is faster than any Android phone... we have to see if that scales to PC-class chips, of course, but it's a start.

[saltycaramel]

We’re 2-3 years from the release of the first Mac where I feel Apple would have some trouble replacing the current AMD offering with Apple GPUs, and that Mac would be the Apple Silicon Mac Pro.

Why would they spoil their strategy 2-3 years in advance in a random developer session? Is that even decided for good already?

 

Every other Mac that is not a Mac Pro, I have little trouble imagining apple creating a big fat integrated gpu that would be as good or better than current (mobile) AMD offerings...just looking at their consistent dominance on mobile makes me think there’s a lot we still don’t know about their laptop/desktop silicon..

 

By the way next gen game consoles are based on a big fat SoC with integrated GPU too...Macs are becoming glorified consoles, so what? Not like you could upgrade the hardware before? (except the 2012 and 2019 Mac Pro) If anything, it will all be far more optimized...and Mac gaming will benefit from being lumped with iPads in terms of development...

[saltycaramel]

As for the MacPro in 2022 or 2023, the only way to compete with AMD would be to introduce discrete pcie daughter-board style Apple GPUs...the current MacPro already is designed with proprietary “MPX Modules” in mind...maybe the next one will only work with MPX Modules (based on Apple GPU and not AMD like the current ones) and not support regular GPUs anymore...they cut Nvidia, they can cut Nvidia+AMD...

[mr moose]

2 hours ago, saltycaramel said:

As for the MacPro in 2022 or 2023, the only way to compete with AMD would be to introduce discrete pcie daughter-board style Apple GPUs...the current MacPro already is designed with proprietary “MPX Modules” in mind...maybe the next one will only work with MPX Modules (based on Apple GPU and not AMD like the current ones) and not support regular GPUs anymore...they cut Nvidia, they can cut Nvidia+AMD...

 

I think the reality is exactly that, but not because they want to get rid of AMD, but because if they want mac to be anything over the next decade they are going to have to make something that has no alternatives and going full in house hardware that performs notably better is the only way to do that.  

 

Currently the only thing unique about mac is a handful of software options (all with alternatives).  What made mac successful in the first place was being the better option, people are happy to pay more for better (that's why mechanics are happy to pay 5 times more for a spanner).  However when they aren't offering anything tangibly better but still have the "tangibly better" price tag then we get the sales stagnation.  Mac peaked in 2015 according to most analytics,  I think Apple have decided there is no other way to make themselves uniquely better than other options on the market.

 

 

[mr moose]

4 hours ago, Kaloob said:

???

Hondas are pretty good sellers and most of their buyers aren't looking at 0-60 times. 

Honda's are among the best selling cars.  I believe the civic is the second most sold car in the US.

[leadeater]

On 7/10/2020 at 9:28 PM, Master Disaster said:

My understanding is back in the day Apple had the edge in graphics due to OOTB Adobe RGB colour space coverage certification which was still pretty rare on Windows displays and in audio because the Mac OSes sound hardware had significantly less latency than Windows.

 

Now colour space certified displays are much more commonplace and Windows sound cards have all but caught up on the latency there's really no benefit to using one over the other.

A display is just a display, the problem was Windows graphics drivers and windows support for extended colour spaces wasn't that great. But that also doesn't mean it wasn't possible either. Back in the day I used to game on a CRT monitor that came out of the hospital that was used for medical imaging, had I bothered to use the 5 BNC connectors on it over the VGA and had say a supporting Matrox GPU I would of been gaming at medical grade colour (so long as the games/applications were themselves accurate).

 

Colour is more an end to end issue rather than a single part of it as if only one is wrong it's all wrong, getting all that ticked off was just easier on Apple/Mac back then but it wasn't exclusive.

 

Same goes for audio, the problem with that whole latency discussion actually relates to Windows Visa and after where Microsoft changed the audio subsystem and that itself has latency. I have not looked in to the situation recently but I'm sure if it were a problem then audio interfaces will have DMA support and you just bypass all of that with supporting software, any problem can be solved if required or already has been.

[mr moose]

1 hour ago, leadeater said:

 I have not looked in to the situation recently but I'm sure if it were a problem then audio interfaces will have DMA support and you just bypass all of that with supporting software, any problem can be solved if required or already has been.

