Analyzing the iPhone 5 Geekbench Resultsby Anand Lal Shimpi on September 16, 2012 8:31 PM EST
- Posted in
- iPhone 5
While working on our Haswell piece, I've been religiously checking the Geekbench and GLBenchmark results browsers to see if anyone ran either benchmark and decided to tap upload. This usually happens before every major smartphone launch, but in the case of the iPhone 5 the details these applications can give us are even more important.
Yesterday we confirmed that Apple is using its own custom designed ARM based CPU cores in its A6 SoC. Apple opted not to design in a vanilla ARM Cortex A9 likely to avoid relying on pure voltage/frequency scaling to improve performance, and chose not to integrate a Cortex A15 likely because of power consumption concerns as well.
There's absolutely no chance of Apple sending us a nice block diagram of the A6 CPU cores, so we have to work with what clues we can get elsewhere. Geekbench is particularly useful because it reports clock speed. Why does clock speed matter? Because, if reported accurately, it can tell us a lot about how the A6's CPU design has improved from an IPC standpoint. Remember that clock speed doesn't matter, but rather the combination of clock frequency and instructions executed per clock that define single threaded performance.
|Apple iPhone 5 Models|
|iPhone 5 Model||GSM/EDGE Bands||WCDMA Bands||CDMA 1x/EVDO Rev.A/B Bands||LTE Bands (FCC+Apple)|
|A1428 "GSM"||850/900/1800/1900 MHz||850/900/1900/2100 MHz||N/A||2/4/5/17|
|A1429 "CDMA"||850/900/1800/1900 MHz||850/900/1900/2100 MHz||800/1900/2100 MHz||1/3/5/13/25|
|A1429 "GSM"||850/900/1800/1900 MHz||850/900/1900/2100 MHz||NA||1/3/5 (13/25 unused)|
A short while ago, Geekbench results for a device identifying itself as an iPhone5,2 appeared. Brian believes this is likely the A1429 Verizon device (A1428 being iPhone 5,1) - perhaps one presampled to a reviewer looking to test their luck.
MacRumors appears to be first on the scene, having been tipped by an employee at PrimateLabs (the creators of Geekbench).
I need to preface the rest of this post with a giant caution sign: I have no inside knowledge of whether or not these results are legitimate. They seem believable, but anything can happen. The rest of this post is simply my initial thoughts on what these mean, should the results be accurate. Update: The first iPhone 5 reviews are out and this Geekbench data looks accurate.
Cache sizes haven't changed, which either tells us Apple isn't feeling as generous with die size as perhaps it once was or that working sets in iOS are still small enough to fit inside of a 1MB L2. I suspect it's mostly the latter, although all microprocessor design is a constantly evaluated series of tradeoffs (often made through giant, awesomely protected spreadsheets).
The first real change is clock speed. Apple clocked its A4 and A5 CPU core(s) at 800MHz, although these Geekbench results point to a 25% increase in frequency at 1GHz. Some of the headroom is likely enabled by the move to 32nm, although it's very possible that Apple also went with a slightly deeper pipeline to gain frequency headroom. The latter makes sense. We've seen conservative/manageable increases in pipeline depth to hit frequency targets and improve performance before.
The fairly low clock speed also points to an increase in IPC (instructions executed per clock) over the Cortex A9 design. As I mentioned in our A6 analysis post, simple voltage/frequency scaling is a very power inefficient way to scale performance. A combination of IPC and frequency increases are necessary. If these results are accurate and the CPU cores are only running at 1GHz, it does lend credibility to the idea of a tangibly wider design.
It's also unclear if Apple is doing any sort of dynamic thermal allocation here, ala Intel's Turbo Boost. You can't get more power constrained than in a smartphone, and power gating is already common within ARM based SoCs, so that 1GHz value could be under load for both cores. A single core could run at higher frequencies for short bursts.
The next thing that stood out to me was the memory data:
|Memory Performance||iPhone 4S||iPhone 5 (unconfirmed)||Scaling|
|Read Sequential ST||0.32 GB/s||1.78 GB/s||5.63x|
|Write Sequential ST||0.86 GB/s||1.35 GB/s||1.57x|
|Stdlib Allocate ST||1.44 Mallocs/s||1.92 Mallocs/s||1.33x|
|Stdlib Write||2.7 GB/s||6.06 GB/s||2.24x|
|Stdlib Copy||0.55 GB/s||2.26 GB/s||4.13x|
|Stream Performance||iPhone 4S||iPhone 5 (unconfirmed)||Scaling|
|Stream Copy||0.42 GB/s||1.9 GB/s||4.55x|
|Stream Scale||319 MB/s||994 MB/s||3.11x|
|Stream Add||0.59 GB/s||1.39 GB/s||2.34x|
|Stream Triad||377 MB/s||1019 MB/s||2.70x|
It's well known that ARM's Cortex A9 doesn't have the world's best interface outside of the compute core and its memory performance suffered as a result. If this data is accurate, it points to significantly overhauled cache and memory interfaces. Perhaps an additional load port, deeper buffers, etc...
