RIP Mac gaming. It was never actually a vital group, and it's been on life support after Catalina nuked everyone's back catalogs. But this is probably gonna kill Bootcamp too.
RIP Mac gaming. It was never actually a vital group, and it's been on life support after Catalina nuked everyone's back catalogs. But this is probably gonna kill Bootcamp too.
Well, Apple is about the only company to have successfully pulled off two different architecture transitions so far, so if anyone can do it, it's them. Although I imagine that the Mac Pros will...
Well, Apple is about the only company to have successfully pulled off two different architecture transitions so far, so if anyone can do it, it's them.
Although I imagine that the Mac Pros will remain x86-64 for the time being (until they can get ARM chips with Xeon-level performance, which (to my layman's eye at least) requires a different set of optimizations than the ones made for the A series as is (performance per watt being a huge consideration in a phone chip), plus getting all the pro software ported to ARM as well (because an emulated solution, even on powerful hardware) isn't going to perform at the levels expected of a $50k pro workstation.
It'll be interesting to see regardless, given the great work they've done with their mobile processors so far.
Apple's CPUs are actually very competitive with x86 chips.
Apple's CPUs are actually very competitive with x86 chips.
The Apple A13 - even its implementation in the iPhone SE, in microbenchmarks achieves on par single core performance [1] with the Core i7 8086k [2] and Ryzen 9 3950X [3]. That's the highest single core performance you can buy in PCs in principle.
That's pretty interesting to see. One thing I was thinking of, Apple could (with their enormous cash haul) purchase VIA and that way acquire an x86(64) license that works more along the lines of...
That's pretty interesting to see.
One thing I was thinking of, Apple could (with their enormous cash haul) purchase VIA and that way acquire an x86(64) license that works more along the lines of the license they have from ARM in order to produce the A series. They could bring more of their A series work to that, but keep compatibility with more existing software for pro machines at the least.
I doubt it'd ever happen, but it's fun to think about what could come of something like that.
I'm always skeptical when it comes to iPhone benchmarks. I have seen so many unfair comparisons, and since so much of Apple's iPhone CPUs are black boxes you can never be sure how real world...
I'm always skeptical when it comes to iPhone benchmarks. I have seen so many unfair comparisons, and since so much of Apple's iPhone CPUs are black boxes you can never be sure how real world performance is supposed to be, especially since there is essentially no cross-platform software available for iOS.
If you were wondering what I meant by unfairness, the last time I saw a comparitive benchmark between iPhone and any given Android handset, they tested for speed of h.264 encoding. What they completely missed was that one of them had a dedicated hardware encoder on them and the other did not. Of course one was going to blow the other away.
I have limited trust for geekbench. From what I can tell it is not open source so you don't know what is actually happening behind the scenes. In any case, they are still somewhat limited in...
I have limited trust for geekbench. From what I can tell it is not open source so you don't know what is actually happening behind the scenes. In any case, they are still somewhat limited in usefulness as a metric because even though they advertise that their testing mirrors real usage, it is still a purely synthetic benchmark.
Being proprietary isn't necessarily bad in this context, but the fact that geekbench is frequently the only benchmark run for comparison. There should be a wide range of benchmarks and those should all be based on a shared codebase as much as possible.
When Intel makes an incremental improvement on their CPUs, reputable tech news outlets will run a battery of 10 or more benchmarks of various types. Apple is claiming to have the most efficient and powerful mobile CPUs on the market, so why don't we have the same level of scrutiny?
You don’t have to take Apple’s word for it, but if you reject the cross-platform benchmarks that exist, what do you want? What evidence would convince you? If such evidence doesn’t exist, I guess...
Apple is claiming to have the most efficient and powerful mobile CPUs on the market, so why don't we have the same level of scrutiny?
You don’t have to take Apple’s word for it, but if you reject the cross-platform benchmarks that exist, what do you want? What evidence would convince you? If such evidence doesn’t exist, I guess you can just remain doubtful, but I don’t find the arguments against Geekbench convincing. It’s a huge effort to make fair, cross-platform, representative benchmarks, and I’m not surprised there isn’t any open source offering, because it’s not only hard work, it’s also not even very interesting work. So, unless someone writes a research grant proposal or something to fund that work, I’m not surprised there aren’t other offerings that have seen wide adoption.
