Blog

Discretes are dead–long live discretes

The death of discrete GPUs by integrated graphics has been predicted every year for the past ten years and every year for the past ten years, the reports of death have been highly exagerated — even this year. If you look over time at the ratio of discrete GPUs to integrated GPUs (including the embedded GPUs in CPUs like Fusion ...

Robert Dow

The death of discrete GPUs by integrated graphics has been predicted every year for the past ten years and every year for the past ten years, the reports of death have been highly exagerated — even this year. If you look over time at the ratio of discrete GPUs to integrated GPUs (including the embedded GPUs in CPUs like Fusion and Sandy Bridge) you can see an obvious trend.

(I’m using the term “GPU” liberally to represent all graphics controller/processors. In reality, we didn’t have GPUs prior to 2001.)

But what the chart doesn’t show is the effect of a discrete and an integrated GPU being in one system, which has come to be known as “double-attach”; which started in 2002.

Double-attach is due to several factors: After market purchases of AIBs by DIY’ers and IT departments trying to squeeze more life out of a PC, OEMs building multiple SKUs with a one-size-fits all motherboard that has integrated graphics on it, and most recently dual mode or hybrid operation techniques like AMD’s Switched Graphics and Nvidia’s Optimus.

More GPUs ship than PCs. That’s because many systems using PC parts are built for POS terminals (from ATMs to cash registers at the supermarket), industrial, scientific, military/transportation systems, and CE devices like STBs. Those items don’t get counted by Gartner and IDC as PCs. So the graphics that go into them have to be factored out—it’s not an exact science because there are also some parts that are drifting through the channel and/or in inventory at ODMs. We think about 35% of the GPUs are in double-attach PCs, and about 3% of the discrete GPUs are double-load (two AIBs per PC or an AIB with two GPUs).

When you compare the Discrete GPUs to Integrated relative to just PCs you get a decidedly different perspective.

And when you look at the rate of change of Discretes and Integrated you can see the glory days of growth for integrated are over.

But the story isn’t over.

AMD’s Fusion and Intel’s Sandy Bridge are selling well (see Market Watch, and story in this week’s issue of TechWatch). However, they are replacing the older IGPs and not (yet) cannibalizing low-end AIBs with discrete GPUs.

Also we’re going through a secondary economic slump and when that happens people tend to put off buying new products, or they buy low cost items. This is particularly true in the US where buy cheap is the norm. In BRIC and Western Europe there is a stronger appreciation of the benefit of discrete GPUs. And the OEMs like them for the sell-up opportunities and higher margins they bring.

The US is an anomaly and still 40% of the market so it will influence the results. However the growth rates in the BRIC countries are promising and will, combined with other factors like GPU-compute, keep the growth rate of Discretes up. We may not see that this year but our forecast indicates a good future, and a long future for Discretes—long live the Discrete.

Integrated GPUs are growing 38 times faster than discrete GPUs

Integrated GPU growth has slowed while discrete has picked up

Growth rates of Discrete and Integrated since introduction of Integrated