And those problems are rare in professional audio solutions.  It's really only the crap domestic stuff like creative and Asus that have issues due to the way their drivers are written (designed to hook into windows to provide god knows what features).  Even on cheap hardware with asios4all drivers I don't experience latency on mutlitrack recording.

[Kisai]

1 hour ago, leadeater said:

A display is just a display, the problem was Windows graphics drivers and windows support for extended colour spaces wasn't that great. But that also doesn't mean it wasn't possible either. Back in the day I used to game on a CRT monitor that came out of the hospital that was used for medical imaging, had I bothered to use the 5 BNC connectors on it over the VGA and had say a supporting Matrox GPU I would of been gaming at medical grade colour (so long as the games/applications were themselves accurate).

 

Colour is more an end to end issue rather than a single part of it as if only one is wrong it's all wrong, getting all that ticked off was just easier on Apple/Mac back then but it wasn't exclusive.

 

Same goes for audio, the problem with that whole latency discussion actually relates to Windows Visa and after where Microsoft changed the audio subsystem and that itself has latency. I have not looked in to the situation recently but I'm sure if it were a problem then audio interfaces will have DMA support and you just bypass all of that with supporting software, any problem can be solved if required or already has been.

Windows basically has the same problem that All PC's do, including Mac's. We stopped having dedicated hard audio parts and what exists on motherboards is largely a software solution.

 

Like when you look at the marketing for a "HD Audio Codec" (and in this term CODEC is different from how it's normally used)

https://www.realtek.com/en/products/computer-peripheral-ics/item/alc898

All it talks about is the ADC and DAC. It doesn't talk about any kind of hardware mixing.

 

Then go look at a much older sound card:

https://web.archive.org/web/20110713125655/http://us.store.creative.com/Creative-Sound-Blaster-XFi-Titanium-PCI/M/B001E25KDK.htm

(This is the sound card I have)

Quote

Intense audio for your games, music and movies

• EAX® 5.0 sound effects and 3D positional audio for total game immersion
• Precise and detailed surround sound - even with headphones
• Hardware acceleration that boosts gaming performance
• Restores the details and vibrance to your MP3s
• Works with your PCI Express equipped PC

Get extreme audio performance from your PC with the PCI Express Sound Blaster X-Fi Titanium sound card. You'll hear realistic EAX® 5.0 sound effects and 3D positional audio in your games that's so accurate you can locate opponents by sound - even over normal stereo headphones. Plus, get unbeatable performance with hardware accelerated audio. All your music and movies will sound better too with Xtreme Fidelity audio technology that restores the detail and clarity to compressed music files like MP3s.

The CA20K2 and EMU20K2 basically have their own processors, memory and do the EAX and SRC in hardware. The CA20K2 actually supports Microsoft UAA (which is what Vista+ requires)

 

And the reason I used the x-fi over the onboard audio is the onboard audio was so much noisier. There is a better model, but really it is a case of diminishing returns since Windows doesn't really make use of the hardware acceleration and all you really get from the dedicated card is the ASIO support. 

 

What, or where the audio acceleration really belongs now is on the GPU. AMD was doing this with the trueaudio tech, nVidia did something like this with their RT tech.

 

I remember back in 1998 when the difference between using the crappy AC97 and a dedicated soundcard was bout 10%. More if you were forced to use the software MIDI, since sound cards at that time had hardware MIDI synths.

 

[mr moose]

15 minutes ago, Kisai said:

Windows basically has the same problem that All PC's do, including Mac's. We stopped having dedicated hard audio parts and what exists on motherboards is largely a software solution.

 

Like when you look at the marketing for a "HD Audio Codec" (and in this term CODEC is different from how it's normally used)

https://www.realtek.com/en/products/computer-peripheral-ics/item/alc898

All it talks about is the ADC and DAC. It doesn't talk about any kind of hardware mixing.

 

Then go look at a much older sound card:

https://web.archive.org/web/20110713125655/http://us.store.creative.com/Creative-Sound-Blaster-XFi-Titanium-PCI/M/B001E25KDK.htm

(This is the sound card I have)

The CA20K2 and EMU20K2 basically have their own processors, memory and do the EAX and SRC in hardware. The CA20K2 actually supports Microsoft UAA (which is what Vista+ requires)

 

And the reason I used the x-fi over the onboard audio is the onboard audio was so much noisier. There is a better model, but really it is a case of diminishing returns since Windows doesn't really make use of the hardware acceleration and all you really get from the dedicated card is the ASIO support. 

