iPad Ipad A10 CPU 2x faster than iPhone 11 A13 chip with DreamLab

[ericinboston]

Hi all! I run an app called DreamLab that looks for cures for Cancer and COVID-19. I am running the app on an iPad with the A10 Fusion chip and also on the iPhone 11 with the A13 chip.

I noticed, shockingly, that the old iPad crunches DreamLab work 2x faster than the iPhone 11! As far as I can think, there can only be 2 reasons for this:

1)DreamLab is somehow more tuned to the older A10 chip and/or the Fusion chip does something 2x better than the A11 Bionic chip/architecture.

2)Apple is throttling the iPhone 11 CPU.


Does anyone have an explanation for this?

Thanks for your input!

p.s. Is anyone running DreamLab on the newest iPad 8th gen? If so, how many calculations are you getting per hour? My iPad is getting about 8 per hour for the COVID-19 project.

 

[AutomaticApple]

2)Apple is throttling the iPhone 11 CPU.

There’s been no word so far on Apple throttling iPhone 11 CPUs.

 

[Starscape]

This is probably a situation where DreamLab just happens to be better optimized for the A10 and therefore performs very well with it. This is the case with most applications, where optimizations for specific hardware will make a big difference in performance. I wouldn't expect Apple to be throttling the iPhone 11 unless it's overheating or the battery is low or running at reduced output. Just try other apps and compare performance between your iPad and iPhone and you'll have a better idea of relative performance.

Also, make sure your iPhone is not running a lot of apps concurrently, which would also slow it down.

 

[ericinboston]

This is probably a situation where DreamLab just happens to be better optimized for the A10 and therefore performs very well with it. This is the case with most applications, where optimizations for specific hardware will make a big difference in performance.

Well, Apple stated back in 2017: "A11 Bionic, the most powerful and smartest chip ever in a smartphone, features a six-core CPU design with two performance cores that are 25 percent faster and four efficiency cores that are 70 percent faster than the A10 Fusion,"

And the Bionic chip supposedly includes/builds on the Fusion architecture. So why the complete opposite performance of the A10? I am sure DreamLab was not optimized for the 3+ year old CPU because DreamLab runs on a few platforms and DreamLab updates itself pretty regularly on the App Store. The devil is in the details but Apple really shouldn't claim such bold performance metrics when I have an app that specifically puts the chip through a lot of calculations yet it performs 2x better on 3+ year old architecture.

I may try purchasing an 8th gen iPad with the A12 chip and see how it compares to the A10 chip. My goal is to buy a new(er) iOS device that will be faster than the current A10 iPad for a reasonable price. If the new iPad for $299 (sale price at many locations) will be significantly faster, I'll buy that.

 

[Moyapilot]

I think its a heat throttling issue. All phones, due to their tiny size in comparison to the large ipad, suffer from this to various degrees. They can be fast in short bursts, but ipads will win out in longer sustained tests.

The entire aluminum casing of the ipad and iphone is basically acting like a heatsink.

 

[rui no onna]

Is it an actual A10 (iPad 6th & 7th gen) or is it A10X (iPad Pro)? Most distributed computing projects make extensive use of the GPU and afaik, A10X beats even the A12 on that front.

Check out Metal score comparison here:

iOS Benchmarks - Geekbench

 

[rui no onna]

Okay, installed DreamLab on a few iPads:

iPad Air 3 (A12)
iPad Pro 12.9 (A10X)
iPad Pro 10.5 (A10X)
iPad 6th gen (A10)

Interestingly enough, at the 40 minute mark, the A10-based iPads have completed 5 calculations each while the Air 3 had only completed 2.

I do wonder if perhaps the scheduler on the A12 considers DreamLab to be a light workload and shunting it to the low power cores. Another possibility is the Apple Bionic chipsets might be getting assigned more complex calculations. Wish I had one of those multi-port charges with display showing power consumption on each port.

 

[Starscape]

Well, Apple stated back in 2017: "A11 Bionic, the most powerful and smartest chip ever in a smartphone, features a six-core CPU design with two performance cores that are 25 percent faster and four efficiency cores that are 70 percent faster than the A10 Fusion,"

And the Bionic chip supposedly includes/builds on the Fusion architecture. So why the complete opposite performance of the A10? I am sure DreamLab was not optimized for the 3+ year old CPU because DreamLab runs on a few platforms and DreamLab updates itself pretty regularly on the App Store. The devil is in the details but Apple really shouldn't claim such bold performance metrics when I have an app that specifically puts the chip through a lot of calculations yet it performs 2x better on 3+ year old architecture.

I definitely understand but again, generally, the software needs to be coded to support hardware features, which includes the number of CPU cores and other features. I‘m not saying that’s what is taking place here, but it’s possible. I’m a former developer myself, though not on mobile (N64 and PC), so I can only speak from my personal experience. I don’t know anything about DreamLab, so I have no idea how old the codebase is. Also remember that it’s crucial the optimizations are made especially for lower-end or older devices so that software could run on the widest array of devices as possible.