The problem with shared codebases there is that it can encourage people to game the system and optimize to make the benchmark go smoother at the expense of general performance. Of course, if...
There should be a wide range of benchmarks and those should all be based on a shared codebase as much as possible.
The problem with shared codebases there is that it can encourage people to game the system and optimize to make the benchmark go smoother at the expense of general performance.
Of course, if everyone uses the same benchmark they end up doing that anyway just with a blackbox in the middle somewhere so it's kind of a wash.
I could have been clearer when I made that requirement. I meant that singular cross-platform benchmarks should have the same codebase, not that every benchmark should be using the same codebase....
I could have been clearer when I made that requirement. I meant that singular cross-platform benchmarks should have the same codebase, not that every benchmark should be using the same codebase.
Ideally, we would be running the exact same system software as well when running comparitive benchmarks, but that is not going to happen for a number of reasons.
Single core benchmarks on a mobile OS vs x86 isn't even comparing apples (no pun intended) to oranges, it's like comparing apples to to a multi-course meal. One is infinitely more complex and...
Single core benchmarks on a mobile OS vs x86 isn't even comparing apples (no pun intended) to oranges, it's like comparing apples to to a multi-course meal. One is infinitely more complex and doing a million other things at the same time. What freaking blind fanboy wrote that?
Here are some benchmarks that are as close to “apples to apples” as I know how to find (if you have suggestion of a better benchmark, I’m all ears): iMac (27-inch Retina Early 2019) Windows 10...
Here are some benchmarks that are as close to “apples to apples” as I know how to find (if you have suggestion of a better benchmark, I’m all ears):
No, the way that an iPhone is configured, it’s not going to best the highest end desktop components of today in single-core performance. But, if Apple designs a system with decent thermals and active cooling, it’s really not unfathomable that they can be competitive. If not today, then imminently. And, if they throw some more cores in, I don’t see why multi-core performance can’t scale up as well.
I think part of their long-term strategy is to just assume that most of the intensive multi-core compute tasks are going to increasingly be moved off to GPUs or some other dedicated card. Given...
No, the way that an iPhone is configured, it’s not going to best the highest end desktop components of today in single-core performance. But, if Apple designs a system with decent thermals and active cooling, it’s really not unfathomable that they can be competitive. If not today, then imminently.
I think part of their long-term strategy is to just assume that most of the intensive multi-core compute tasks are going to increasingly be moved off to GPUs or some other dedicated card. Given that Apple's view on intensive workloads assumes you're either running scientific models or doing high end video editing that seems like a safe assumption. It leaves normal people's intensive use cases out in the cold though, like playing games or having too many tabs open in Chrome (jk).
Yeah I assume Apple will continue to encourage external GPU enclosures for those scenarios. Just daisy-chain an additional enclosure for each 10 Chrome tabs you plan to open. /s
It leaves normal people's intensive use cases out in the cold though, like playing games or having too many tabs open in Chrome (jk).
Yeah I assume Apple will continue to encourage external GPU enclosures for those scenarios.
Just daisy-chain an additional enclosure for each 10 Chrome tabs you plan to open. /s
RIP Mac gaming. It was never actually a vital group, and it's been on life support after Catalina nuked everyone's back catalogs. But this is probably gonna kill Bootcamp too.
So are they going to be ARM? That seems risky (but I don't know anything about business or that much about computing, to being fair).
Well, Apple is about the only company to have successfully pulled off two different architecture transitions so far, so if anyone can do it, it's them.
Although I imagine that the Mac Pros will remain x86-64 for the time being (until they can get ARM chips with Xeon-level performance, which (to my layman's eye at least) requires a different set of optimizations than the ones made for the A series as is (performance per watt being a huge consideration in a phone chip), plus getting all the pro software ported to ARM as well (because an emulated solution, even on powerful hardware) isn't going to perform at the levels expected of a $50k pro workstation.
It'll be interesting to see regardless, given the great work they've done with their mobile processors so far.
Apple's CPUs are actually very competitive with x86 chips.
That's pretty interesting to see.
