First A14 Benchmarks from iPad Air 4 Appear Online

[TheMacDaddy1]

It is quite risky comparing a small powered cpu with small set of instructions to a 45~125 watt chip with a lot more instructions (remember atom series with micro ops that is only run at half the performance of their counterpart with CISC at the same clockspeed). I wouldn't be surprised if the next mac with Apple Silicon would run some softwares (including benchmarks) better than the Intel one, while the other run worse.

Regarding the DTK, it is strange that i haven't been able to found benchmark result other than geekbench score. Which is not a good benchmark for a sustained workload, the kind of thing that traditional computer do best compared to phone or tablet. We need a lot more than that to measure real world performance.

This! Comparing geekbench on a iPhone, which is just a burst of CPU usage for a very short time does not mean it is more powerful than the top end Intel Laptop chips at 45w. Render a video for 2 hours on that A14 and then compare the too. Using the iPhone or iPad it would melt or shut down. The Mac ARM is probably a much bigger ARM chip....ARM really just being an instruction set so comparing phone CPU's to full on computer CPU's is not and Apples to Apples comparision.

 

[cmaier]

Don’t know what you are going on about. There are two fast cores and four low-power slow cores. Of course they run simultaneously. Why wouldn’t they?

Believe it or not, on a lot of multi-core designs, there are limitations as to which cores can run simultaneously and how many can run simultaneously. This is done for several reasons, but primarily: 1) power/thermal. Those designs cannot supply sufficient power to power all cores at once, or they need to power down cores periodically to allow thermal hotspots to dissipate. 2) resource contention. Buses to things like shared caches and the outside bus in these designs are frequently limited and allowing all cores to simultaneously use them would cause contention issues that would slow the machine well beyond whatever is gained by allowing the cores to compute simultaneously. The main issue in most of these cpus, however, is (1), at least from my experience.

Of course, the A-series chips don’t have these problems, at least when used in iphones and ipads.

 

[PPietra]

In that case, it seems you're agreeing with the point I've been making all along, which is that Geekbench scores on iOS can't be used (except in the roughest of senses) to predict real-world performance on MacOS.

After all, if they could, then we would indeed expect the SC (single-core) performance of the AS chip in the first Mac laptops to beat the SC performance of the fastest Intel chips—because, according to GB, even the iPad's A14 is already beating the fastest (non-overclocked) Intel chips: 1583 for the A14, vs. 1474 for the 2.8 GHz/4.7 GHz Intel Core i7-1165G7 Tiger Lake (which beats the 1413 of the 3.7 GHz/5.3 GHz Intel Core i9-10900K Comet Lake).

In other words (again, talking about SC here): The iPad A14 already beats fastest Intel chip on GB. The fact that you don't expect the first AS Mac laptop (which, for thermal reasons, should be faster still) will actually be faster than the fastest Intel chip is tantamount to acknowledging GB scores can't be used to predict real-world performance (at least in this context).

Given this, why do you keep quoting the GB scores as predictors of performance? [I do mention GB scores as well, but always (or nearly always) with the caveat that they may not mean much.]

Yes, the fastest next-gen Intel desktop chip (Rocket Lake) will certainly be faster than that U-series Tiger Lake laptop/ultrabook chip. But that won't release until late 2021, a year after the first AS Mac laptop, so it won't even be available for comparison. And even if it were, the % increase for the AS chip, in going from an iPad to a Mac laptop, should be at least as great.

[For those who may not be familiar with the model designations, I just looked this following up. Please correct me if I'm wrong: The i9-10900K has the fastest SC performance of any 10th-generation Intel chip. The 1165G7 Tiger Lake is a low-powered (U-series) laptop/ultrabook chip from Intel's upcoming 10 nm 11th-generation laptop chips. It will be offered with different TDPs and, at that base frequency, the TDP should be ~25W.]

I don’t think that is what he said!
He was careful to only make the comparison with the Intel chips that Apple uses, to avoid generalising, since there are results with higher single core GB scores for Intel chips.
Geekbench should give identical scores wether it runs iOS or macOS. We have already seen this with the Apple developer transition kit (A12Z) when it runs Geekbench natively - similar score to the iPad Pro.

 

[chucker23n1]

you mentioned in your other post that A14 doesn’t have 6 cores 🤦‍♂️ (You seriously said something that stupid) despite you also saying the cores are divided into 2 high performance cores and 4 high efficiency cores which is basically 6 cores

But two performance cores rather than six matters. The multicore scores would be quite different.

 

[chucker23n1]

This! Comparing geekbench on a iPhone, which is just a burst of CPU usage for a very short time does not mean it is more powerful than the top end Intel Laptop chips at 45w.