 

[ericinboston]

Okay, installed DreamLab on a few iPads:

iPad Air 3 (A12)
iPad Pro 12.9 (A10X)
iPad Pro 10.5 (A10X)
iPad 6th gen (A10)

Interestingly enough, at the 40 minute mark, the A10-based iPads have completed 5 calculations each while the Air 3 had only completed 2.

I do wonder if perhaps the scheduler on the A12 considers DreamLab to be a light workload and shunting it to the low power cores. Another possibility is the Apple Bionic chipsets might be getting assigned more complex calculations. Wish I had one of those multi-port charges with display showing power consumption on each port.

Thank you very much, rui no onna, for your details!!

One item I forgot about was that a lot of distributed computing projects (like BOINC and World Community Grid) send different size/complex work units depending on the hardware of the machine.

Unfortunately DreamLab has 0 Support and communities so there is no way to ask these types of technical questions.

As a contributor to DreamLab, it's not about the calculations I have completed, but rather the % of help I have given to the project(s). If an A10 process completes 8 an hour and an A14 completes 2 per hour, yet the software is giving the A14 a "work unit" a far, far more complex than the one it is giving the A10 machine, then the whole statistics of counting how many Calculations I have done is pointless.

If anyone at DreamLab is reading this thread, DL should really come up with a better mechanism of how I am contributing to the project using various machines.

On a side note, it would be nice if anyone reading this thread installs DreamLab and lets it crunch as much as possible. My A10 iPad has crunched 6200+ Calculations over the past 40 days (running 24x7) for the COVID-19 project. My hope is to buy a few iPads (maybe even A10 to keep it consistent) and support DreamLab as much as I can.

 

[ericinboston]

Also remember that it’s crucial the optimizations are made especially for lower-end or older devices so that software could run on the widest array of devices as possible.

True...but DreamLab has a pretty bad track record of supporting older devices. When I installed it 2 years ago it supported all the projects...then after a DreamLab update, I had to update my iOS so I could run some other, newer, DreamLab projects (very annoying). And this happened again this year with the COVID-19 project and I was not able to update some of my iOS devices because (as the trickle-down-effect works) would require me to upgrade to a major version of iTunes which in turn would require a ton of work and re-setting defaults for all my music (50,000+ songs), settings, sharing, etc. My bank does the same thing with their app: update it every bleeping 2 months and often requires me to update iOS or it refuses to work. No. I'm not going to update my iOS every few months just to make 1 app happy while also creating me hours of anxiety and work to see what else trickles out.

It's very sad DreamLab doesn't have any kind of Support team or Community to get answers (and give feedback while also receiving acknowledgement of that feedback).

DreamLab has the awesomeness of being new, cool, and hip (unlike World Community Grid) but lacks the community of basic software support (especially when 500,000+ users are helping the project(s)).

 

[rui no onna]

On a side note, it would be nice if anyone reading this thread installs DreamLab and lets it crunch as much as possible. My A10 iPad has crunched 6200+ Calculations over the past 40 days (running 24x7) for the COVID-19 project. My hope is to buy a few iPads (maybe even A10 to keep it consistent) and support DreamLab as much as I can.

Why iPad? An iPhone 7 will probably do the job just as well and likely consume less power and space.

 

[ericinboston]

Why iPad? An iPhone 7 will probably do the job just as well and likely consume less power and space.

A few answers:

1)I have an iPhone 7 Plus...the problem is about the upgrading of iOS on my phone whenever DreamLab has a major new project. I don't want to do it on MY PHONE. I also am not worried about an iPad size vs. iPhone 7 size.

2)However, I could try to find some iPhone 7 devices (do you know of a good place to buy?).

But, ultimately it comes down to 3 things: a)cost, b)performance for the cost, c)reselling of the machine 2-3 years from now. If I buy a 4+ year old A10 chip iOS device now, am I going to be able to even sell it 3 years from now? Who would want it? If I can find a great 8th gen iPad on sale for $279 or $299 and buy 10 of them, they will last years and can be re-used for other things if I don't want to use DreamLab any more or whatever.

My employer sells old devices and has the iPhone 7 for $139. Sure, compared to an older iPad A10's current price that's a pretty good price right now for my project of building a DreamLab farm, but, is it more advantageous for me to spend 2x the price on a brand spanking new model iPad? Does the 8th gen iPad churn out 2x the amount of work (not "Calculations") for DreamLab? That's the question that I don't think I will ever get answered...therefore I would lean towards buying something brand new, luckily during holiday season, that is much newer in case I want to re-purpose it 1-3 years from now.

On a side note, I am eyeing an iPad Air 4th gen all for myself, regardless of DreamLab. Hoping for a great sale in a few weeks.

