Page 1:From The Desk Of Andrew Ku And Chris Angelini
Page 2:The Motherboard Survey: Early Q4 2010
Page 3:The Multi-Core Trend
Page 4:CPU/GPU Hybrids And The Death Of The Graphics Business?
Page 5:The Future Of Nvidia
Page 6:GPGPU Programming, Where Is It?
Page 7:CPU/GPU Hybrids And Performance Integrated Graphics
CPU/GPU Hybrids And The Death Of The Graphics Business?
Question: Will there come a time when integrated graphics with programmable logic (like Sandy Bridge and Llano) make discrete graphics unnecessary?
- High-end gamers would still require a more powerful 3D experience, which is difficult for the [CPU/GPU] hybrid architecture.
- … so far, the performance of integrated graphics is not strong enough to replace discrete. As for enthusiasts, they still need discrete graphics for fluid gaming performance.
- Integrated graphics is still not powerful enough for gamers.
- You need discrete graphics for multi-display setups.
- Discrete solutions provide more flexibility to upgrade graphics for DIY users.
- The gaming PCs still need discrete graphic solutions.
- Yes, for most users. But power users will always want discrete solutions.
- There will be always be discrete graphic cards, but they'll be limited to the extreme gaming market.
Unlike the Graphics Card Survey, we are getting a decent mix of people foreseeing the declining impact of discrete graphics. Right off the bat, one third of our experts responded “yes, integrated graphics will de-emphasize discrete.” Two-thirds of those gave the caveat that the high-end market will exist in perpetuity. Meanwhile, 60% of those who responded "no" gave that same addendum. If you put that all together, it should be 77% responding yes, integrated graphics can replace discrete graphics except in the high-end space. If we normalize to that caveat, you can compare this to 16% of the Graphics Card Survey participants. This is a pretty big difference.
A lot of people are going to write this off because, in their minds, the motherboard business has more to gain from a potentially powerful hybrid. This isn’t really true. A video card user will always have a motherboard, but a motherboard user doesn’t necessarily need a video card, provided there is sufficient IGP performance. Remember that Sandy Bridge and Llano should come armed with integrated graphics across the entire processor lineup. If anything, this means that there is nothing to be gained in the number of motherboards sold. In this scenario, the company that makes both video cards and motherboards actually loses out on video card sales.
So what is the real reason behind the difference in opinions? There are a couple of factors at play we should consider:
- In order to design motherboards and get them manufactured on time, motherboard makers have access to technology at a much earlier stage for qualification purposes. For example, it looks like they had Sandy Bridge samples almost a full year before we were able to see a demo.
- We are asking about the future, not just Sandy Bridge and Llano. Motherboard makers are also in on the long-term roadmaps plans for Intel and AMD.
If we are talking about the long term (say five years), it is possible that hybrids could corner the graphics market on everything but the $150+ range. The people that are most likely to be in the know (outside the walls of Intel and AMD) certainly think that this is the eventual end game. Intel already has a strong hold on over half of the graphics market, which basically translates into a domination of the integrated graphics space. AMD’s design looks more likely to deliver something similar to what a discrete solution might offer (which makes sense, given the core's pedigree). So, the fight for the mainstream market is going to be very competitive.
We aren’t discounting Nvidia, but it is hard to have a discussion on Nvidia in the context of hybrids because there is so much uncertainty. What is the company's future? We felt that deserved a question of its own, so read on.