Also pay close attention to peak bandwidth utilization. The 4S had 6.4GB/s of theoretical bandwidth out to main memory, the 5 raises that to 8.5GB/s. In the Stdlib write test the 4S couldn't even hit 50% of that peak bandwidth. The iPhone 5 on the other hand manages to hit over 70% of its peak memory bandwidth. I will say that if these numbers are indeed faked, whoever faked them was smart enough not to violate reality when coming up with these memory bandwidth numbers (e.g. no 95% efficiency numbers show up). It's also clear that these results aren't a simply doubling across the board over the 4S, lending some credibility to them.
Some of the largest performance improvements promised by the Geekbench data appear here in the memory results. It's whatever work Apple did here that helped enable the gains in the integer and floating point results below:
|Integer Performance||iPhone 4S||iPhone 5 (unconfirmed)||Scaling|
|Blowfish ST||10.7 MB/s||23.4 MB/s||2.18x|
|Blowfish MT||20.7 MB/s||45.6 MB/s||2.20x|
|Text Compress ST||1.21 MB/s||2.79 MB/s||2.30x|
|Text Compress MT||2.28 MB/s||5.19 MB/s||2.27x|
|Text Decompress ST||1.71 MB/s||3.82 MB/s||2.23x|
|Text Decompress MT||2.84 MB/s||5.60 MB/s||2.67x|
|Image Compress ST||3.32 Mpixels/s||7.31 Mpixels/s||2.20x|
|Image Compress MT||6.59 Mpixels/s||14.2 Mpixels/s||2.15x|
|Image Decompress ST||5.32 Mpixels/s||12.4 Mpixels/s||2.33x|
|Image Decompress MT||10.5 Mpixels/s||23.0 Mpixels/s||2.19x|
|Lua ST||215.4 Knodes/s||455 Knodes/s||2.11x|
|Lua MT||425.6 Knodes/s||887 Knodes/s||2.08x|
On average we see around 2.2x scaling from the 4S to the 5 in Geekbench's integer tests. There's no major improvement in multicore scaling, confirming what Geekbench tells us that we're looking at a two core/two thread machine.
The gains here are huge and are likely directly embodied in the performance claims that Apple made at the iPhone 5 launch event. Many smartphone workloads (under Android, iOS and Windows Phone despite what Microsoft may tell you) are still very CPU bound. Big increases in integer performance will be apparent in application level improvements.
Don't be surprised to see greater than 2x scaling here even though Apple only promised 2x at the event. Remember that what you're looking at is raw compute tests without many of the constraints that apply to application level benchmarks. While Apple has used benchmarks in the past to showcase performance, all of its performance claims at launch were application level tests. Those types of tests are more constrained and will show less scaling. That being said, I am surprised to see application level tests that were so close to the 2.2x average scaling we see here. Apple could have moved to faster NAND/storage controller here as well, which could help most if not all of these situations.
|Floating Point Performance||iPhone 4S||iPhone 5 (unconfirmed)||Scaling|
|Mandelbrot ST||223 MFLOPS||397 MFLOPS||1.77x|
|Mandelbrot MT||438 MFLOPS||766 MFLOPS||1.74x|
|Dot Product ST||177 MFLOPS||322 MFLOPS||1.81x|
|Dot Product MT||353 MFLOPS||627 MFLOPS||1.77x|
|LU Decomposition ST||171 MFLOPS||387 MFLOPS||2.25x|
|LU Decomposition MT||348 MFLOPS||767 MFLOPS||2.20x|
|Primality ST||142 MFLOPS||370 MFLOPS||2.59x|
|Primality MT||260 MFLOPS||676 MFLOPS||2.59x|
|Sharpen Image ST||1.35 Mpixels/s||4.85 Mpixels/s||3.59x|
|Sharpen Image MT||2.67 Mpixels/s||9.28 Mpixels/s||3.47x|
|Blur Image ST||0.53 Mpixels/s||1.96 Mpixels/s||3.68x|
|Blur Image MT||1.06 Mpixels/s||3.78 Mpixels/s||3.56x|
The floating point benchmarks show "milder" scaling in the first few tests (sub-2x) but big scaling in the latter ones. My guess here is we're seeing some of the impacts of increased memory bandwidth at the end there. If you look at our iPhone 5 hands-on video you'll see Brian talking about how super fast the new flyover mode in iOS 6 Maps is on the 5 compared to the 4S. That's likely due in no small part to the improved memory interface.
