Print 86 comment(s) - last by StoveMeister.. on Mar 17 at 5:48 AM

Apple claims that the iPad 3 is "four times faster than Tegra 3"... does that claim hold up?

Apple, Inc. (AAPL), as usual, astounded the world with its latest product launch.  The company who made tablets a hot commodity a week ago pulled the wraps off its third generation tablet.  Featuring a high-definition "Retina" screen from LG Electronics Inc. (KS:066570) and an LTE modem, the new tablet kicked off a crazed frenzy of pre-orders.  

I. Same Results, Different CPU

Apple now has stated it will be out of stock of the popular tablet for several weeks, following the sales of its small-quantity of in-store stock. So what are buyers getting?  Well the first benchmarks have leaked out from Vietnamese forums site, and, if accurate, the picture is not as rosy as some fans had hoped.  

Using the Benchmark from GeekBench that measures integer, floating point performance, stream processing, and memory, the tester reveals that the computing power on the unspecified 1.0 GHz ARMv7 instruction set purported A5X dual-core CPU remains unchanged.

GeekBench iPad 3

[Source: Tienhte]

II. Better Graphics

The results would hint that when Apple said that its system-on-a-chip was "four times faster than Tegra 3", it was referring to the graphics processing unit, not the ARM central-processing unit. The GPU core is expected to be a quad-core variant of Kings Langley, UK-based Imagination Technologies Plc's (LON:IMG) PowerVR SGX543MP2, dubbed the SGX543MP4.

The SGX543MP4 is expected to pack a theoretical 134 MPolygon/s and 4 GPixel/s fill rate.  AnandTech's benchmarking of the iPad 2 vs. NVIDIA Corp.'s (NVDA) Tegra in GPU-centric GLBenchmark showed the last generation iPad to be anywhere from 30 to 80 percent faster than Tegra 3 in different benchmarks, versus Apple's claim that it was twice as fast.  Thus it could be expected that the new core in some cases would be 2x as fast as Tegra 3, or perhaps a bit better.

Given that Tegra 3 has a higher clock, perhaps it would be fair to say that to some extent in clock-per-clock Apple's GPU could be close to legitimately delivering 4x performance speed-up vs. Tegra 3, which is quite impressive (but again, the credit here goes first to Imagination Technologies for making the GPU, and second to Apple for securing the stock).

iPad 2 v. Tegra
[Source: Anandtech]

III. Rivals Prepare Counterstrike

NVIDIA is reportedly rushing its Tegra 4 counterpunch to market.  Named "Wayne", in honor of the DC Comics' fictional billionaire Bruce Wayne (aka Batman), NVIDIA's fourth gen. system-on-a-chip will be built on Taiwan Semiconductor Manufacturing Comp., Ltd.'s (TPE:2330new 28 nm process.  

Tegra 4 will pack a significantly increased core count and faster clock speeds, but NVIDIA hopes to keep power consumption and die size low, thanks to the die shrink and new circuit technologies.

A modified Tegra 4, code-named "Grey", after the iconic Jean Grey of X-Men fame, is packing an on-die 4G LTE radio from Icera Inc.  The shipping date has been accelerated to an earlier 2012 release, but is unclear whether designs sporting the chips will appear in time for the holidays.

"Grey" will be followed by a chip code-named after her at times flame "Logan" (aka "Wolverine").  Logan will pack a greatly improved GPU, and will launch in 2013.

If not, Apple's main competition will likely be the Qualcomm, Inc.'s (QCOMSnapdragon 4, which is expected to see deep pickup in the smartphone and tablet space, as well as in Windows 8 laptops.  Qualcomm is a close ally of Microsoft Corp. (MSFT) who recorded it by making it the exclusive CPU supplier of the Windows Phone platform at launch.

The Qualcomm CPU may prove a threat to Apple's tablet dominance as AnandTech’s described its performance as "insane".

IV. Samsung Truce Could Guarantee Steady Component Supply for iPad 3

AnandTech has published some additional info on the CPU, stating that it's Cortex-A9 MPCore, which indicates that Apple may have purchased an IP core from ARM Technologies Plc. (LON:ARM) and then modified it.  Our sources had previously hinted that the iPad 3 might carry Apple's first in-house designed CPU, but this claim has been difficult to verify due to Apple's extreme secrecy.

