The problem with AMD is volume. Can they meet Apple’s volume requirement? I don’t have to look at Apple. In my country, for retail prebuilt machines, AMD chips are only readily available in their low end offerings for lower end laptops and all in one. All their fancy Ryzen are only available as a component for system builders. Retail wise, I just don’t see the OEMs like Asus or HP pushing their AMD models in volume. It’s all intel world.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
AMD
- Thread starter UBS28
- Start date
- Sort by reaction score
As someone who has been following CPU development two decades, there are various ways to look at this. Renoir was the largest generational uplift in mobile x86 performance in approximately 14 years. Apple on its first attempt has fielded a SOC with comparable CPU performance with superior performance per watt. The M1 is amazing and so is Renoir. Period.
There are some people saying that this is a budget chip, compared to AMD’s greatest but that’s not really a fair assessment. As of this writing its currently the most powerful CPU Apple has even designed. In addition, at approximately 16 billion transistors on a cutting 5mn process, its almost certainly a more expensive chip to produce than anything AMD has produced in mobile. Of course, Apple has an advantage over Intel/AMD/PC OEM because they can they are the manufacturer and they can absorb some of the costs in the total cost of the system. AMD is really limited in mobile because Intel pays OEMs not to put them poor designs, and they design of the laptop really makes a difference in performance and battery life. Apple designs the entire system for optimal performance/power characteristics.
And there are some who say Apple will always been on a new process node when compared to AMD. I hope you realize that makes absolutely no sense whatsoever. AMD and both use TSMC, Apple has no advantage here, except having more money to spend on the latest node. There will be a 5nm overlap of the companies just like it was with 7nm.
For me I don’t care about Apple/AMD, but I was curious to see x86 vs. ARM, and round one was inconclusive for me. Having that said, Apple has the advantage here because they don’t have to sell these chips, which cuts into the cost/design decisions. In addition Apple has more money to pour into research by an order of magnitude. AMD is tiny compared to Apple. In addition, PC OEMs are pretty dumb, and Intel still dominate with 80 percent of the market despite having vastly inferior chips to AMD’s offerings. AMD will struggle to compete with Apple long term who seem to have equally good engineers, but much more money to spend.
There are some people saying that this is a budget chip, compared to AMD’s greatest but that’s not really a fair assessment. As of this writing its currently the most powerful CPU Apple has even designed. In addition, at approximately 16 billion transistors on a cutting 5mn process, its almost certainly a more expensive chip to produce than anything AMD has produced in mobile. Of course, Apple has an advantage over Intel/AMD/PC OEM because they can they are the manufacturer and they can absorb some of the costs in the total cost of the system. AMD is really limited in mobile because Intel pays OEMs not to put them poor designs, and they design of the laptop really makes a difference in performance and battery life. Apple designs the entire system for optimal performance/power characteristics.
And there are some who say Apple will always been on a new process node when compared to AMD. I hope you realize that makes absolutely no sense whatsoever. AMD and both use TSMC, Apple has no advantage here, except having more money to spend on the latest node. There will be a 5nm overlap of the companies just like it was with 7nm.
For me I don’t care about Apple/AMD, but I was curious to see x86 vs. ARM, and round one was inconclusive for me. Having that said, Apple has the advantage here because they don’t have to sell these chips, which cuts into the cost/design decisions. In addition Apple has more money to pour into research by an order of magnitude. AMD is tiny compared to Apple. In addition, PC OEMs are pretty dumb, and Intel still dominate with 80 percent of the market despite having vastly inferior chips to AMD’s offerings. AMD will struggle to compete with Apple long term who seem to have equally good engineers, but much more money to spend.
You have to realise the AMD chips do not actually run at 15w. Multiple test have shown that it will peak at 50w or higher for benchmarks. Even running cinebench the AMD laptops fans rev up like a jet. That also why they won’t ever run fanless like the air - actual power consumption and heat is too high.
That’s also why you won’t get any claims of 20 hour battery life on any Thin AMD laptop. It is closer to 10 hours which is half of the MacBook Pro with m1.