It is, though. Why would you think giving it nine times(!!) as much thermal headroom wouldn’t also allow it for sustain perf for much longer?

 

[cmaier]

I don’t think that is not what he said!
He was careful to only make the comparison with the Intel chips that Apple uses, to avoid generalising, since there are results with higher single core GB scores for Intel chips.
Geekbench should give identical scores wether it runs iOS or macOS. We have already seen this with the Apple developer transition kit (A12Z) when it runs Geekbench natively - similar score to the iPad Pro.

Yep, that’s what i was careful to say.

 

[PPietra]

This! Comparing geekbench on a iPhone, which is just a burst of CPU usage for a very short time does not mean it is more powerful than the top end Intel Laptop chips at 45w. Render a video for 2 hours on that A14 and then compare the too. Using the iPhone or iPad it would melt or shut down. The Mac ARM is probably a much bigger ARM chip....ARM really just being an instruction set so comparing phone CPU's to full on computer CPU's is not and Apples to Apples comparision.

There are people doing that (video rendering) on the iPad Pro without a problem.

 

[theorist9]

And you yourself are predicting the first Mac AS chips, even the laptop chips, will have significantly faster single-core performance than even the fastest Intel desktop chips, correct? Here is some of your earlier language, though I don't know which AS chips you are referring to here:

I don’t think that is what he said!
He was careful to only make the comparison with the Intel chips that Apple uses, to avoid generalising, since there are results with higher single core GB scores for Intel chips.
Geekbench should give identical scores wether it runs iOS or macOS. We have already seen this with the Apple developer transition kit (A12Z) when it runs Geekbench natively - similar score to the iPad Pro.

You don't understand what I wrote. Based on GB, even the A14 in the iPad is faster, per-core, than the fastest current Intel desktop chip, period. The fact that he doesn't think *actual* per-core performance of the upcoming AS Mac laptop (which will be even faster than the iPad A14) will exceed that of the fastest desktop chip is, effectively, an acknowledgedment that GB scores can't be used to predict peformance in this context.

I.e.:

According to GB: A14 iPad is, per-core, faster than fastest Intel chip
According to cmaier: Actual per-core performance of first AS Mac laptop (which will be even faster than the iPad) will be slower than the fastest Intel chip.

Conclusion: cmaier is effectively acknowledging GB scores can't be used to predict actual performance (in this context).

 

[PPietra]

You don't understand what I wrote. Based on GB, even the A14 in the iPad is faster, per-core, than the fastest current Intel desktop chip, period. The fact that he doesn't think *actual* per-core performance of the upcoming AS Mac laptop (which will be even faster than the iPad A14) will exceed that of the fastest desktop chip is, effectively, an acknowledgedment that GB scores can't be used to predict peformance in this context.

read his new comment!
There are current Intel desktop chips with faster single core scores, you just need to look for it.
But even if there weren’t any chips, any careful person wouldn’t assume that there weren’t, since it is not even necessary for the argument.

 

[theorist9]

Yep, that’s what i was careful to say.

Logically, it's precisely because you needed to make that qualifier that you were effectively acknowledging GB scores can't be used to predict real-world performance in this context. If they could be used in that way, no qualifier would be needed.

Recall I've said things in the past you didn't follow until I re-explained them (not laying blame here, maybe I didn't do a good job explaining the first time). I think that's the case here. Understanding this point is going to require a careful (as opposed to casual) read of what I wrote.

It's perfectly fine to disagree with me (many people here do ), but I can tell from your reply that this isn't you disagreeing with what I wrote, this is you not yet understanding what I wrote.

 

[theorist9]

read his new comment!
There are current Intel desktop chips with faster single core scores, you just need to look for it.
But even if there weren’t any chips, any careful person wouldn’t assume that there weren’t, since it is not even necessary for the argument.

Nope, you're still not understanding. And neither is he. There is a logical inconsistency between using GB scores as though they are predictors of real-world performance, and realizing you need to add that strong qualifier (i.e., that strong restriction) when talking about real-world predictions.

Not trying to criticize you or him here. What I'm saying is subtle. You need to read what I wrote carefully, and think about it carefully. I think you are both doing a quick scan and disagreeing without really understanding what I am writing.

And I already did look through the GB Processor Benchmark. The SC scores I quoted were the fastest SC scores listed there. Are there any Intel chips you believe have faster SC performance that aren't among those listed here?:

Processor Benchmarks - Geekbench

[You may be referring to the system benchmark scores, but they're not reliable indicators of actual GB CPU performance, because they're filled with strange results (perhaps from overclocking). For instance, according to that table, the fastest SC CPU score (2000) belongs to a 9th-gen Intel Core i7-9700KF, whose stock configuration is 3.6 GHz/4.9 GHz. That chip's stock SC performance is certainly less than that of the 10-th gen 3.7 GHz/5.3 GHz Intel Core i9-10900K.]