Regardless, both AnandTech’s analysis, the GeekBench metrics, and our own sources indicate that while the CPU may be an advance for Apple in terms of internalizing its design, it is hardly a step forward -- let alone a leap forward -- in terms of performance.

The GeekBench benchmark did confirm (as previously leaked by mobile engine developer Epic Games during its iPad 3-related press comments) that the new iPad 3 has 1 GB of RAM -- twice the memory of its predecessor.  That should help keep those hungry graphics GPU cores (which use the DRAM for slow-storage) fed.  

The core -- like the last generation model -- is expected to be printed by Samsung Electronics Comp., Ltd. (KS:005930) at its Texas facility, a mere miles away from one of Apple's top call centers.  Samsung -- reportedly the sole remaining profitable DRAM supplier -- is also expected to supply the design and process of the on-die DDR2 RAM (memory).  

Samsung Austin Texas
Samsung's CPU supplying plant is located in Austin, Tex. near Texas Instruments and Apple.
[Image Source: Let's Go Digital]

Samsung and Apple are reportedly negotiating and uneasy licensing truce, a testament to their deep mutual dependence from a supplier-client perspective, that runs counter to their heated market competition and legal rivalry [1][2][3][4][5][6][7][8].

V. The Beefy Battery

According to AnandTech’s math, the iPad 3 would last a mere 6 hours with a 25 watt-hour battery, based on the fact that Apple claims 10 hours of battery life on the new 42.5 watt-hour battery.  The new Apple tablet adds several power hungry components -- an LTE modem, a more intense, higher resolution screen, and the aforementioned higher-core GPU.

Apple's iPad 3 is a hungry monster necessitating a fatter battery. [Image Source: Apple]

No benchmarks on the LTE performance (surely network dependent) or battery life have been published yet, to our knowledge.  It remains to be seen whether the larger battery delivers better or worse results than Apple's promised 10 hours, in the real world.  

AnandTech founder Anand Shimpi, a veteran iPad user considered the larger battery a necessary evil, but complained, "The new iPad isn't as heavy as the original model, but it's clearly heavier than the iPad 2. I don't believe the added weight is a deal breaker, but it is a step backwards."

VI. LTE -- Fast and Dangerous

In addition to the $130 USD "LTE tax" you'll pay Apple to trade up from Wi-Fi, AnandTech reports that data plans will remain relatively pricey and capped at lower limits -- the highest of which is 5 GB.  The U.S. plans are available from Verizon Wireless, a joint venture between Verizon Communications Inc. (VZ) and Vodafone Group Plc. (LON:VOD), and AT&T, Inc. (T) -- America's top two wireless carriers in terms of subscribers.

iPad 3 LTE
[Source: Anandtech]

AT&T has the faster network, according to past tests.  However, Verizon Wireless's LTE network covers a much wider geographic range, covering an estimated 200+ million Americans to date.  That range should balloon if Verizon latest big spectrum purchase is given the greenlight by the U.S. Federal Communication Commission.  However, customers on both networks should be wary of hitting their caps and being smacked with fees -- those 10 second app downloads might be a dangerous privilege for some.

On the plus side AnandTech reports that the new tablet is not SIM-locked.  Apple's decision to open up on the SIM side means an easier path to international roaming, as you can just pop in different cards to hop on different subscribed-to networks.

Sources: Tinhte [Vietnamese], AnandTech, AP

Comments     Threshold

This article is over a month old, voting and posting comments is disabled

RE: Maybe I'm wrong but...
By extra_baggage on 3/13/2012 4:29:22 PM , Rating: 1
You are forgetting the increased ram!

RE: Maybe I'm wrong but...
By B3an on 3/13/2012 4:39:34 PM , Rating: 3
More RAM will not magically make it perform better, it will just help compensate for the extra pixels it has to push.

At native resolution games should actually run slower than the iPad 2.

RE: Maybe I'm wrong but...
By tayb on 3/13/2012 6:40:27 PM , Rating: 2
Except that games don't scale that way. As an example from a recent Anand article (pasted below) scaling from 1680x1050 to 2560x1600 resulted in a 61.75% drop in FPS but the increase in pixels was 132.20%. The "new iPad" is pushing 4 times as many pixels as the iPad 2 but that does not mean you should expect 1/4 of the performance. They have also doubled the theoretical performance of the GPU, may have increased the CPU speed, could have optimized the software, and might have doubled the amount of RAM. This is a "we won't know until we know" kind of thing.