 

What, or where the audio acceleration really belongs now is on the GPU. AMD was doing this with the trueaudio tech, nVidia did something like this with their RT tech.

 

I remember back in 1998 when the difference between using the crappy AC97 and a dedicated soundcard was bout 10%. More if you were forced to use the software MIDI, since sound cards at that time had hardware MIDI synths.

 

Sound cards do not need DSP's in them to be a "sound card" or even just a good solution,  all a sound card is is a ADC/DAC pair and digital interface (usually PCI or USB),  anything on top of that is feature fluff which is either important to the end user or not important.    Codecs are both legitimately either hardware or software depending on application.  Even "high end"++ soundcards with dedicated DSP's can suffer terrible latency meanwhile a properly designed 16 channel fire-wire/optic interface can simultaneously record all channels with negligible latency and loop to outboard gear. 

 

 

++ I put high end in quotation marks because the idea that any domestic sound card these days being high end is a laugh, there is some terrible rubbish at stupid prices that do little more than make you think you have something that sounds good. People are so feeble with their hearing that you only need to increase the volume for 99% of people to report an improvement in sound quality.   And that's before we even get into the science of electronics and why majority of it is just marketing.

 

 

[leadeater]

53 minutes ago, Kisai said:

Windows basically has the same problem that All PC's do, including Mac's. We stopped having dedicated hard audio parts and what exists on motherboards is largely a software solution.

But we're talking about audio recording, and in a more "professional" sense so those are not used.

 

You're not going to be connecting a Yamaha QL5 to integrated onboard audio, it has Ethernet and USB connection for connectivity to a computer, but either way nothing along those lines would regardless of Ethernet/USB connectivity or not. You don't pay $20k USD to plug in to a 3.5mm Realtek line input.

 

Like the entire industry has moved away from relying on or utilizing audio processing in line with a PC and is fully digital, everything has been externalized as much as possible with digital audio streams being passed to the PC and if you need to do something you either do it outboard of the PC or directly in the software. I'm not as well versed with the current goings on as the person I used to talk to about it moved to a different city but Pro Tools has supported Ethernet for a long ass time now and I'm sure the people most affected by the Windows audio subsystem changes were the smaller indies and hobbyist that simply had to move to USB connected equipment as that start coming out and got cheaper over time as well, everything else was already using better methods and tools.

 

Like I said if latency was important a connection that supports DMA would be used, or is being used I don't know (Ethernet supports that), but I just don't think it is anymore with the way everything is digitally interfaced now and synchronized.

[ToboRobot]

This isn’t a confirmation that Apple is abandoning outside GPU vendors.

 

Id expect low power AS Macs to run Apple GPUs but there will be GPUs for heavy compute in the future for high performance workloads.

[Kisai]

1 hour ago, mr moose said:

 

++ I put high end in quotation marks because the idea that any domestic sound card these days being high end is a laugh, there is some terrible rubbish at stupid prices that do little more than make you think you have something that sounds good. People are so feeble with their hearing that you only need to increase the volume for 99% of people to report an improvement in sound quality.   And that's before we even get into the science of electronics and why majority of it is just marketing.

 

 

Yep. To me the onboard had an audible hiss at any volume level, where as the X-fi only had that hiss if the microphone was turned on. So I knew where the blame lied. If I buy a new system I probably will just keep using the same card. I listen to it at "10" (out of 100) 

 

Creative labs current AE5 card... has the same SNR as the "X-fi Titanium HD" (the one above my card) and it's just like... I bought the two titanium models off eBay ($35 ea), and gave one to my neighbor who was having a rubbish time with the newer creative labs card (I think it was a Audigy Rx, the thing would never work on Win8.1), but all he needed was ASIO. Anyway, the AE cards are just all marketing nonsense. Would you really pay $180 for capability that you basically will never practically use?

 

Pretty much everything else out there is not really any better

https://www.newegg.com/evga-nu-audio/p/N82E16829276001 $400.

https://www.newegg.com/asus-essence-stx-ii/p/N82E16829132072 $300

 

The AE9 is right in the middle of those for prices. I'll pass.

[Kisai]

1 hour ago, leadeater said:

But we're talking about audio recording, and in a more "professional" sense so those are not used.

 

You're not going to be connecting a Yamaha QL5 to integrated onboard audio, it has Ethernet and USB connection for connectivity to a computer, but either way nothing along those lines would regardless of Ethernet/USB connectivity or not. You don't pay $20k USD to plug in to a 3.5mm Realtek line input.

