📜 ⬆️ ⬇️

Clarification of the great quad-core confusion with Apple A5X

A5X
When Apple introduced the new iPad (the third one), the Apple A5X chip piece created a lot of confusion and excitement, put dust in the eyes of someone, and angered someone. Now, many believe that the new iPad is a quad-core tablet, and also the A5X chip is four times faster, according to Apple, than NVIDIA Tegra 3 , which is currently the fastest chip for Android devices and tablets with Windows 8. Tegra 3 is also the first chip for mobile devices with four cores. Apple used the term quad-core in an unusual sense (graphics core), which led to confusion in the understanding of the word in many people. That was exactly what Apple needed from a marketing point of view. In short, the Apple A5X is NOT a quad-core processor, but let's figure it out ...

Definitely not a “quad-core processor”


Usually, the term quad-core is used to denote the number of central processing units (CPUs) or cores in a chip. These processors perform common tasks, such as parsing web pages, installing and running applications, as well as most other things that computers do. To help with this CPU, manufacturers usually add a graphics processor (GPU) that takes on graphical tasks such as user interface, games, etc. Or, for example, many install individual video cards that encode and decode video streams. When they say "dual-core" by this means a dual-core CPU. In the presentation of Apple, they talked about graphics cores - this is completely different.

Mobile processors are highly integrated and are called Systems On a Chip or SoCs. For more information about them, you can find out by reading the All Smartphones and Tablets .

So what is the kernel GPU?


On desktop computers, GPUs can hold 512 "cores" and more.
On desktop computers, GPUs can hold 512 "cores" and more.
')
Graphic processors usually use a set of basic computing "building blocks", which are also called "cores". Due to the fact that computer graphics can be highly parallelized, the addition of cores is one way to increase graphics performance. GPU manufacturers, such as AMD, Power VG, Qualcomm, NVIDIA and others, have been doing this for more than ten years, and modern desktop GPUs can accommodate more than 512 "cores".

The fact is that GPU cores are so common that the term has lost its meaning. It makes sense to compare several GPU models of the same generation, and not the same brand with the other, because the definition of the graphics core can vary greatly. For example, AMD usually has much more cores than NVIDIA, but as a result, actual performance does not reflect this difference.

In short, the last two paragraphs indicate that the GPU cores have turned into a marketing ploy, but in reality only real performance matters.

In the end ... is the Apple A5X really four times faster than the NVIDIA Tegra 3 in graphics?


This is where the next controversy begins. It is best to test performance on both platforms with the same applications. Depending on your goal (artificial or real indicators), you choose the criteria by which the load is checked in a certain way. Currently, there are no criteria by which the A5X can be compared with the Tegra 3, but there are some by which the A5 and the Tegra 3 can be compared.

Screenshot from GLBenchmark Egypt
Screenshot from GLBenchmark Egypt

Based on the results of an independent comparison of the A5 and Tegra3 processors, I assume that Apple claims a fourfold speed based on the GLBenchmark “Egypt” indicators with the PRO Offscreen settings. The A5 processor (iPad 2) has twice the performance of Tegra 3, with fps 150 (A5) versus 82 fps (Tegra 3). Fps is frames per second - the number of frames per second. This is an interesting criterion, but, frankly, old and does not meet the capabilities of modern games. He simply overestimates the advantages that the A5 chip has (a drawing with a simple texture and very simple lighting).

Since the GLbenchmark “Egypt” doesn’t even closely reflect modern games (compare it with Infinity Blade , the “old” game), the numbers are not very informative. In the end, we really need built-in gaming cross-platform test criteria, but they are not. Why? This is another story ... my experience in the developer community suggests that policy gets in the way of creating in-game criteria. Why? Because most companies, releasing their products, do not want to depend on the performance of the equipment.

It can be concluded that the A5X is four times faster than the Tegra 3, and another way. Given the total computing power (multiply-adds or MADs) of each GPU (in GFLOPS ), in theory you can come up with figures that 100% use of all transistors allows. Depending on the clock-rates and provided that Apple uses a CPU with the PowerVR SGX543MP4 design, there is a real chance to reach the magic number "4".

Answering the question in the paragraph heading, I will say that the A5X seems to use a very fast graphics engine and under certain conditions, it is undoubtedly faster than Tegra 3. Speaking it, I do not think that in any real game you you will see something close to four times the difference in speed (will Riptide run at a speed of 240 fps at a resolution of 1280? 720 on the A5X?).
My opinion is that Apple’s priority was to handle more pixels to provide better fillrate, while NVIDIA’s priority was to focus on more complex rendering (shading).

In reality, it all depends on how applications use current equipment (again, at the head of all software). Someone may decide to play on the strengths of the A5 / A5X and to focus on their advantages in game development. Others may decide to move the console games, which will work fine on Tegra 3. As a result, both platforms will be in the black.

So anyway ... is the new iPad a quad-core tablet? (Not)


Simply no". Because the "quad" implies the presence of four CPU. The A5X is a dual-core chip, therefore, the iPad is a dual-core tablet.

This is not to say that it is bad. Several people complained about the performance of the chip in the iPad 2, but in terms of CPU usage, a real quad-core chip, such as NVIDIA Tegra 3, may be twice as fast as the A5X, based on the fact that they both use the ARM Cortex A9 chip CPU (also need to take into account the speed of MHz). If multi-core applications are friendly, then a chip with a large number of cores will work better. It's simple. If the applications on the contrary are not multi-core friendly, then the number of cores is irrelevant.

Why then all the noise?


For several reasons, quad-core chips have been a hot topic lately. First, it is a simple thing to “capture” potential buyers. It's like megapixels for cameras. "Four" certainly sounds better than "two", and Intel and AMD have advertised quad-core processors for many years, so users are aware of all this.

Secondly, Android-based tablets began using quad-core chips as early as November with ASUS Transformer Prime, so in a way, new tablets that have “only” two cores lose by quad-core because they look old and not so powerful. The manufacturer is very anxious when releasing a dual-core tablet. And in the end, the “specs-sensitive” crowd will not approve of the “old” dual-core technologies, and Apple will have to contend with the quad-core argument for the rest of 2012 (at least with respect to the new iPad).

This explains why Apple released the “something (nothing?) Quad-core.” Of course, to say that the graphic element A5X is “quad-core” absolutely correct from a semantic point of view, but this leads to a misunderstanding. Was this a deliberate attempt to expand the meaning of the term “quad”? I dont know. It is important that a misunderstanding arose, therefore, at best, the thought was not clear.

Conclusion: it is always important to test everything.


Misunderstanding will never lead to anything good when you are going to spend a significant amount of money on any device. At Ubergizmo, we believe that informed buyers become happy users, so it was important to talk about it. The new iPad is a dual-core tablet PC with a graphics engine twice as fast as the iPad 2, we know that for sure.

When we conduct independent testing, we expect to see the results reflected in the theory formulated above, and we look forward to the opportunity to test actual applications ourselves. Wait for the full review of Ubergizmo of the new iPad on March 19, as our device will come on the 16th. We are waiting for your comments, because we are interested in your opinion. If you want to save money by buying an old model, read our review of iPad 1 and iPad 2 .

Translator's Note

I left some terms without translation, as I thought that it would be wiser to use them without translation and the translated terms would be less clear.

Source: https://habr.com/ru/post/139868/


All Articles