RE: Maybe I'm wrong but...
By snorldown on 3/13/2012 7:33:44 PM , Rating: 4
Your example just reinforces B3an's point. Assuming performance scales linearly with the number of pixels, a 132% increase in the pixel count would result in a 57% drop in FPS, so the fact that the benchmark showed a 61.75% drop means that the scaling was even worse.

RE: Maybe I'm wrong but...
By testerguy on 3/14/2012 2:49:31 AM , Rating: 3
Sorry but tayb is correct, he just used a bad example.

Take the same link of his.

Look at the HD 7950 (only selected because it's the first one on the list). At 1920 x 1200, it achieves an FPS of 47.1

Now, lets look at the result for 2560 x 1600. This is a move from 2304000 pixels to 4096000 pixels. This represents 1.78x as many pixels. As a result, if performance scales linearly, we would expect an FPS of 47.1/1.78, which is 26.46.

Instead, we see an FPS of 31.1.

Proof, by counter example, that FPS does not necessarily scale linearly with number of pixels. This also proves that 4x the pixels does not necessarily mean 1/4 of the FPS. Which was tayb's original point.

RE: Maybe I'm wrong but...
By testerguy on 3/14/2012 2:52:23 AM , Rating: 2
Edit: Sorry the 47.1 figure I quoted is actually 47.7 on the graph.

The point still stands though, 47.7/1.78 is 26.8 FPS (rather than the 26.46). Still less than the 31.1 (14% less) and still proves the point.

RE: Maybe I'm wrong but...
By theapparition on 3/14/2012 9:28:18 AM , Rating: 2
Despite your horrendous position on benchmarking in the previous article, I do agree with you that benchmarks don't scale linearly with resolution.

There's a whole system that is in place that needs to be considered, including the amount of graphics ram, it's speed, and it's data path to the CPU.

RE: Maybe I'm wrong but...
By therealnickdanger on 3/14/2012 11:28:16 AM , Rating: 3
I'm sorry, but did Apple claim "4X GPU performance" at the native resolution? I don't remember that being the direct claim. Seems like fanboys are getting their collective panties in a bunch over very little. They seem to forget that rendered resolution and displayed resolution are not always equal. Most Xbox/PS3 games are rendered at 720p (or lower in some cases) but that doesn't prevent the console or the display from scaling it up to 1080p.

Will the iPad 3 GPU be 4X faster than a 800p Tegra 3 device at native resolution? Likely not. Will it be 4X faster at a common resolution like 1024x768? It's possible.

By therealnickdanger on 3/15/2012 1:43:41 PM , Rating: 2
And the results are in. Compare:

Tegra 3
GL Egypt Offscreen 720p: 64fps
GL Pro Offscreen 720p: 78fps

iPad 3
GL Egypt Offscreen 720p: 140fps (2.2X faster)
GL Pro Offscreen 720p: 241fps (3.1X faster)

So it's not literally 4X faster, at least in this benchmark, but it is the very reason words like "pwnd" exist.

That being said, I'm very happy with my Acer Iconia A100.

RE: Maybe I'm wrong but...
By retrospooty on 3/14/2012 11:53:30 AM , Rating: 2
"Despite your horrendous position on benchmarking in the previous article, I do agree with you that benchmarks don't scale linearly with resolution."

That is because the fanboy's perspective will change when it benefits Apple's viewpoint. On this particular subject, the concept that performance doesn't scale linear with res. benefits Apple, so the fanboy takes that tact. It just happens to be correct in this particular subject.

RE: Maybe I'm wrong but...
By testerguy on 3/14/2012 2:45:18 PM , Rating: 1
That is because the fanboy's perspective will change when it benefits Apple's viewpoint. On this particular subject, the concept that performance doesn't scale linear with res. benefits Apple, so the fanboy takes that tact. It just happens to be correct in this particular subject.

Clueless hate comment fuelled by bitterness. If you actually read my former stances on benchmarks, it's ALWAYS been that they have to be compared on equal resolutions. The exact same point I've made here. It's common knowledge, and a well followed principle that when benchmarking GPU's, you use the same resolution. That is what I was stating in the previous article. If you disagree with that obvious reality, take it up with Anandtech or any other true tech site who always benchmark that way.