Last edited:
You can’t compare the multicore when AMD has twice the big cores.
Anyways.. Why would Apple rely on AMD and their roadmap? Apple has a better architecture, better performance per watt, more efficient GPU tuned with their own API. Just an overall complete package that they can iterate for their own products and control their own destiny.
You’re thinking very short term if you’re still clinging to 32-bit and x86 Windows. The future of Windows is also Arm.
The future of Windows is ARM is a bit premature. PC landscape is very different than Apple...
This is another factor. It’s my understanding that you still barely even find these Renoir laptops.
In the PC world “paper launches” are the norm. Look at how mad Nvidia fans are with them that they can’t get the new GPUs. Same with the new AMD GPUs.
Why would Apple want to deal with this? They can order exactly as much product as they need from TSMC and they already have a relationship with them from the iPhone and iPad chips. AMD is not first priority at TSMC, Apple is. They get the best stuff first. Vertical integration is the dream of every company and Apple has it.
That 3nm is probably getting out in early 2022 but we already know that the 2nm can be out late 2022 or 2023, 4nm is out of the question for apple at leastThat is probably not true. The next TSMC mass production node to come online will be 3nm ; not 2nm.
3nm very high volume isn't coming until 2022.
There is a "4nm" in-between process node that Apple might use in 2021 to squeeze more into the A15, but they may skip it. AMD, if stay on their 12-16 month "tick tock" schedule, should be rolling onto 5nm in 2022 after a "optimization step" on 7nm in 2021.
Likewise AMD should be on DDR5/LPDDR5 in 2022 also.
All of this stuff has long lead times. TSMC working on 2nm now is so that it is ready like 3-4 years for volume production. Apple is going to do better in the sub 30W range but above that and in desktop zone they don't have a "slam dunk" advantage if AMD continues their incremental progress. ( e.g., AMD applies something like "infinity cache" to the CPUs. )
Apple is probably not going to "Run away" from AMD. Or Intel if they would pull their head out of there butt and get their "Intel 7nm" kinks worked out.
"TSMC plans on using FinFET transistors for its 3nm mode before switching to GAAFET (gate all around) for 2nm chips(expected for late 2023). Unlike FinFET, which doesn't surround a channel on all sides, GAA surrounds a channel using a Gate. The latter method makes current leakage almost negligible"
So its clear in 2-3 years Apple arm chips will get to 3nm and even to 2nm while others will still be on 10nm and amd probably still on 5nm...but dont get me wrong, i hope Amd to succeed but is still behind and amd is faceing the wall that is x86 arhitecture...and cannot compete until they come with a proper perf/w
Lets hope amd will go arm also...nvidia based on the money spend for arm will probably do the same
Probably none, simply because AMD doesn’t seem to ship them.
Which laptop would that be? I looked around - and 4800U seems to be unavailable. Lenovo has some cheapo laptops with 4700U but that’s about it.
Probably looked up one of 4800H gaming laptops with nVidia RTX 2060 and decided it's comparable, seriously. No wonder the same people suggest Apple could've gone with AMD.
Legion 5 with 144 Hz screen. Have they sold out? It is a really fast laptop, but also unwieldy and would work best as a desktop replacement.
Presumption makes for nonsense posts. I have one, a Legion 5. Also have an EE degree, and have built computers for 40 years. One of the fastest and best speced laptops I have owned. IF looking for a desktop replacement that is portable, the MUCH increased spec and comparable performance is there. IF looking for an easy small portable solution, the Mac M1 has that advantage, a better but smaller portable screen, and comparable performance until it reaches memory and storage limits or until running software that utilizes the GPU, at which point the AMD will eat it’s lunch.
People loose sight that the M1 is handsomely beating these x86 chips in single threaded performance with a fan-less design and performing about the same MT with half big, half little cores for a fraction of the power.
Can these AMD chips be fitted inside the MBA's chasis? Can they run fan-less with 15-hour battery life? how cool does these x86 chips run after a 30-minute full load?
When you think of it, the M1 and Apple Silicon in general are not just great for laptops and mini-desktops where thermals are key, but they would also be incredible chips for data centers where energy and cooling costs are crucial. I know Apple is not in the business of selling chips, but the surely could shake that market if they were to enter it.
That is a budget gaming laptop with a 4800H (not U), it weights over 2 kg, is actually bigger than the 16" MBP and has the same battery capacity as the 13" MBP. Bringing this class of laptops into the discussion makes no sense whatsoever.
That will certainly be an interesting one to watch. I expect OEMs to give more attention to WoA now, but for the moment at least Samsung is probably the only one in a position to make anything remotely comparable to Apple Silicon Macs. Others will have to use Qualcomm chips like the 8CX.
It makes sense IF, AS I SAID, a person wants a desktop replacement that is also portable. But NOT, AS I SAID, IF a person wants a small easily handled laptop. And, a 80Whr battery is available. WITH the 80 Whr battery, 4 TB nvme, 64gb Ram, it is roughly same cost as 16gb/1tb MBP M1. And sorry about the “u”, brain fart.
Yeah but the processors such as the 15w 7nm SQ1 or the 7w 8CX isn’t that interesting. Its not in the same ballpark as competing x86 designs such Ryzen 7 4800u. Even in its 10w TDP down option, the AMD chip is faster than both of Qualcomm chips. Samsung’s Exynos chips historically have been the slowest and most power hungry ARM chips on the market. They will need to claw up a lot of lost ground to compete with Qualcomm, let alone Intel or AMD.
4nm out of the question?
Apple Expected to Use TSMC's Advanced 5nm+ and 4nm Technologies for Future iPhone Chips, Likely Macs Too
While the A14 Bionic chip in iPhone 12 models was the first chip in the smartphone industry to be manufactured based on a 5nm process, Apple and its...
www.macrumors.com
I wouldn't expect Apple to use 5P on the high profile Apple implementations. but 4nm may come into play.
Not really for the M-series. very good chance that N4 (4nm) has lower costs than N3 will have. Uses the same design rules as N5. For larger dies it probably will be a more mature option than N3 in the late 2022 - first half 2023 timeframe.
Macs ( and iPad Pros ) don't have the volumes of iPhones. Also if Apple isn't too interested in craming lots more "stuff" into their 3nm iPhone die they can go smaller and offset the substantial increase in wafer costs and upfront R&D.
N4 is there because bigger die implementors aren't going to chase off to 3nm with the larger designs as quickly as possible. As Apple's dies get substantively bigger they are going to be increasingly be in the same group for at least part of their portfolio.
Apple isn't magically immune to the same factors that AMD is facing with larger dies.
AMD could chuck x86 without dropping x86_64. That is in part what Apple has done ( they ejected the 32 Arm to ease complexity). Incrementally it is getting there. Microsoft isn't doing 32-bit Windows 10 for OEMs anymore. The data center market has already minimized 32-bit OS usage. If AMD gets a decent widely perceived lead over Intel it wouldn't be such a bold move to do. Intel is probably the one more hooked to "way , way , way backwards compatibility is everything" crack.
As for Nvidia moving to ARM .... errr they are already there. That's one of the major issues with the acquisition is that they are already an ARM implementers and would be licensing out to other implementors in possible non-competitive fashion. ( Yes , Nvidia now in the pre-acquisition phase says that it will be separate ... but that isn't a rock solid contract. A substantial part of that is them wanting to close the acquisition. The deal would probably be dead if they said anything else )
There is a notion that Apple not just has more money to spend , but that Apple has more money to do a 'lock out' for several months. In part, Apple can pay upfront money to ASML Twinscan and other bleeding edge equipment and then gets a 'early investor' deal for exclusive access to the equipment for a period of time (since they pragmatically paid for it).
