The current state of the pro GPU market has transformed from an acceptably capitalistic venture to the point where one looks around wide eyed at the rising costs of a piece of silicon and wonders how no trust laws have been broken. Nvidia's (NASDAQ: NVDA) current pricing tactics are something which have raised many an eyebrow but this particular pricing revision was one that, I must admit, resulted in an exasperated sigh from my part.
The game Nvidia Corporation (NASDAQ: NVDA) is playing is one of pure numbers, and involves (technically) legal tactics such as price skimming and utilizing the gap in its competitor's lineup, all the while skirting the edge of legality and trust. The price point strategy and positioning of the new TITAN-X and the Quadro M6000 is what we will be exploring in this opinion editorial.
A TITAN Idea - Some of the perks, without the full cost of going Pro
Let's start with the TITAN branding first. The TITAN branding was originally designed to establish a brand new market. A market which Nvidia dubbed as "Semi-Pro". The logic and rationale was clear. TITAN series GPUs would get unlocked double precision performance but not the ISV and Drivers support of the Quadro Series. A sort of midway between the mostly single precision based GPUs of the usual nomenclature. Because of the unlocked DP, the price tag had received a begrudging nod from me. Having firsthand experience of how much double precision matters in some programs (AutoCAD and the video industry being prime examples), it was something new, and something exciting. As a result, Green gained some more momentum in the Industry, establishing an even stronger foothold in what was already more or less set concrete.
The TITAN-X - None of the perks, but still with some of the cost of going Pro
At the same time, the Quadro K6000 Flagship (with all its lineup) stood proud, with full ISV and drivers support. Nvidia (NASDAQ: NVDA) charged a price tag accordingly, and because of its monopolistic position in the professional industry, AMD could not compete. To be fair, the Quadro K5000 pre-dated the Quadro K6000 with 1/24 FP32 but also a relatively marginal price tag of just over 2000 USD. Something that once again, granted the nature of Quadro branding, could be ignored.
The Quadro M6000: Some of the perks with all the cost of going Pro
Then came the first brand new iteration of GM204 based Quadros, which had Double Precision completely crippled at 1/32 FP32 and were still priced as top entrants. It was a GX XX4 die with the Quadro branding (and all the perks that come with it) so they still got a free ticket from the Industry. The nomenclature also suggested that the GM200 would be the die to look out for. The GM200 finally arrived, with much fanfare, but once again the emperor had no clothes.
The Nvidia (NASDAQ: NVDA) Quadro M6000 flagship was supposed to be the Quadro K6000's successor, will probably be priced accordingly (upwards of $5000), and has 1/32 double precision performance. So basically for ECC, 16-bit color, full ISV and driver support consumers will pay a price thats roughly five times of the same chip without it. And speaking of Error Correcting Memory, I think its worth pointing out that the card has exactly 12 GB worth of GDDR5 memory on board (ECC memory usually has 2 extra chips that handles the ECC calculations, which is not the case here). The significance of that fact becomes clear when you consider the fact that with ECC enabled, the effective memory should fall to around 10 GB which means for the end user to use the ECC perks they would have to give up approximately 2GB worth of vRAM.
In Nvidia's Defense,
As I already mentioned above, while trust laws advocate fair pricing, they do not hinder the pricing tactic called Price Skimming (initially set a very high price and then gradually lower it to fair value). If you take the Titan Z which was originally priced at $3000 but then dropped to $1500, you will get what I mean by price skimming. However, fair value is very hard to determine. In the case of Quadro GPUs, the fair value is mostly determined by the supply and demand of the market, and considering the completely green ecosystem, it becomes increasingly hard to be anything other than what Nvidia wants it to be.
Another point to mention is that the Quadro series of GPUs was not actually designed for FP64 calculations - a very valid counterpoint I imagine Nvidia (NASDAQ: NVDA) could have. The Tesla lineup was, and mostly deals with scientific studies where 64 bit precision is of critical import. With that reflection, this particular op-ed comes to a close.
While in terms of sheer technicality alone, Nvidia appears to be in the clear, but in terms of the pricing model, the question remains: does Nvidia Corporation face an ethical conundrum and potentially a trust problem?