 

There's a lot of different kind of professional audio stuff, but just to stay on point, midi kit can be ethernet, usb, or 5-pin din, and it's really just a question of what you're trying to hook up. Vintage synths don't have any of that connectivity other than 5-pin din midi and analog output.

[leadeater]

40 minutes ago, Kisai said:

Vintage synths don't have any of that connectivity other than 5-pin din midi and analog output.

Which along with everything else analogue were the only things affected by the changes and if you really want to keep using some of what you have connect it first to an audio interface which has a USB connection back to the PC. It actually makes zero difference if it's a $20K console or a $300 one, long as it's digital and directly controlled by all the standard software that is used the whole "Windows has latency" is pretty well unrelated.

 

Audio people can be the worst, "I just have to use my [insert here] from 1983 because [invalid reason]". And that doesn't mean what ever it is isn't excellent in quality but deaf ears to complaints about 20+ year old equipment not working with modern things, not that it isn't possible.

 

Basically if you had more than just a passing interest to record something this entire thing was a non issue, across all market price ranges.

[Donut417]

On 7/10/2020 at 1:00 PM, Spindel said:

The question not enough people in this thread don’t ask (and for obvoius reasons can’t answer) is: How well will Apples GPUs perform compared to AMDs offerings?

The real questions are how well the GPUs are, BUT also how well the software is optimized for the hardware. Its like game consoles. Devs are able to optimize for the specific hardware, that's why a $500 console can technically do 4K. Id imagine that Apples in house software ill be well optimized but I wonder how Adobe will do? 

[Lord Vile]

On 7/10/2020 at 1:19 AM, emosun said:

all I know is this is just another nail in the coffin for the forever long battle of not being able to play games on a mac

They did run tomb raider at decent quality whilst being emulated on what is essentially an Alpha piece or hardware.  

[iBabySlapper]

Correct me if I'm wrong (since I don't know much about ARM stuff), but I thought this was a given since all ARM based stuff seems to be an SoC design

[PC LOAD LETTER]

On 7/9/2020 at 8:30 PM, Jeppes said:

People who buy Apple computers dont care about performance. They would a Threadripper instead of Mac pro if they cared, theyll buy an ipad pro packed into a pc case fpr 8000$ if its just marketed properly to them.

Is that so?

[mr moose]

16 minutes ago, PC LOAD LETTER said:

Is that so?

Well, I have encountered people trying to recommend a $2000 macbook for a grandma who wants to check emails and facebook.  these types of recommendations can only be the result of marketing.

 

[Kisai]

7 hours ago, iBabySlapper said:

Correct me if I'm wrong (since I don't know much about ARM stuff), but I thought this was a given since all ARM based stuff seems to be an SoC design

The flash memory in the iphone 6S and ipad are NVMe devices. They can put whatever they want in the SoC.

 

The question will be if they can put the PCIe lanes and memory controllers needed for a Mac Pro, or even a iMac that makes it competitive with an Intel part. Like on white-box PC's, Intel CPU's do not have enough PCIe lanes for the number of ports they have. Like just to point out the big joke. Motherboards typically have 3 x 16 lane slots, and 3 m2 NVMe drive slots and 3, 1x slots for everything else, plus maybe 4 usb-c ports. So add all of that up (16 x 3) + (4 x 3) + (1 x 3) + (4 x 4) + anything else, requires a minimum of 80 lanes. Most of those chipsets Intel puts out, put only 20 or 24 lanes on the chipset and 16 on the cpu. If you install any other card into the system, you get either a 8+8+0 or a 8+4+4 or 8+8+4, depending if those lanes exist.

 

So I don't see Apple making a crippled Mac Pro. I think we may see a huge problem with devices not having drivers, but that's not even a change from the status quo.

[leadeater]

2 hours ago, Kisai said:

The flash memory in the iphone 6S and ipad are NVMe devices. They can put whatever they want in the SoC.