That said AMD isn't exactly "broke" anymore. But they are running a multiple front war on Intel and Nvidia. ( Xlinx just makes the multiple fronts more complicated. Not a losing front but more complications. )
If you can sell the chips made at higher margins, then it doesn't matter if the chip dies cost more to make on the bleeding edge processes. AMD tends to make bigger dies than Apple does. But bigger dies will cut into everyone (including Apple when they get there). But if apple "overpays" to buy out just about all of the wafer starts then it is more so about spending cash to stifle the competition. But with AMD not having to borrow money to keep the lights on they can start to bid up early wafer starts also. It is only if Apple is buying something outside the bid process where their money could block folks.
Amd will face the issue on perf/w in the chip fight if they stay on x86 architecture too long
Think what dual M1 can offer or how the arm can scale from now on....x86 will face the wall in perf/w when they will hit 2nm or 3nm
When Apple will offer 2nm in 2024, others will still offer x86 7nm or even 10nm with BS of iGpu and so on(M1 has around 2.8 terraflops/10W)...and think what will be the perf/w into the cheapest mac , like mac mini...not to mention the 14" mbp or the imacs
The best acquisition for apple was to steal the chip brain from Intel , years ago
Think what dual M1 can offer or how the arm can scale from now on....x86 will face the wall in perf/w when they will hit 2nm or 3nm
When Apple will offer 2nm in 2024, others will still offer x86 7nm or even 10nm with BS of iGpu and so on(M1 has around 2.8 terraflops/10W)...and think what will be the perf/w into the cheapest mac , like mac mini...not to mention the 14" mbp or the imacs
The best acquisition for apple was to steal the chip brain from Intel , years ago
True, that's really just swapping Intel for another chip supplier (though the possibilities with Arm are still interesting in their own right) but Qualcomm as you say have a long way to go to even be competitive in the U series space. I think they will have an inherent advantage with fanless computing as frankly Intel's Y series just never have really worked well without active cooling. Interesting about Samsung, I didn't realise they had that reputation. Maybe this will spur them to up their game as well?
x86 (as a label for the broad historical to future scope / lineage ) architecture isn't the problem. ARM doesn't have a magically better op codes. They have more modular opcodes that allow one to dump old stuff if you choose to. That's the bigger problem AMD has. Having to compete with "we do super old stuff" Intel and having to play the "Match feature list " game with system vendors and super ultra conservative customers .
Cut loose the 32-bit baggage and the legacy SIMD stuff that isn't very good SSE variations. AMD could loose a boat anchor that really doesn't matter much on modern code. It more is the guts to leave some old 'dusty' software behind than some magic benefit in opcodes.
Dual M1 isn't going to do jack squat because there is no interpackage bus to commnicate over. This is largely an iPad Pro die. There isn't good evidence that Apple's System cache and total focus on Unified Memory is going to scale extremely well to any large double digit number they choose. Apple is squeezing out some extra performance here in part because it is optimized for lower counts in a more narrow silo.
ARM and x86-64 scale doesn't really matter to macOS that doesn't go past 64 cores. High double digit core counts is a path Apple probably isn't going to get on. Neither is much higher than average clock rates.
that is ton of hand waving. Nothing about the opcodes make that inherently true. How the opcodes are implemented is up to the designers (which can be good or bad. ) and how much time they spend.
One of Apple's upside is that they do less. They make few designs which means they spend more time polished 2-3 rather than trying to make everything for everybody and having to do 3-6 designs. Not spreading yourself too thin matters. AMD got better when the stopped trying to chase off into every nook and cranny that Intel was pursuing at the same pace. And Intel has gotten worse (compared to themselves) as they expanded past what they were doing 10 years ago.
AMD is only about 8-14 months behind Apple now depending upon size of die being implemented. They aren't going to be further back in 2024. As they have more cash on reserve that will probably shrink some over the next 4 years rather than widen.
x86 isn't a synonym for Intel. It really wasn't before and it is even less so now.
