Why are there many gpu manufacturers (zotac, Asus, msi…) but onlz 2 pc CPU manufacturers?

739 views

Why are there many gpu manufacturers (zotac, Asus, msi…) but onlz 2 pc CPU manufacturers?

In: Technology

It comes down to the stunning cost of being a major CPU manufacturer. The industry is so competitive, you need to build a major new factory every two years or so, and such a factory can easily cost over a billion dollars, and must be supported by an army of extremely advanced design engineers.

There isn’t room for many companies to keep up with this spending race and still make a profit.

They’re graphics card manufacturers, not GPU manufacturers. For the GPU the two largest manufacturers are nVidia and AMD, it’s pretty much like with CPU’s.

They develop and create the GPU. Third party graphics card manufacturers can then make a deal with Nvidia to use their GPU. The GPU design can’t be changed, but everything else that’s on a graphics card, such as ram or voltage controllers. Besides that, they can increase the MHz of the GPU.

If you buy a CPU you already buy the chip itself. There’s pretty much nothing, besides overclocking, that a 3rd party manufacturer could change. And Intel and AMD rather sell their own products.

I must say though that this only applies for the Desktop/Laptop market. When it comes to other things, such as smartphones, the list of actual GPU and CPU manufacturers significantly increases.

Motherboard makers are like restaurants. They take available ingredients (computer chips, resistors, capacitors, etc.) and cook (wire them in the board and program software drivers) them into a dish (motherboard, GPU board). Since the ingredients can be easily sourced, making a new restaurant (brand) is not difficult.

CPUs are difficult-to-make ingredients that are required by certain dishes. Like high quality olive oil that requires huge fields of old olive trees and years of experience in processing the olives, CPUs require billions of dollars to build the fabrication factories and very specialized engineers to design them. This makes it difficult for new companies to compete with the two entrenched CPU companies. This also applies to GPUs.

There used to many more CPU manufacturers in the old days.

Pre-pentium when you had CPUs with names like 486 every company that could make CPUs could make their own version of the chip and they would all fit the same slot on the motherboard have the same instructions and work the same way with the rest of the system.

Then Intel and to a lesser degree AMD made their chips a brand with slogans like “Intel Inside” and dancing people in multicolored cleanroom suits. They also changed their products to better set them apart from the competition and to make them incompatible with one another.

One by one the other chipmakers shut down their own x86 lines so that today only Intel and AMD (and a Chinese attempt to make their own x86 cpu) remain.

A big issue was that for the CPU you would need a different board for each maker and the OS would need to support them all.

There simply isn’t enough room in the market for many different ways of doing things.

Even Intel and AMD are now on sort of the own standard again. When it came to going from the old 32-bit CPUs to new 64-bit systems. Both AMD and Intel created their own way architecture. For a while there was a Windows 64-bit OS for the Intel 64 and one for the AMD 64 bit CPUs. The Intel one is history now and both AMD and Intel create their CPUs according to the AMD created standard. Nobody wants to deal with two different incompatible versions of the same OS.

For GPUs things are a lot easier. The OS and the mainboard don’t really care which manufacturer puts their cards into the computer as long as they adhere to the relevant standards and provide working drivers nobody on the other side of the abstraction layer needs to worry too much about it.

GPU’s are much more complex than CPU’s. They’re essentially computers all on their own, minus an OS. nVidia and AMD mostly design just the chips and the basic PCB necessary to run them. They then license the designs to manufacturers such as ASUS and MSI, who put forth the resources to design custom boards, coolers and BIOS. This allows nVidia and AMD to run more efficiently, as they do not have to worry about production, only design.

CPU’s in contrast are a much simpler design, as it’s just one chip. Unlike GPU’s, CPU’s do not have voltage control, memory, fan controllers, etc. AMD outsources production to TSMC, but Intel fabricates their own CPU’s.

There are really only three GPU manufacturers, and only two of them have any gaming or performance credentials: nVidia and AMD, with Intel being bulk but no performance.

