JPR Reports Q3 Graphics Numbers

And Everyone’s Smiling…

Good news everybody! Q3 was up. Overall graphics shipments were up 16.7 percent from Q2 and 18.4 percent over Q3 2010. This comes after an 8.8 percent increase in worldwide PC shipments, meaning that new graphics shipments out stripped new PC shipments. This has had the effect of raising the average number of GPUs per computer to almost 1.6 GPUs. In addition Intel gained 6.4 percent market share while Nvidia was drastically hurt by the loss of its integrated graphics business and dropped 4 percent.

Now let’s take a deeper look at one of these numbers in particular, 1.6 GPUs per PC.  JPR notes that in 2001 the average number of GPUs per PC was 1.15. For basic operation a computer requires input and output devices, and visual displays have proven to be the perennial output choice of manufactures and consumers alike. Thus it’s not surprising that the number of GPUs per PC is trending up, but at the same time you only need one GPU per PC for basic functionality. So why are these extra GPUs showing up in most PCs? Well, as it turns out, there are a lot of additive factors; so let’s run down the list.

The biggest factor supporting this trend has to be Intel’s GPUs, or more specifically Intel’s lack of a fully competitive GPU. Intel’s shipping on-die GPUs with most of their chips these days, the problem though is that these GPUs aren’t enough to really cut the mustard for certain applications, like hardware accelerated video play back, notably because they lack proper frame rate support. They also don’t support the same level of GPU compute capabilities and variety of APIs that chips from Nvidia and AMD offer. Thus it’s not surprising to find most Core-i5 and i7 laptops kited out with a discrete GPUs from one of Intel’s competitors.

In addition, adding a discrete GPU to a laptop makes more sense than ever because of technologies like Nvidia’s Optimus and AMD’s Switchable graphics. These technologies allow the user to pick between low power on-die GPUs or higher power discrete GPUs without wasting power by letting one or the other idle away. Essentially, the biggest factor contributing to the growth of this ratio is OEMs choosing to add a discrete GPU on top of Intel’s less-than-perfect on-die GPU.

Another factor is AMD’s dual graphics options on its mainstream APU based offerings. OEMs like HP have been selling laptops that can use the power of AMD’s A8 and A6 APUs in conjunction with discrete class solutions like the Radeon HD 6750M. AMD was heavily promoting solutions like these at this year’s AFDS and even has its own supplementary naming scheme to describe these setups. Additionally there are more traditional factors like high-end gaming PCs which often use Intel’s i5 and i7 products in conjunction with mid to high-end discrete class graphics and Enthusiast PC that can have up to four discrete graphics cards at once.

Credit - AMD and Nvidia

Now let’s shake our magic eight ball and look into the future for a moment. Will this trend continue? Maybe, it really depends on how well Intel executes with its upcoming Ivybridge chip. Admittedly we do know quite a bit about Ivybridge. For example Ivybridge should effectively kill the market for super low-end  mobile and desktop GPUs, by offering significantly better performance than Sandybridge. But at the same time other details like specific API support and its hardware video acceleration capabilities remain in flux depending on who you ask. On the AMD side of the market Trinity and Wichita are going to continue to reduce the need for more GPUs in the low to midrange PC segment. Thus Intel and AMD are both in the process of killing the low-end of the market to some degree.

So is 1.6 GPUs per PC the crest of the GPU attach rate? Maybe, it’s hard to really pin down, and chances are we’ll see this number continue to move closer to 1.7 over the next two quarters. But after Trinity, Wichita, and Ivybridge all make it to market, in volume, it won’t be surprising to see this number start slipping back down towards one or even below if the definition of APU begins to stick.S|A

The following two tabs change content below.
Thomas Ryan is a freelance technology writer and photographer from Seattle, living in Austin. You can also find his work on SemiAccurate and PCWorld. He has a BA in Geography from the University of Washington with a minor in Urban Design and Planning and specializes in geospatial data science. If you have a hardware performance question or an interesting data set Thomas has you covered.