Chicago (IL) - Intel is proud of its dominant position in semiconductor in production technology and especially the fact that, for as long as we can remember, has led the industry in terms of the smallest chip structures. Its 45 nm technology is still at least one year ahead of AMD. But it appears that Intel will have to give up that lead next year, at least for a few months, when GPUs will make their transition to 40 nm.
This was bound to happen sooner or later. After speaking with several of our sources at ATI (AMD GPG) and Nvidia, we were told that a 40 nm GPU manufacturing process is on the way for first half of 2009. In fact, both companies are working on parts that should capture the spotlights at CeBit 2009 in Hannover. Both low-end and mainstream products are ready to be manufactured in 40 nm soon and should be on display at the tradeshow.
It appears that TSMC’s previously announced $10 billion investment in manufacturing technology is yielding results already, since the company is now able to develop 45 nm and 40 nm processes at the same time. The next step for TSMC is either 32 nm or 30 nm - or below. Samsung is investing heavily in 30 nm, but that is for DRAM only.
Intel has 32 nm CPUs still in development at its research, development and production facilities in Hillsboro, Oregon. 45 nm Nehalem CPUs will be the focus at the upcoming fall IDF, but it is generally expected that prototype 32 nm processors will be first shown at the company’s spring developer forum in H1 2009. Production of the chips should begin early in H2 2009, with volume shipments beginning in late Q3 or early Q4. First chips should surface in commercial products in late 2009, while 32 nm will be a 2010 topic for the mainstream buyer.
However, by that time, millions of 40 nm GPUs will have shipped already and you can bet the farm on the fact that both AMD and Nvidia will be pitching that story to the media in the same way Intel did in previous years.
According to TSMC’s internal roadmap, the company will offer three different production nodes. CLN40G is the general purpose process and will be used by the GPU manufacturers. CLN40LP is the low-power process and will be used for the production of notebook derivatives of GPUs manufactured in the general purpose process. The third 40nm node is CLN40LPG, which is targeted at manufacturing chips for handheld devices. So expect Nvidia’s Tegra chips to shrink all the way down to 40 nm.
TSMC is likely to launch a 32nm process in a similar frame with Intel. CLN32G is scheduled for a roll-out in Q4 2009. A low-power version will follow about one quarter later. If TSMC is able to stay with its roadmap then we may see 32 nm GPUs in 2009. Intel is expected to launch its Larrabee cGPU on 45 nm in 2010 and move it to 32 nm as soon as possible.
As things shape up right now, GPUs will overtake CPUs for the first time in history of IT in terms of production nodes. Of course, a lead in chip production tech is defined by many more components than size, but the fact that GPU production nodes will surpass Intel CPUs is a significant event.
If we look at the conversion of the complete GPU line-up from 55nm and 65nm to 40 nm - ahead of CPU cycle - it showcases how much more important the GPU has become and will attract more attention to the technology than before. GPUs were introduced by 3Dfx (big "D") and Nvidia, trailing CPU manufacturing processes from Intel and AMD by two to three generations. Now GPUs are about to leap ahead.
Kudos to TSMC for developing the 45 nm and 40 nm half-node die-shrink at the same time.