Although Geekbench is cross platform, I wouldn't recommend using this data to do anything other than compare iOS devices. I've looked at using Geekbench to compare iOS to Android in the past and I've sometimes seen odd results.
I'm sure we'll learn a lot more about the A6 SoC over the coming days/weeks.
Post Your CommentPlease log in or sign up to comment.
View All Comments
MrAwax - Monday, September 17, 2012 - linkbtw, how could Samsung have made a "it doesn't take a genius" ad listing 1GB of RAM while no one has talked about this and this "credible but not confirmed" Geekbench score wasn't out yet ?
That's a pretty efficient firewall you have in place between your component and cellphone divisions, Samsung.
piroroadkill - Monday, September 17, 2012 - linkUh, sorry, there have been phones with dual core CPUs and 1GB RAM for a while.
There are even phones with 2GB RAM and a 4.3" display (Xiaomi Mi-Two).
agoyal - Sunday, September 16, 2012 - linkGive the numbers this 1 GHZ processor is performing at about 1750mhz level or about 1.75 times faster than A9. Krait is about 1.3 times as fast as A9. Is apple is able to achive this kind of performance at this power levels, this is nothing short of amazing. Wow!! And this is there 1st custom ARM chip.
kpb321 - Monday, September 17, 2012 - linkIt's different targets. The Krait or A15 processors are designed to run at higher speeds. Krait runs at 1-1.7 ghz so far and I wouldn't be surprised to see it hit 2ghz. The a15 is suppose to hit 1.5 ghz in phones and 2.5 ghz in server/router uses. Going higher mhz almost always means some loss in IPC if everything else remains roughly the same. Apple clearly wants to keep the clock speed lower on the A6 to keep the voltage down which lowers the power draw.
serversurfer - Monday, September 17, 2012 - linkHow do you know how high Apple plan to scale the clock in the A6? You seem to be assuming it'll be 1GHz forever. Maybe it'll be 1.5GHz in the iPad Mini, and 2.2GHz in the iPad 4. /shrug
Zink - Monday, September 17, 2012 - linkWow yeah with the big battery crammed into the new iPad it would be neat to see them go all out on CPU performance next gen and have twice the frequency and twice the memory bandwidth of the iPhone 5. Maybe a companion core set up to keep the high clock sections off most of the time?
The next gen iPads are probably just going to get thinner with 2x performance boost like the iPhone and iPods did though.
MrAwax - Monday, September 17, 2012 - linkWe'll maybe because it will burn too hot in the iPad and we'll get another warmgate.
The iPad 4 will be blazingly hot, like 80 freaking degrees (F).
serversurfer - Monday, September 17, 2012 - linkWell, I don't think it'll actually be 2.2GHz. My point was, we have no idea how high Apple intend to scale this thing. For that matter, we also don't know how much heat it produces. We do know that Apple said per-Watt performance was one of their primary design goals, which would have the knock-on effect of reducing heat output.
In any case, kpb321 points out that Krait is running at 1-1.7GHz, but then looks at the very first A6, running 1GHz, and assumes that at or near its top speed. That strikes me as a fairly baseless assumption. They bump the clock speed of the A5 by 25% when moving from the iPhone to the iPad. If they do the same with the A6, then you're looking at a score in the 2000 range, and that's with no other memory improvements, etc. (The iPads have a lot more memory bandwidth than the iPhones, mind.) And who knows if Apple's aggressive power saving measures will allow them to clock it even faster than that? At only 1.5GHz you'd be looking at scores of 2400 or so.
kyuu - Monday, September 17, 2012 - linkIt's not that interesting, really. Krait is clocked at 1.5GHz in recent phones, and at that clock speed it should equal if not beat the custom SoC in the iPhone 5 in terms of raw compute. Given that the rated battery life of, for example, the Nokia Lumia 920 is equal or better than the ratings for the iPhone 5, it seems to do it with a roughly equal power draw (though of course there's much more to battery life than the CPU).
So, basically, it's probably roughly equal to Krait while coming much later, but with a better GPU than the Krait S4 (though I'm not sure how it compares to the Krait S4 Pro).
doobydoo - Monday, September 17, 2012 - linkIsn't the point of these benchmarks that it's beating the performance of all the Android phones?
GPU wise yeah it will beat them all by a large margin too.