Sign in with
Sign up | Sign in

CPU/GPU Hybrids And The Death Of The Graphics Business?

Talking Heads: VGA Manager Edition, September 2010

Question: Will there ever come a time when integrated graphics make discrete graphics cards unnecessary?

  • There are still a lot of room on the performance improvement front for discrete graphics and 3D applications. As long as the applications do not satisfy our need for realism, discrete will still be necessary.
  • Discrete graphics cards always have their place, especially for the high-end segment, which the power envelope will not fit into CPU/GPU hybrid solution.
  • At the current stage, we don't think the CPU/GPU hybrids can replace high-end discrete graphics. The design architecture of CPU-based CPU/GPU hybrids cannot really compete with what discrete graphics can do.
  • The levels of integration necessary to deliver even today's high-end graphics performance, together with full CPU logic, are unrealistic for the foreseeable future. In any case, this is a moving target, as graphics performance continues to multiply year on year. There will always be specialist applications in design or animation, for example, where the highest levels of graphics performance are still not enough to deliver results in a short time, and multiple graphics arrays or even higher-performing GPUs are demanded.
  • Discrete graphics will always be necessary. The pace of development and power restrictions on integrated GPUs will always keep them a minimum of one generation behind. The average user will also see increased benefits of a discrete graphics card as more and more programs/operating systems take advantage of the parallel processing power of the GPU. I think these new applications, in addition to games of course, will continue to showcase the benefit of a graphics card. Very similar to the reason "wireless" will never replace "wired" networking entirely. There is always more demand for bandwidth and "wireless," or integrated graphics in this case, may be good enough...but never fully support the latest generation.
  • Discrete graphic cards should still have better performance, and the end-user should consider the total cost of whole PC. CPU/GPU hybrid solution should have higher cost premium than normal one.

The outlier here really has us scratching our heads. While we were curious to see if anyone had an apocalyptic vision on the horizon, we honestly never expected an outright “yes.” The need for discrete graphic solutions harkens to the very existence of these companies (or a very large portion of their business). That respondent didn’t provide details behind his answer, so this only raises more questions. While not a tier-one  company, they are still a large manufacturer/supplier in terms of worldwide unit sales. Does this mean they are planning to retreat from the graphic card market, either in the short-term or long-term? Or do they see this as a very long-term technological advancement at an unforeseeable date?

In our opinion, we don’t see the video card industry disappearing, one, two, or even five years forward. One has to wonder a bit if CPU/GPU hybrids are as much as a “game changer” as AMD and Intel hope them to be (or if they’re only game-changers for each company’s bottom line, as they cut cost via integration and maintain pricing by offering better performance). Even if they are, these folks have a valid point. The applications and demands made of graphic solutions will continue to multiply, most likely to a degree that outpaces the development of integrated graphics for some time to come.

Combine this with a desire to enable general-purpose computing using GPUs. It is easy to see how these companies will exist in the long term. If you recall the emergence of 64-bit computing, Intel and AMD were both heavily vested in pushing adoption. Fast forward to the present day. We are still lacking a concerted effort by the software development community to adopt 64-bit programming. We still lack a 64-bit version of Firefox, and there is no ETA on a 64-bit Flash plug-in. While the benefits of 64-bit in these two scenarios may in fact be negligible, it shows how slow the software community has been in contrast to what today’s hardware provides. Only recently did Adobe update its suite of apps to support a 64-bit architecture, and we’ve already shown the effect to be massive.

If all of AMD’s and Intel’s wishes come to fruition, the video card industry will no doubt be shaken up. In a worst-case scenario, it seems more likely that within a couple of years, we will see the number of third-party board vendors condense (indeed, industry veteran BFG has already disappeared, and customers have been told directly that the company wasn’t getting enough support from Nvidia).  What doesn’t seem to be changing is the focus on the high-end graphic space. This market will never really be satisfied by integrated graphics.

This makes sense considering what one representative pointed out. “The pace of development and power restrictions on integrated GPUs will always keep them a minimum of one generation behind. The average user will also see increased benefits of a discrete graphics card as more and more programs/operating systems take advantage of the parallel processing power of the GPU.” With general-purpose GPU programming, there should still be a huge performance delta between a high-end discrete graphics card and a “high-end” IGP solution.

If you talk to the people in the motherboard industry, everyone is more confident in their own strengths after watching competitors exit the market amid what we might think of as the second tech bust. It was a wake up call that many needed to focus on what really made brands stand out in the first place: overclockablity, stability, feature set, pricing, quality of service, and so on. In the short-term, the only business really being threatened is the low-end discrete space. While there is volume here, it really isn’t in the channel, if you are looking at the marginal profit. It is a decent revenue stream for tier-one and tier-two board vendors, simply because of the large volumes on lower contracted pricing. However, the real money is in the “sweet spot” from $125 to $175. If this means card makers are going to invest more time and money into making better cards in that range, everyone might be better off having the IGP solutions take the low-end of the spectrum. The war for the low-end discrete space is going to drag out for at least a year and a half, because success depends on multiple factors (price, driver support, discretionary spending, economic climate, etc...), so it feels like there is ample time for all card makers to stay ahead of the curve.

React To This Article