 

[chucker23n1]

Nope, you're still not understanding. And neither is he. There is a logical inconsistency between using GB scores to predict real-world scores,

What qualifies as “real-world”?

 

[Homme]

But two performance cores rather than six matters. The multicore scores would be quite different.

obviously yes. 6 high performance cores is overkill even for the A14. I’m not sure if even the CPU itself could survive on 6 high performance cores unless it adopted the cooling technology from the X line of SoC. That’s why the low performance cores matter and because all the cores run simultaneously it will show up as 6 on Geekbench (A11 and up) compared to 2 for the A10 and 3 on the A10X

 

[theorist9]

What qualifies as “real-world”?

It's whatever he was thinking of as real-world when he wrote "I don't think they will beat the *fastest* intel chips".

 

[chucker23n1]

obviously yes. 6 high performance cores is overkill even for the A14. I’m not sure if even the CPU itself could survive on 6 high performance cores unless it adopted the cooling technology from the X line of SoC. That’s why the low performance cores matter and because all the cores run simultaneously it will show up as 6 on Geekbench (A11 and up) compared to 2 for the A10 and 3 on the A10X

I think that’s and increasingly relevant falling of Geekbench, though. It should slow specs of individual cores.

It's whatever he was thinking of as real-world when he wrote "I don't think they will beat the *fastest* intel chips".

So, no answer?

 

[cmaier]

It's whatever he was thinking of as real-world when he wrote "I don't think they will beat the *fastest* intel chips".

To be clear, and probably my fault, when I provided that answer I didn‘t have a definition of “real world” in mind, nor do I think there is one that works uniformly. For some workloads, then yes, Axx will beat everything Intel‘s got. For others, the highest end intel stuff will beat Axx. Which workloads are which are not clear to me yet, but I’m sure Apple is has targeted the workloads it thinks drive purchases for each product category. So, for example, MacBooks would likely be office, photo processing, audio, etc. MBP’s/iMacs would be video editing, Xcode, etc. MP would be high end video processing, etc. Things outside what Apple is targeting, e.g. HPC, scientific computing that relies on FP, etc., would likely be superior on high end intel chips (of the sort Apple doesn’t tend to put in its machines anyway)

 

[EugW]

This! Comparing geekbench on a iPhone, which is just a burst of CPU usage for a very short time does not mean it is more powerful than the top end Intel Laptop chips at 45w. Render a video for 2 hours on that A14 and then compare the too. Using the iPhone or iPad it would melt or shut down.

Funny you should mention that, since most video encoding on the iPad is hardware accelerated. In this case, power usage is quite low. Furthermore, Apple’s iDevice hardware video encoder seems to be very robust (esp. on the X variant chips) and is faster than a lot of Intel Macs.

 

[PPietra]

Nope, you're still not understanding. And neither is he. There is a logical inconsistency between using GB scores as though they are predictors of real-world performance, and realizing you need to add that strong qualifier (i.e., that strong restriction) when talking about real-world predictions.

Not trying to criticize you or him here. What I'm saying is subtle. You need to read what I wrote carefully, and think about it carefully. I think you are both doing a quick scan and disagreeing without really understanding what I am writing.

And I already did look through the GB Processor Benchmark. The SC scores I quoted were the fastest SC scores listed there. Are there any Intel chips you believe have faster SC performance that aren't among those listed here?:

Processor Benchmarks - Geekbench

[You may be referring to the system benchmark scores, but they're not reliable indicators of actual GB CPU performance, because they're filled with strange results (perhaps from overclocking). For instance, according to that table, the fastest SC CPU score (2000) belongs to a 9th-gen Intel Core i7-9700KF, whose stock configuration is 3.6 GHz/4.9 GHz. That chip's stock SC performance is certainly less than that of the 10-th gen 3.7 GHz/5.3 GHz Intel Core i9-10900K.]

It doesn’t matter if it is overclocking or not... once you know results exist and without proof of otherwise, no one should assume in an argument that there aren’t any faster chips. Classifying as the "best", "faster" than any other, is a rabbit hole... As such you are making way too many assumptions about what someone says, that have nothing to do with what people are saying.
Synthetic Benchmarks have their problems but so do real world performance measurements, because you can not encompass the general potential performance with just some real world performance measurements, since computing needs are extremely varied - that is why synthetic benchmarks are developed, even if we can argue about their quality. In truth a good evaluation should be an extensive mixture of both, but we don’t have them yet, so we can only speculate with geekbench.

