Ask About Computers Blog

The largest maker of computer graphics in the world is not NVIDIA or AMD. It’s Intel. Intel is the largest by virtue of its integrated graphics.

Because integrated graphics makes up most computer graphics, and Intel makes most integrated graphics, this makes Intel the largest supplier of graphics in the world. Bigger than NVIDIA, bigger than AMD.

To be sure, not all integrated graphics is made by Intel, but most of it is. To this end of tackling integrated graphics, NVIDIA has positioned its motherboard GPUs.

Integrated graphics is integrated into the chipset. The installed base is so large that the CEO of NVIDIA recently called it the “integrated graphics opportunity”.

The Largest Possible Market

In order to offer motherboard GPUs as a viable alternative to integrated graphics, NVIDIA had to lower costs. Integrated graphics occupies “the most sensitive price segments”.

In the past, NVIDIA seems to have been content to let ATI and Intel have much of the pie of volume graphics with its lower profit margins. All this may be changing.

By lowering costs, NVIDIA is targeting motherboard GPUs at “the largest possible market”.

Off to a Great Start

While integrated graphics have been around for some time, motherboard GPUs on the other hand are a recent development.

According to NVIDIA, motherboard GPUs for both AMD and Intel systems are “off to a great start”.

The First Of That Category

The motherboard GPU for Intel is the first of its kind.

NVIDIA initially introduced motherboard GPUs for AMD computers. Only after there were motherboard GPUs for AMD did NVIDIA introduce a motherboard GPU for Intel systems, the MCP73.

The motherboard GPU for Intel is the “first product of that category”.

Traction in the Intel Platform

Motherboard GPUs show promise. The mGPUs for AMD seem to be doing well.

The motherboard GPUs on the Intel side recently began “to see traction”.

Early Adoption Wonderful

The motherboard GPU for Intel was one of two classes of devices in which desire for the GPU recently exceeded NVIDIA’s ability to make. Demand outran supply.

Of the mGPU for Intel, the “early adoption has been wonderful”.

The other GPU that NVIDIA couldn’t make enough of, and still can’t, was the GeForce 8800 GT.

GeForce 8800 GT

In order to meet burgeoning demand, NVIDIA has expanded its foundry partners (DIGITIMES). Still these things take time, one would think.

In the days and weeks that followed the launch, places that sold graphics cards couldn’t keep the GeForce 8800 GT in stock.

There was some discussion as to why this was, with wild accusations of poor execution on NVIDIA’s part and that NVIDIA simply hadn’t made enough cards. There was also a rumor of component shortages for certain parts that went on the boards. Whatever the case may be, there were certainly not enough cards.

At the time of this writing, availability is still problematic, often as not.

NVIDIA will soon release another version of the GeForce 8800 GT, this time with less memory, the GeForce 8800 GT 256MB. The performance is almost as good under most circumstances, and the price a tad lower.

Just above the 8800 GT in the graphics-card hierarchy sits a new card set to be released December 11, the GeForce 8800 GTS with 512MB of memory, not to be confused with the card of the same name, GeForce 8800 GTS but with 640MB of memory. The new GTS is based on a new manufacturing technology of 65nm. The old GTS was built on 90nm technology. The new GTS card is also based on the same graphics chip as the 8800 GT, the G92.

NVIDIA F3Q08 (Qtr End 10/28/07) Earnings Call Transcript

NVIDIA recently had an amazing quarter. The company more than doubled its earnings. It scored its first billion dollars. NVIDIA is the first dedicated “GPU company to reach the billion dollar quarter level”.

Nevertheless things could have been better. For NVIDIA was and still is capacity constrained.

Capacity Limited

In the quarter before last, NVIDIA had phenomenal results too. The stock, however, took a dive. Why?

The reason the stock took a dive, quarter before last, was NVIDIA’s foundry partner–TSMC–was running at or near capacity. Things were thus “tight from a capacity perspective”.

In the event, it turned out that NVIDIA was production limited in two primary areas. Motherboard GPUs and some GeForce 8 parts were “capacity limited or production limited”.

In particular it was motherboard GPUs for Intel and GeForce 8800 GT devices that were capacity constrained. Capacity in fact still is constrained. These GPUs NVIDIA “just can’t make enough of”.

Source

NVIDIA F3Q08 (Qtr End 10/28/07) Earnings Call Transcript

In spite of the recent setbacks in housing and the credit markets in the U.S., most component makers of computer equipment are in the midst of a boom. Results of the previous quarter ended even better than most third quarters, which are usually pretty strong.

The strength that was manifested in the previous period shows signs of continuing into the present period. For instance, the last two weeks of September saw an inordinate number of orders for Intel chipsets. Since chipsets precede CPUs, this boded well for the period we are currently in.

Chipsets “tend to lead microprocessors in terms of purchases because they get put onto a motherboard before the processor”.

The time it takes between ordering chipsets and ordering CPUs to go with the chipsets can take anywhere from about “four to ten weeks”.

The danger to the current period is not one of insufficient demand, but of too much demand: not being able to make all the CPUs and related chips that people want, or having a glitch in production and too little inventory to fall back on.

An overabundance of demand may sound like a nice problem to have, but maybe not if it’s severe.

Below the Comfort Zone

Chips and chipset reserves were already low from the exceptionally robust third quarter. Inventories were “below the comfort level”.

Inventories now appear to be lower still for certain types of chips, like mobile chips and chipsets.

Intel recently entertained the possibility that they would be unable to meet demand for processors in the present period: “if demand is significantly above the range that we’re expecting we’ll be hard pressed”.

This scenario of too-much demand appears to have in fact materialized.

Maybe the first to break the news was Fudo. “Intel expects massive shortages” (Fudzilla).

Other reports soon followed. “Intel isn’t meeting demand”, said the president of one computer company (IDG).

And “the most needed components are Intel’s CPU and WLAN modules” (DIGITIMES).

Not All Silicon Wafers Qualify

Lean inventories and tight supply might not hurt so much so long as the ramp of Penryn–Intel’s first 45-nm family of microprocessors–proceeds as planned.

The potential problem is that this is a period of transition for Intel, which is moving away from 65-nm to 45-nm transistors, and in any transition between process technologies there will be large numbers of silicon wafers that are not good enough to sell. In the previous period, Intel built lots of 45-nm processors “that did not qualify for sale”.

While this appears normal for the ramp of a new process, if Intel experiences complications, things could turn ugly: Intel supplies most of the world’s CPUs and has little 65-nm inventory in reserve.

Yield rates should improve as process technology and microprocessor designs mature. For the sake of the industry, let’s hope so.

Source

Intel Q3 2007 Earnings Call Transcript