RE: Maybe I'm wrong but...
By retrospooty on 3/15/2012 10:14:18 AM , Rating: 2
"Clueless hate comment fuelled by bitterness."

??? huh

Again, reading comprehension is a problem for you. I do agree, and said you are correct above with regards to benchmarks.

What I am also saying is that you always side with Apple, that is your sole purpose here and that is all you comment on. You are here on an agenda. With regards to benchmarks, lets say the tables were turned and the current benchmarks were not favoring Apple. your stance WOULD change or you would simply not comment at all.

When all you do is post in Apple articles, attack anyone that has anything negative to say about Apple, and come down on Apple's side 100% of the time no matter what the issue is, people dont take you seriously. Flake off now fanboy.

RE: Maybe I'm wrong but...
By B3an on 3/13/2012 8:54:25 PM , Rating: 2
CPU speed is the same @ 1GHz. And as i said, more RAM will not make games run better! Unless games are limited by not enough RAM, which no iPAd games are being as all the developers know exactly how much RAM they have to work with.

Another example would be that i have 32GB RAM in my PC, but taking much of it out and just leaving 6GB does not make games run any worse (apart from slightly longer loading screens), being as no PC game uses more than 4GB. Adding more RAM when it's not going to be used will not increase FPS.

If anything, games that decide to use the extra RAM in the iPad 3 will see even more of a performance hit from the higher res textures used and so on. Or all that extra RAM could easily be used up from just enabling AA, which needs a lot of RAM for that res, which again will see a performance hit. The iPad 3 will not run games better at native res with just these GPU changes.

RE: Maybe I'm wrong but...
By reginhild2 on 3/14/2012 3:01:40 PM , Rating: 3
Good example. The geekbench number above is interesting as the Tegra 3 in the Transformer Prime kicks it to the curb.

The first independent benchmark comparison test is out and the Tegra 3 CPU is over 2x faster than the CPU in the iPad 3:

RE: Maybe I'm wrong but...
By quiksilvr on 3/13/2012 9:05:56 PM , Rating: 1
One thing worth mentioning:

Those Tegra 3 benchmarks were taken BEFORE ICS. Those are Honeycomb benchmarks.

RE: Maybe I'm wrong but...
By MrMilli on 3/14/2012 8:53:09 AM , Rating: 3
Those numbers only relate to how the HD 7950 scales but that doesn't mean this kind of a scaling just transfers to the A5X. As a matter of a fact I can assure you that it doesn't.
That a videocard with 240GB/s dedicated memory bandwidth and 25.6 GPixel/s fill rate can take a punch, we all know. But the A5X with it's around 5-6GB/s shared memory bandwidth and 4GPixel/s fill rate (PowerVR: All fill rate figures stated assuming a scene depth complexity of x2.5. That means the actual fill rate is 1.6GPixel/s) will have a much harder time with this 4x increase in pixel count.
My guess is that the performance drop will be even worse than linear because of the many bottlenecks.
So as stated, the performance of games on the new iPad will be worse than the iPad 2 on their respective native resolutions.
To fight this effect, developers won't run 3D games at 2048x1536 but will keep the engine running at 1024x768 internally. They will apply the (almost free) 2xAA and then scale the output to 2048x1536 (just the like a XBOX360 would run FullHD).

"The Space Elevator will be built about 50 years after everyone stops laughing" -- Sir Arthur C. Clarke

Latest Headlines
Inspiron Laptops & 2-in-1 PCs
September 25, 2016, 9:00 AM
The Samsung Galaxy S7
September 14, 2016, 6:00 AM
Apple Watch 2 – Coming September 7th
September 3, 2016, 6:30 AM
Apple says “See you on the 7th.”
September 1, 2016, 6:30 AM

Most Popular Articles5 Cases for iPhone 7 and 7 iPhone Plus
September 18, 2016, 10:08 AM
No More Turtlenecks - Try Snakables
September 19, 2016, 7:44 AM
ADHD Diagnosis and Treatment in Children: Problem or Paranoia?
September 19, 2016, 5:30 AM
Walmart may get "Robot Shopping Carts?"
September 17, 2016, 6:01 AM
Automaker Porsche may expand range of Panamera Coupe design.
September 18, 2016, 11:00 AM

Copyright 2016 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki