Since our previous article, waves of new pictures in greater detail have hit the internet – many of which are revealing. These detailed pictures have started a rather speedily rotating rumour mill claiming a conspiracy theory that Nvidia does not in fact have a working Fermi sample and the entire architecture only exists on paper. To cover up for this, Nvidia hastily fabricated a non-working dummy for the audience. The previous pictures shot from distance were rather convincing – the close-ups being questionable.

Fermi closeup [Rumour] Nvidia demonstrated a non functional Fermi dummy?

Read more next page.


Since our previous article, waves of new pictures in greater detail have hit the internet – many of which are revealing. These detailed pictures have started a rather speedily rotating rumour mill claiming a conspiracy theory that Nvidia does not in fact have a working Fermi sample and the entire architecture only exists on paper. To cover up for this, Nvidia hastily fabricated a non-working dummy for the audience. The previous pictures shot from distance were rather convincing – the close-ups being questionable.

Fermi closeup [Rumour] Nvidia demonstrated a non functional Fermi dummy?

One of the more interesting pictures leaked is the GF100 die covered by a heatspreader. This is undoubtedly a massive chip – visually it appears to be as large as a GT200b. We are talking A1 silicon (A2 is the silicon shipped to consumers – if all goes according to plan) fabricated late-August, early-September. Incidentally, this is when the <10% yield rumours hit the web. The chip is marked “7”. Obviously, no one would be marking chips unless they were rather rare.

Other peculiarities are an exhaust vent that is blocked, a PCB which appears to be abruptly sawed off, components appear to have no connections to the PCB, two DVI links on the PCB for a Tesla product (which usually don’t have display outputs; whereas Geforce products do – even though there is one out of place DVI port in the Fermi “prototype”) and so on. All such evidence has been well documented at the SemiAccurate and TPU links below. What is most interesting is – why would Nvidia claim to bring a working Fermi board to the demonstration, but use GT200b based products for the actual demonstrations? Sure, drivers could be a reason.

Fermi is undoubtedly a promising architecture, on paper, and all evidence seems to point to the fact that it exists only on paper at this point.

We will wait for the Nvidia side of rumours on this rather serious accusation. It sounds too sensationalist to be true – if indeed Nvidia wanted to pull off a dummy – a multi-billion dollar company like Nvidia would surely pull off a more convincing dummy? (Naysayers suggest they realized at the last moment that they wouldn’t be having a working silicon, and this was a rush job) Perhaps they were just showing “what Fermi will look like” all along and something got lost in translation? Let’s see how this controversy pans out. 

If true, we are probably looking at well into Q1 2010 before GF100 finally hits retail in quantities.

Reference: TechPowerUp, SemiAccurate