When AMD gets to 5nm then can talk smack. (likewise when the Samsung/AMD GPU mash up finally comes out on 7nm or 5nm ) . But AMD also has 23+ TFLOPs now.
https://adoredtv.com/amd-announces-mi100-better-performance-than-expected/
That is ten times faster now. Apple has nothing. For those who need 23+ TFLOPS now doing it under 10W probably doesn't matter much.
There is no iMac now because Apple doesn't have an iMac solution now. Apple is rumored to be working on a "half" size Mac Pro. Mostly likley because they don't have a full size appropriate SoC coming any time soon (or even next year , 2021).
AMD is mainly trying to get back into the top end space that Intel and Nvidia cover. That's a very good thing for AMD's profitability. Trying to go low and compete with Apple, Qualcomm, Samsung, Huawei/HiSilicon and others would have probably turned out worse for AMD than it did for Intel ( which was major capital loss that Intel could afford to write off).
The best "acquistion" Apple did was not doing one. It was saying "no" more than saying yes. They built fewer chips and kept their objectives narrow. When they broad the scope to cover the Mac line up it will be questionable which side takes the bigger "hit". The scope of the Mac line up ( fewer substantive different systems ) or Apple's performance path in some SoC segments.
The A-series X models have been on a slower path than the nominal iPhone SoC. Apple has waited fro process shrinks to move A-nnX models forward. Those process shrinks are coming slower and more expensive over time. It isn't going to stop but the pace isn't necessarily going to be the same speed.
Now that ARM has establish the Neoverse ( N1 , N2 ) and the V series for data center they are turning their attention more to the inbetween. The X1 is a path toward Qualsomm , Samsung , Microsoft , and some others pooling their money to get more traction in the out of the smarphone optiimzed space.
Cortex-X Custom CPU | Empowering Next-Gen CPU Performance
The Arm Cortex-X Custom (CXC) program reshaping CPU standards for future devices, optimizing for unparalleled user experiences.
www.arm.com
Part of the issue is that Samsung and Qualcomm have (or had for previous versions) architecture licenses and aren't fully utilizing them. Part of this is down to again trying to chase everything for everyone options for every phone , tablet , etc possible.
What is needed is a larger pool of money that is tarted to hit primarily the Windows / Chromebook market specifically. It will be harder to catch AMD and Intel in performance when splitting focus in 4-5 different direcitons at the same time with a relatively small pot of money.
when the X1 designs trickle out the performance 'hole" that 8CX variants are in won't be as deep. Apple has been "hiding" a more higher tier aimed processor inside of more expensive phones and tablets as they worked their way toward the lower "half" of the Mac line up. They have paid more and gotten more in a narrower area.
How scalable is Apple’s design? This is one of the main questions, isn‘t it? Srouji said:
“You want to deliver the highest performance at the lowest power consumption. And that's exactly where we want to take the Mac. . . . [O]ur plan is to give the Mac a much higher level of performance, while at the same time, consuming less power. . . . But that's just part of the story. Our scalable architecture includes many custom technologies, that . . . will bring even more innovation to the Mac. . . . This will give the Mac a unique set of features and incredible performance“ (highlighting by me, source: https://www.zdnet.com/article/how-the-apple-silicon-mac-will-attack-the-pc/)
On the other hand:
Who is going to end up being right?
“You want to deliver the highest performance at the lowest power consumption. And that's exactly where we want to take the Mac. . . . [O]ur plan is to give the Mac a much higher level of performance, while at the same time, consuming less power. . . . But that's just part of the story. Our scalable architecture includes many custom technologies, that . . . will bring even more innovation to the Mac. . . . This will give the Mac a unique set of features and incredible performance“ (highlighting by me, source: https://www.zdnet.com/article/how-the-apple-silicon-mac-will-attack-the-pc/)
On the other hand:
Who is going to end up being right?
How does AMD compete with this?
Apple's move to M1 chips will save $2.5B this year – IBM exec - 9to5Mac
An analysis by a senior IMB exec suggests that Apple's move to M1 chips will save the company around $2.5B this year, with far larger savings to come ...
9to5mac.com
I'd guess that the savings for making their own chips at the high-end are much greater than at the low end.