Brother Jie found a very serious problem.
That is, whenever we discuss the performance of mobile phones.
In the comment area, there is no lack of organic friends who feel -
The current mobile phone performance is already excessive
Although, Brother Ji does not want to admit the remarks of "excess performance", after all, nowadays, there is no machine that can truly conquer the original god...
But in fact, it is.
For the vast majority, especially those who don't play games much.
Now the performance of the phone, indeed... is excessive.
But time will not go backwards, and technology will not stand still.
Whether you hate it or reject it.
Can't stop those, the crazy ♂ chip manufacturers, changing the law, hooking and carving on this small silicon-based chip~
Yes, you read that right.
Brother MediaTek, actually and NVIDIA Lao Huang... Cooperated.
Netizens who saw this news called out a mess.
Ah, this, this...
How do the mobile phone chip manufacturers at the good end get mixed with the graphics card manufacturers.
Even if it is the so-called "cross-border X joint branding".
But don't take it like that...
Well, maybe you don't know something.
The relationship between this mobile phone chip and the computer chip.
It's really going to be broken and unfolded.
It is not as "cross-border" and "broken circle" as everyone thinks.
It's nothing new.
If you say that Ago, relying on its own superior "banknote" ability conditions, steady and steady, and finally with the mobile architecture ARM body, comparable to the X86 desktop graphics card.
That next door to Qualcomm's success.
More or less, it is inseparable from years of in-depth ♂ cooperation with the graphics card manufacturer AMD.
Xingzi also bought AMD's GPU architecture long ago and used it on the self-developed Exynos chip.
Although the effect is not very dripping...
So just this time, Brother Fa and Lao Huang "suddenly" cooperated.
According to Brother Ji, it is both an accident and an inevitability of history.
Even if it's 10,000 steps back.
Neverly, no.
Brother Fa go to the toothpaste factory to cooperate...
Now.
Surely the organic friend will ask:
"Brother, I understand the truth, but why did mobile phone chip manufacturers give up the GPU architecture that was used well before, and turn to cooperate with the graphics card manufacturers on the PC side?"
Oh yes.
Because, mobile phone chips, has reached the bottleneck period at the object level.
True.
There is a fundamental difference between a mobile platform and a computer platform.
There is still a big gap between mobile chips and computer chips.
But when we really pull out the chip of the graphics card and the phone, we will find out.
The difference between the two is not as big as we think.
For example, scale --
Regular graphics cards, although still dominant.
But if it is an "entry-level" graphics card such as RTX 3050, the number of transistors in the entire GPU core chip is only 12 billion.
And what about the A16 next door?
There are already 16 billion.
In addition to the transistor size, the chip area is even greater.
Like the just-baked RTX 4060, the GPU chip area is 146mm².
And what about the A16.
It's already 132mm².
So, before you know it.
The scale of our mobile phone's SoC has secretly caught up with the desktop.
Especially in this one.
"Small but beautiful" and "Moore's Law" collective failure environment.
In order to achieve breakthroughs in performance and data.
The "vigorous effort" that was once scorned by everyone has now become a strategic policy of a number of manufacturers, all of whom agree and coincide.
Naturally, this has bred M1 Ultra, a "monster" that is three times larger than conventional chips.
Though.
Take the integrated CPU, NPU, memory... SoC chips with many cores, compared to graphics chips with only GPUs, are indeed a bit unkind.
But there is a little bit.
Presumably no one will deny it.
Throughout the history of the development of the machine industry, there may never be such a mobile chip process, scale, volume... In all directions, it is infinitely close to the moment of the desktop chip.
So, in a Moment.
If you still take the design idea of "small chips" to make "big chips"...
Apparently, out of place.
The safest and most feasible way is to let manufacturers who are accustomed to doing "big chips" intervene and use their handy desktop design ideas to design our already small mobile phone chips.
Especially when ray tracing, AI complementing, game engines, and so on, all kinds of bells and whistles are added.
GPUs led by ARM and PowerVR are traditional reduced instruction set architectures.
Performance, can't feed these guys anymore.
Let's let the professionals do professional things.
But that's the truth.
But the problem is --
Can they really, really make mobile phone chips?
Let's not talk about the NVIDIA family, there is no physical Orin Ampere structure.
Let's take a look at the more familiar Snapdragon and AMD~
Speaking up.
Do you know why the relationship between Qualcomm and AMD is so good that every time there is a little thing and a break, they will pull each other together?
Brother Suo yelled at you.
This, all for a reason drop~
Snapdragon's Adreno, presumably friends will not be unfamiliar.
So, what about AMD's Radeon?
In 2006, AMD, which was originally only a processor, suddenly smashed the pot and sold iron to save 5.6 billion US dollars (about 39.3 billion yuan) and acquired a graphics card company called ATI.
In order to enter the graphics card market.
Prepare to fight the ring match with Nvidia Lao Huang next door.
And this company, called ATI, is not only developing desktop graphics card GPUs.
It also launched "Imageon", a GPU chip for mobile phones.
Brother Ji mentioned in passing.
Qualcomm also cooperated with "Imageon", that is, ATI.
After AMD officially acquired ATI.
"Imageon", as a sub-project, is naturally also included in AMD.
But AMD had no interest in mobile at the time.
I only think about how to line up with NVIDIA and tear up.
Seeing Qualcomm sitting on a cold bench with his old friend "Imageon" "couldn't bear it."
In 2009, the entire department of "Imageon" was packaged and taken away for a high price of 65 million US dollars (about 450 million yuan).
We all know what happened later.
AMD, the entire ATI team, has invested in the research and development of the "Radeon" graphics card.
The "Imageon" team, which split off and went to Qualcomm's home, lost its real name, so it was renamed "Adreno".
Eh and so on.
Let's take a closer look at the logos of Radeon and Adreno——
Found the difference?
Therefore, there is no such thing as "mobile is becoming more and more like desktop".
There is no need to worry at all, graphics card manufacturers will not make ARM chips.
Originally, weren't they the same thing (bushi
Image from the web
It turned out to be such a relationship!