One thing I was thinking of, Apple could (with their enormous cash haul) purchase VIA and that way acquire an x86(64) license that works more along the lines of the license they have from ARM in order to produce the A series. They could bring more of their A series work to that, but keep compatibility with more existing software for pro machines at the least.
I doubt it'd ever happen, but it's fun to think about what could come of something like that.
I'm always skeptical when it comes to iPhone benchmarks. I have seen so many unfair comparisons, and since so much of Apple's iPhone CPUs are black boxes you can never be sure how real world performance is supposed to be, especially since there is essentially no cross-platform software available for iOS.
If you were wondering what I meant by unfairness, the last time I saw a comparitive benchmark between iPhone and any given Android handset, they tested for speed of h.264 encoding. What they completely missed was that one of them had a dedicated hardware encoder on them and the other did not. Of course one was going to blow the other away.
Are you familiar with benchmarks like Geekbench? AFAIK, they try really hard to keep things fair across platforms.
I have limited trust for geekbench. From what I can tell it is not open source so you don't know what is actually happening behind the scenes. In any case, they are still somewhat limited in usefulness as a metric because even though they advertise that their testing mirrors real usage, it is still a purely synthetic benchmark.
Being proprietary isn't necessarily bad in this context, but the fact that geekbench is frequently the only benchmark run for comparison. There should be a wide range of benchmarks and those should all be based on a shared codebase as much as possible.
When Intel makes an incremental improvement on their CPUs, reputable tech news outlets will run a battery of 10 or more benchmarks of various types. Apple is claiming to have the most efficient and powerful mobile CPUs on the market, so why don't we have the same level of scrutiny?
You don’t have to take Apple’s word for it, but if you reject the cross-platform benchmarks that exist, what do you want? What evidence would convince you? If such evidence doesn’t exist, I guess you can just remain doubtful, but I don’t find the arguments against Geekbench convincing. It’s a huge effort to make fair, cross-platform, representative benchmarks, and I’m not surprised there isn’t any open source offering, because it’s not only hard work, it’s also not even very interesting work. So, unless someone writes a research grant proposal or something to fund that work, I’m not surprised there aren’t other offerings that have seen wide adoption.
Like I said, I don't outright reject geekbench, I just refuse to take one singular benchmark as the word of God. I'm doubtful, not unbelieving.
The problem with shared codebases there is that it can encourage people to game the system and optimize to make the benchmark go smoother at the expense of general performance.
Of course, if everyone uses the same benchmark they end up doing that anyway just with a blackbox in the middle somewhere so it's kind of a wash.
I could have been clearer when I made that requirement. I meant that singular cross-platform benchmarks should have the same codebase, not that every benchmark should be using the same codebase.
Ideally, we would be running the exact same system software as well when running comparitive benchmarks, but that is not going to happen for a number of reasons.
Single core benchmarks on a mobile OS vs x86 isn't even comparing apples (no pun intended) to oranges, it's like comparing apples to to a multi-course meal. One is infinitely more complex and doing a million other things at the same time. What freaking blind fanboy wrote that?
Here are some benchmarks that are as close to “apples to apples” as I know how to find (if you have suggestion of a better benchmark, I’m all ears):
iMac (27-inch Retina Early 2019)
Windows 10 machine with the same Core i9 as the iMac, for reference
iPhone 11 Pro Max
So, iPhones are competitive with desktop offerings from Intel in single-core performance since last year.
No, the way that an iPhone is configured, it’s not going to best the highest end desktop components of today in single-core performance. But, if Apple designs a system with decent thermals and active cooling, it’s really not unfathomable that they can be competitive. If not today, then imminently. And, if they throw some more cores in, I don’t see why multi-core performance can’t scale up as well.
I think part of their long-term strategy is to just assume that most of the intensive multi-core compute tasks are going to increasingly be moved off to GPUs or some other dedicated card. Given that Apple's view on intensive workloads assumes you're either running scientific models or doing high end video editing that seems like a safe assumption. It leaves normal people's intensive use cases out in the cold though, like playing games or having too many tabs open in Chrome (jk).
Yeah I assume Apple will continue to encourage external GPU enclosures for those scenarios.
Just daisy-chain an additional enclosure for each 10 Chrome tabs you plan to open. /s