 

The question will be if they can put the PCIe lanes and memory controllers needed for a Mac Pro, or even a iMac that makes it competitive with an Intel part. Like on white-box PC's, Intel CPU's do not have enough PCIe lanes for the number of ports they have. Like just to point out the big joke. Motherboards typically have 3 x 16 lane slots, and 3 m2 NVMe drive slots and 3, 1x slots for everything else, plus maybe 4 usb-c ports. So add all of that up (16 x 3) + (4 x 3) + (1 x 3) + (4 x 4) + anything else, requires a minimum of 80 lanes. Most of those chipsets Intel puts out, put only 20 or 24 lanes on the chipset and 16 on the cpu. If you install any other card into the system, you get either a 8+8+0 or a 8+4+4 or 8+8+4, depending if those lanes exist.

 

So I don't see Apple making a crippled Mac Pro. I think we may see a huge problem with devices not having drivers, but that's not even a change from the status quo.

Your math is slightly wrong on those USB ports, also worth noting if you want to tally things up then for Intel you need to use HSIO lane count rather than PCIe lane count off the chipset. HSIO are the actual things used and can be assigned for purposes like PCIe lanes for PCIe slots, there is a total of 30 of these with 6 mandatory USB lanes which leaves those 24 HSIO lanes that can be assigned for different things.

 

Each USB 3.1 G2 port only requires a single HSIO lane not 4 so it would be 4 total, but there are 6 mandatory ones so covered anyway.

 

It's all basically moot though as if you are relying on PCH lanes then you're just fooling yourself, DMI 3.0 is only equivalent to 4 PCIe lanes of bandwidth so really you have 16 + 4 PCIe lanes on consumer Intel and I really don't care how many HSIO lanes they cram in to the PCH as you can't use them. The only purpose the serve is for flexibility purposes for what and how many things you can connect but you can only ever utilize x4 worth of them at any one time.

 

Personally I think Apple will go the TR/EPYC route in regards to PCIe lanes and have them all on the CPU/SoC and forgo PCH, makes more sense for 'fixed' hardware configuration devices.

 

This is rather useful

 

 

[Spindel]

14 hours ago, Donut417 said:

The real questions are how well the GPUs are, BUT also how well the software is optimized for the hardware. Its like game consoles. Devs are able to optimize for the specific hardware, that's why a $500 console can technically do 4K. Id imagine that Apples in house software ill be well optimized but I wonder how Adobe will do? 

I hope apples offerings will be at least OK. Just as Apples ARM move is only positive for the competition in the desktop CPU space Apples GPU move is a positive for the competition in the GPU space.

 

You can’t stop to wonder if Apple will be releasing stand alone GPUs also (integrated are a given) for he likes of iMac, Mac Pro and eGPU enclosures.

[Kisai]

1 hour ago, leadeater said:

Your math is slightly wrong on those USB ports, also worth noting if you want to tally things up then for Intel you need to use HSIO lane count rather than PCIe lane count off the chipset. HSIO are the actual things used and can be assigned for purposes like PCIe lanes for PCIe slots, there is a total of 30 of these with 6 mandatory USB lanes which leaves those 24 HSIO lanes that can be assigned for different things.

 

Each USB 3.1 G2 port only requires a single HSIO lane not 4 so it would be 4 total, but there are 6 mandatory ones so covered anyway.

 

It's all basically moot though as if you are relying on PCH lanes then you're just fooling yourself, DMI 3.0 is only equivalent to 4 PCIe lanes of bandwidth so really you have 16 + 4 PCIe lanes on consumer Intel and I really don't care how many HSIO lanes they cram in to the PCH as you can't use them. The only purpose the serve is for flexibility purposes for what and how many things you can connect but you can only ever utilize x4 worth of them at any one time.

 

Personally I think Apple will go the TR/EPYC route in regards to PCIe lanes and have them all on the CPU/SoC and forgo PCH, makes more sense for 'fixed' hardware configuration devices.

 

This is rather useful

 

 

My math was using the USB4/TB3 4-lane configuration but I get your point.

[Orfieus]

On 7/9/2020 at 8:44 PM, Commodus said:

We do care about performance, we just aren't willing to sacrifice an interface and ecosystem we prefer just to say we score higher in a benchmark.  To abuse car analogies: if you go shopping for a car based purely on 0-60 times, you're A) missing out on so much more and B) probably not very fun.

Uh I see it differently.

 

People who buy Apple are people who are buying a $100,000 sports car to only go to the mall and back

 

People who choose Apple for work purposes are people who buy a $100,000 sports car with a trailer hitch

[leadeater]

19 minutes ago, Orfieus said:

People who choose Apple for work purposes are people who buy a $100,000 sports car with a trailer hitch