News

Nvidia’s Intel-based IGP

(from JPR) If you asked any of the executives at Nvidia if they were planning to offer an IGP three years ago, they would have said something like, no way, Who’d want to be in that low margin business? We have way too many other things to do, etc. And, at the time, they probably believed it, or at the ...

Robert Dow

(from JPR)

If you asked any of the executives at Nvidia if they were planning to offer an IGP three years ago, they would have said something like, no way, Who’d want to be in that low margin business? We have way too many other things to do, etc. And, at the time, they probably believed it, or at the least were convincing in their responses.

New semiconductors take time to develop and so when Nvidia introduced their Geforce 6100 family of AMD-based IGPs in September 2005 we knew the earlier comments were a smoke screen, but that’s OK, Nvidia has no obligation to tell us or anyone else what their plans are. However, when asked at that time if they planned an Intel version we got the same song and dance—conclusion—you betcha they were going to come out with an Intel part, now the guessing game becomes one of when.

Yes, there were some technical differences, for AMD Nvidia added a HyperTransport bridge, for Intel they had to do a Front Side Bus interface and have a license. As it turns out the company had FSB licenses from various acquisitions and cross-licenses.

So now Nvidia has brought out their first official Intel IGP, the GeForce 7150, 7100 and 7050 chips which work with Core 2, Pentium and Celeron, as well as the 13333 MHz FSB Penryn due to launch in November (LGA775 socket). Nvidia says the IGPs are aimed at desk¬tops priced at $400 to $600, and expects the MoBo to sell for $150 or less.

In private (and now probably public) showings, Nvidia was comparing the 7150 to the Intel G33. Wasn’t much of a comparison, and wasn’t really a fair fight. Given the G33 (a renamed and polished up i965g) is two years old. When challenged on the point and asked why they didn’t use the G35 as a comparison, Nvidia said they chose the G33 because Nvidia’s GF&150 would compete at the same price point—so that’s a good answer—as good or better than a G35 for a G33 price.

The 19 W, GF7150 can use 800 MHz DDR, offers optional DVI or HDMI output (one or the other) with HDCP The IGP allocates up to 128 MB of system memory and offers all the things you’d want in a chipset (e.g., multi SATA, USB, gigabit Ethernet, USB, etc.)

The graphics core runs at 600 MHz for the 7150 and 7100 and at 500MHz for the 7050 low-cost version and all the parts work with DirectX 9.0, shader model 3.0.

Nvidia says mobos with 7xxx IGPs will be available later this month from Abit, Asrock, Asus, Biostar, Colorful, ECS, EVGA, Foxconn, Galaxy, Giga¬byte, Hassee, Inno3D, Jetway, J&W, Maxsun, MSI, Onda, Palit, PC Partner, PNY, Supox, Unika, XFX, and others.

Nvidia has been extremely successful with their AMD IGPs and that’s one of the reasons AMD didn’t dump Nvidia in favor of ATI after they made the acquisition. Nvidia went from zero in 2005 to about 73% market share of AMD-based non-portable IGPs in Q2’07.

Intel-based IGP machines represented 78% of the desktop IGP PC market in Q2’07, or about 27 million units. In Q2’07 Nvidia sold about 5.6 mil¬lion IGPs for AMD machines. Intel has about 65% market share of Intel non-portable IGP machines (desktop, server, embedded, etc.) The remaining 35% is shared by AMD, SiS, and VIA, with VIA being the largest in the group. Therefore, Nvidia has a TAM of about 9.45 million units (using Q2 data.) If VIA withdraws from the chipset business as is strongly suspected, and at the least doesn’t gain a license for the 1333 MHz FSB and the on-going systems, and if the 690I from AMD is their last Intel-based IGP, then Nvidia is in position to make significant gains, repeating history, and without doubt, Drew Henry and Jen Hsun Huang saw this opportunity when they decided to go ahead and enter the Intel market.

That means Nvidia will face SiS, who isn’t gaining any market share and doesn’t have Nvidia’s graphics capabilities, and AMD and VIA who are backing out of the market, leaving basically Intel for Nvidia to deal with.—JP