 

[PPietra]

"I don't think they will beat the *fastest* intel chips".

He didn’t write exactly that did he!!? His following phrase completely changes the context and meaning...

 

[chucker23n1]

It doesn’t matter if it is overclocking or not...

Depends on the purpose of the argument. Was there an alternate reality in which Apple doesn’t switch to ARM and, before Intel scales 10nm to the desktop, sells an iMac with an overclocked CPU? I dunno about that.

 

[PickUrPoison]

No

Maybe not with geekbench, which doesn’t run long enough tests to show the effects of thermal throttling.

 

[EugW]

Maybe not with geekbench, which doesn’t run long enough tests to show the effects of thermal throttling.

Exactly. However, fortunately for the iPhone, most workloads on the iPhone are bursty. For some that aren’t, like video encoding, it usually still doesn’t matter since it’s usually hardware accelerated. Power usage is low and even an iPhone won’t throttle with this type of usage even after extended periods. There are of course several situations where the iPhone will throttle, but in those situations an iPad will often too, just after a longer period of time. However, there are more situations where an iPad will be at risk of throttling, because an iPad is more often used for content creation than an iPhone would be.

Geekbench 5 isn’t exactly just bursty but it doesn’t run long enough to cause heavy throttling most of the time, at least if the device isn’t in a big thick case and/or charging at the same time in a warm room.

 

[iMi]

Wrong thread?

Wow... you know you’re old when... Yeah, wrong thread. Thanks for pointing it out you whippersnapper you. 😂

 

[theorist9]

To be clear, and probably my fault, when I provided that answer I didn‘t have a definition of “real world” in mind, nor do I think there is one that works uniformly. For some workloads, then yes, Axx will beat everything Intel‘s got. For others, the highest end intel stuff will beat Axx. Which workloads are which are not clear to me yet, but I’m sure Apple is has targeted the workloads it thinks drive purchases for each product category. So, for example, MacBooks would likely be office, photo processing, audio, etc. MBP’s/iMacs would be video editing, Xcode, etc. MP would be high end video processing, etc. Things outside what Apple is targeting, e.g. HPC, scientific computing that relies on FP, etc., would likely be superior on high end intel chips (of the sort Apple doesn’t tend to put in its machines anyway)

Thanks for the clarification. If you're saying that, for the types of apps one would generally expect to be used on a MacBook, you expect the SC performance of the first AS MacBook will (at least typically) be faster than *anything* Intel's got, not merely faster than the comparable Intel laptop chips, then there is no inconsistency with using GB as a predictor.

If so, that would be an exciting prediction. It would certainly blow the computer industry away if it comes to pass.

Moving to the AS MBP and iMac, and regarding their FP performance: While Apple isn't targeting HPC scientific computing, the Mac does have a strong presence among scientists, who use Macs not only for office work, but also for desktop scientific computing, e.g., with apps like Mathematica and Matlab, as well as for prototyping. Apple recognizes this, giving these apps equal prominence, in its marketing of the MBP and iMac, with "creative" apps like Autodesk, Logic Pro, Photoshop, etc. (see screenshot below). Many Mathematica/Matlab/prototyping calculations are FP, so it will be disappointing if this is a weak area for AS.

As an interesting historical note: When Apple introduced the dual-2.0 GHz G5 PowerPC at WWDC 2003, Steve Jobs claimed Mathematica was 2.3 x faster on the PPC than on the fastest PC chip—a dual 3.06 Xeon. However, when I later ran Mathematica timing tests myself, I found it was typically 20% - 40% slower on my G5 than on my plain-vanilla PC (2.8 GHz Pentium IV). When I discussed this with a contact at Wolfram (the maker of Mathematica), he said the performance comparison was based on "a pretty specific function, large integer multiplication", where the G5 is faster. I.e., Apple was only able to make the performance claim because it completely cherry-picked the data.

 

[theorist9]

He didn’t write exactly that did he!!? His following phrase completely changes the context and meaning...

Nope, you still don't understand. Honestly, at this point, this is one of those things that could only be explained by a face-to-face discussion—which is one of the general limitations of forums. The only other thing I could suggest would be to find a friend who is both a native English speaker (or has very high fluency) and has a PhD in the hard sciences with a significant publication record (the rigor of getting a PhD and publishing in those fields provides good training in parsing language), and ask him or her to look over the discussion chain and explain it to you. I'm sorry I can't do any better.

I recognize the above may seem a bit presumptuous, but please remember that you're the one that challenged my understanding of the discussion. I've done my best to explain myself to you, without success.