The current state of the GPU market
Currently, there are 13 companies publicly designing and manufacturing GPUs. For those that sell their chips to Add-In-Board-Partners (AIBs) like AMD and Nvidia, they use up to 29 manufacturers. Nvidia and AMD are likely the two big companies you’ve probably heard of, as they are the two popular options when it comes to ‘discreet’ – add in – graphics cards.
So you can understand the current state – lets take a quick look at how GPUs came to be. GPU type chips were born in the 1970’s to help accelerate arcade games like space invaders – so all you folks saying gaming does nothing for the world… Where would your medical research be without gaming!
Anyway… These chips relied on “Fixed function units”. These are hardware circuits designed to do one specific task. This is relatively inefficient as you are not always doing every function at the same time – you might be processing geometry but not colouring the pixels so the bit of the chip that colours the pixels would be sat idle.
ATI was actually founded in 1985, a full 8 years before Nvidia in 1993, although they were acquired by AMD in 2006 for a whopping $5.4bn. ATI made some rather interesting cards, including what they called the “All-in-wonder” which not only accelerated 3D applications but also has a build in TV tuner too! ATI made the graphics chips for the Game Cube, the Wii and Xbox 360. After the AMD acquisition, AMD also provided the CPU and GPU to the PS4 lineup, and the Xbox one. Oh and the Wii U too.
Nvidia started in 1993 with their NV1 design. In 1999, they released their NV10 architecture which was remarkable due to the hardware Translation and lighting – something previously done using the CPU but thanks to DirectX supporting the feature on DX 7, GPUs were able to massively speed up this process. Nvidia has been a market leader for many years now, although up until 2014 they generally stuck at around 60% to ATI/AMD’s 40%.
As of 2014, the share difference grew massively, with AMD reportedly only just regaining some share to now be at 30% to Nvidia’s 70%. The worst area for AMD was Q2 of 2015 where Nvidia has 82% – almost as high as Intel’s 87.7% share in the CPU market.
Speaking of Intel, they currently include GPU segments in their CPUs – known as iGPUs or Integrated GPUs – which do alright if you are after low power web browsing office work type applications. Even their IRIS PRO lineup of iGPUs aren’t all that powerful, and while it’s a nice inclusion I really wouldn’t recommend trying to game on it (unless necessary).
Chiropractic care has the ability to relieve stress. levitra australia prices There are some certain foods that help an individual purchase levitra improving blood flow in the body. The preparations and treatment suggested by sans prescription viagra these medications prepared with natural ingredients helps with erection problems without causing any side effect to your health then do not panic. Choosing online Kamagra seller, you can you near ones unaware of what you are going to have. viagra tablets 20mg Interestingly, in 2011 Intel agreed to pay Nvidia $1.5bn for access to its graphics technologies. Since that deal came to a close in 2016, rumours have circulated that Intel will now license AMD’s tech to help further its own products. At this point, it’s nothing more than speculation so until you see the contract documents – it hasn’t happened yet.
I’ve focused a lot on desktop GPUs – it’s my area of expertise and all – but to really give you an idea of the full state of the GPU market I have to talk at least a little about the mobile side.
Qualcomm is one of the most popular CPU and GPU manufacturers when it comes to ARM based mobile products. Their Adreno GPU’s are in most Android phones – even in some Samsung phones when they are unable to use their own Exynos chips. The Adreno chips were actually designed by ATI, but were bought by Qualcomm in 2009 and renamed to ‘Adreno’, where they were put into the Snapdragon mobile chips we have in most of our phones today.
The best Adreno GPU currently available is the 530 – which can be ran at 650MHz, is made with a 14nm process and has around 520GFLOPs of compute power – to give you context, an AMD RX 480 has around 5.8 Teraflops meaning it’s a little over 10 times more compute power – in theory anyway.
On the desktop side, we are coming to an interesting point. Nvidia has one card left to show in this generation – the 1080ti. The ti version of their higher of 80 cards is always interesting, since it brings the price of the existing cards down, and often sits relatively close to the same generation Titan card which is normally ludicrously expensive.
AMD however has a very interesting card to play, which is their entire VEGA architecture. They’ve announced that it will be available basically in Q2 of 2017, after their Ryzen CPUs launch, and have released a little information on how VEGA works. The biggest thing I’ve seen is the use of High bandwidth memory – HBM2 specifically – as a ‘High Bandwidth Cache’ and using system RAM, Non-volatile RAM and even your hard drive (through ‘virtual address space’) as the more standard ‘VRAM’ like we see on all current GPUs. This could be a pretty big leap forward when it comes to GPU technologies, although it’s unlikely to give major performance improvements right out the gate as it’s a technology designed for the future more than the present.
So – there ya have it. The current state of the GPU market. We are in a relatively fragile place at the moment – there is a lot of things that could go wrong and mean that we see a pretty bad monopoly that would regularly mess with the customers – especially on the desktop front. My advice for this is make sure you always buy the best card for you – not just the one an advert tells you to!
On the mobile front, I’m not overly knowledgeable I’m afraid, so if you know any more on this front please do let me know in the comments below as I’d love to understand the topic more – but we seem to be on a good path with it. Imagine Technologies’ PowerVR chips currently power Apple’s devices, and the Adreno/Exynos chips are getting better and better each year – bringing the idea of just having a phone you can dock as your office PC, HTPC, phone and other general applications just that little bit closer.
That’s about it. If you enjoyed the video please consider subscribing and sharing it on reddit, tech forums or just with someone you think will find it useful. If you learned something new do let me know down in the comments, and of course if you have any more information on this or if I got anything wrong, let me know in the comments too! Let me know what you thought of the video, and of course if you want to support myself, the channel and these videos more, when ever you are shopping on Amazon or OverclockersUK use my affiliate links too. Thanks for watching and I’ll see you on Friday the next video!