 

[rui no onna]

A few answers:

1)I have an iPhone 7 Plus...the problem is about the upgrading of iOS on my phone whenever DreamLab has a major new project. I don't want to do it on MY PHONE. I also am not worried about an iPad size vs. iPhone 7 size.

2)However, I could try to find some iPhone 7 devices (do you know of a good place to buy?).

But, ultimately it comes down to 3 things: a)cost, b)performance for the cost, c)reselling of the machine 2-3 years from now. If I buy a 4+ year old A10 chip iOS device now, am I going to be able to even sell it 3 years from now? Who would want it? If I can find a great 8th gen iPad on sale for $279 or $299 and buy 10 of them, they will last years and can be re-used for other things if I don't want to use DreamLab any more or whatever.

My employer sells old devices and has the iPhone 7 for $139. Sure, compared to an older iPad A10's current price that's a pretty good price right now for my project of building a DreamLab farm, but, is it more advantageous for me to spend 2x the price on a brand spanking new model iPad? Does the 8th gen iPad churn out 2x the amount of work (not "Calculations") for DreamLab? That's the question that I don't think I will ever get answered...therefore I would lean towards buying something brand new, luckily during holiday season, that is much newer in case I want to re-purpose it 1-3 years from now.

You mentioned getting A10 iPads for consistency on your previous post that's why I mentioned the iPhone 7. The iPhone 7 can be more easily found cheaply than A10 iPads at this point in time.

The iPad 8th gen has the same A12 chipset and 3GB RAM as the Air 3 so we already know its performance. What the numbers actually mean as far as getting the work done, only the DreamLab devs know. I'm trying it now on an iPad 5th gen (A9) and that seems to be finishing calculations at the same speed as the A10.

After 15 hours, the A10-based iPads have completed 105 calculations while the A12 Air 3 has only completed 35. They're all running pretty cool so I'm guessing the DreamLab app is limiting CPU usage intentionally.

Again, we don't know if work complexity is the same. On the off chance it is though, I'd hedge my bets and go with a combo of A10 and A12 devices. Heck, might even be worth getting A9 if you can get those really cheap. I've seen the 1st gen SE go for $50-60. 3x work done at 1/5-1/3 cost seems like a better proposition to me. Assuming the work units are the same, just ~3x of those iPhone 7s at $139ea will do the same job as ~10x iPad 8th gens at $299ea.

 

[Ungibbed]

Wouldn’t the A10X perform better? ?

To me this sounds like the old 100 MHz 486 DX4 chips performing integer calculations better than a comparable Pentium CPU during that time.

 

[rui no onna]

Wouldn’t the A10X perform better? ?

To me this sounds like the old 100 MHz 486 DX4 chips performing integer calculations better than a comparable Pentium CPU during that time.

I've tested on iPad 6th gen, and 2017 Pro 10.5 and 12.9. Both A10 and A10X complete the same number of calculations for the DreamLab app.

 

[Ungibbed]

I've tested on iPad 6th gen, and 2017 Pro 10.5 and 12.9. Both A10 and A10X complete the same number of calculations for the DreamLab app.

Sorry, I wasn’t very clear in my previous post. I was trying to compare the the newer Apple silicon in iPhone/iPad to the A10 as a reference to the old 486 screamers outperforming similar clocked Pentium processors on PC (ages ago...)

So, the stock A10 is faster than the A10X in this app, have you tried older devices and different iOS versions? This is just off the wall but not the first time I’ve seen such a thing happen.

 

[rui no onna]

Sorry, I wasn’t very clear in my previous post. I was trying to compare the the newer Apple silicon in iPhone/iPad to the A10 as a reference to the old 486 screamers outperforming similar clocked Pentium processors on PC (ages ago...)

So, the stock A10 is faster than the A10X in this app, have you tried older devices and different iOS versions? This is just off the wall but not the first time I’ve seen such a thing happen.

Note, A10 is not faster than A10X in this app. They're the same speed. A10 is a wee bit faster than A9 though, albeit the higher clocked A9X might be the same speed. The iPads barely get warm so I believe the app is not really stressing the chipsets to their limits.

Based on results, the firmware version doesn't appear to have any impact. iOS 13 is the earliest supported for the COVID-19 study.

After 15 hours:
iPad Pro 12.9 (A10X, iOS 14): 105
iPad Pro 10.5 (A10X, iOS 13): 105
iPad 6th gen (A10, iOS 14): 105
iPad Pro Air 3 (A12, iOS 13): 35

I wouldn't be surprised if the way the app works is it's just using 1 high power core on A9, A10 and A10X while using 1 low power core on A11 Bionic or newer.

 

[ericinboston]

Any chance anyone here has some results to share? Rui no onna are you still running DreamLab?

Anyone have some time to run DreamLab on the new iPad or iPad Air?

I'd like to see how it runs on the new M1 Macbook Air but I have a feeling it doesn't cool very well and thus not worth spending 2x-3x the cost on a Macbook when the iPad may be better price/performance ratio.