Zotac, EVGA, Sapphire, XFX, Zotac, Power Color – these are all GPU “partners.” Those companies make something more accurately called a “video card.” But, much like how “CPU” used to also refer to an entire *tower*, terminology can sometimes be fuzzy.

The GPU partners that make video cards do so with only either AMD or nVidia chips. Making one of those cards is much more like making a whole computer than it is a discrete card. A modern video card has significant amounts of its own memory, it has its own cooling solution, it has specific power requirement, and it has the GPU chip installed.

Really, you can think of the “board partner” company as being more like Dell, Gateway, HP, Asus, Alienware, Falcon Northwest, or the old IBM – they make variations of a device with components from a smaller list (two, in both cases) of manufacturers. Intel and AMD for CPUs, and nVidia and AMD for GPUs.

Many companies make graphics cards, but only two companies, Nvidia and AMD, make GPU chips. The brand-name third-party companies like Asus and EVGA buy key components from the chipmaker and manufacture them into graphics cards.

Nvidia and AMD engineers don’t just make new chips. They also make a reference design, which is a schematic and circuit board that implements their new GPU chip into a graphics card. When manufacturers like Asus commit to buying Nvidia’s chips, Nvidia licenses the reference design to them, so they can base their GPU on it.

So, when Nvidia launches a new chip, like the GTX 1080 TI, they’re the first ones to bring it to market. You can buy their in-house basic “vanilla” version of a GTX 1080 TI graphics card.

The first batch of third-party cards are almost straight implementations of the reference design. Asus literally takes the schematic from Nvidia and makes minor tweaks so they can use their favorite manufacturing plants and parts suppliers. This first round of graphics cards are nearly identical – Asus, EVGA, Zotac, MSI, etc are all using the same schematic!

As time goes on, third parties improve the reference design to set themselves apart. For instance, Asus replaces the power-management circuit with a stronger version so you can overclock the card easier. Zotac replaces the fan and heatsink with a flashier custom version. EVGA replaces a DisplayPort output with an HDMI. MSI writes improved firmware that gives users better control over the card.

As time goes on, companies improve their proprietary designs and GPU’s start getting really exotic. For instance, EVGA develops a GTX 1030 that doesn’t need a cooling fan and fits in a tiny micro-ATX case.

Processors are a little different – instead of buying a “processor card” with a CPU soldered onto it, you buy the processor and the carrier board – the motherboard – separately. Even though only one company manufactures the Intel i7700, dozens of companies implement its Z370 chipset into compatible motherboards.

As the density of the transistors increased and the processes got smaller (measured in nano-meters or nm) it became more and more expensive to design and make CPUs so companies started getting out of it or failing entirely.

Also, as server processors needed different optimization, and mobile CPUs became lucrative, some companies specialized in one or the other and left PCs to Intel and AMD

A graphics card consists of multiple different parts, the important bits being the actual GPU (the chip that does all the calculations) and then all the other stuff to power, cool, and connect a GPU.

There are really only two big names in making actual GPUs for graphics cards: AMD and Nvidia. They make the GPU chips and then graphics card manufacturers will take them and make custom graphics cards for them using their own different cooling systems and other features, much like how companies like Dell, HP, Lenovo, Apple, etc. will take the CPUs that AMD and Intel make and design computers that use them.

As for why only these two do, it’s much, much harder to develop the circuitry involved in any sort of computing unit, be it a CPU, GPU, or any other sort of ‘-PU’. It takes a lot of money and time in research and development to produce these chips, but it’s easier to then take the chips and design cards that use them.

You’re statement is not entirely correct. There are many manufacturers for CPUs. However, the most dominant ones for general purpose computing are Intel and AMD. The problem with trying to enter the general purpose CPU market is competition from intel and AMD. It takes a lot of money and marketing.

Likewise, for GPUs, there are only a few dominant general purpose GPU makers: AMD, nvidia, and intel. There are others as well, but they are not well known.

Asus, MSI, zotac, etc, are board manufacturers who put GPUs onto their boards and sell them. Likewise, they also sell motherboards to put CPUs on.