A new report suggests Nvidia has halted production of Turing GPUs in preparation for an imminent Ampere ramp-up. An end to production doesn’t mean the launch of Ampere is immediately imminent — there’s going to be a certain amount of volume in the channel to soak up — but it definitely indicates Nvidia is prepping the next-generation GPU’s launch.
Nvidia, apparently, has told its GPU manufacturers to raise prices on high-end cards to reflect the lack of production. I suspect this is a reaction to what happened several years ago, when Turing debuted. Nvidia — which has been accused of deliberately misleading investors as to its product split between gaming GPUs and cryptocurrency back in 2018 — wound up with a glut of GTX Pascal cards it had expected to sell, at exactly the time that Turing was launching at significantly higher prices.
Ampere is expected to retain GTuring’s high-level organization but we don’t know anything about the resources dedicated to each task yet.
Nvidia’s wounds back in 2018 were self-inflicted, driven by the company’s decision to raise prices rather than holding to GTX 1080’s excellent value. The situation undoubtedly improved after this time last year, with the RTX 2080 Super and 2070 Super debuted with higher performance and lower prices, but the damage was done — assuming Nvidia thought it was damage in the first place. It isn’t clear they did.
The price increases contemplated by the report at ITHome appear to apply to Turing cards. A price increase for Ampere over Turing isn’t specifically stated, and I’m not sure Nvidia would communicate that information two months ahead of a theorized September 17 release date. Price is always the last thing a manufacturer finalizes, and handing out its price data this early would make it trivial for AMD to make whatever adjustments are required for RDNA2 to launch aggressively against Ampere. Ending Turing manufacturing early and suggesting OEMs raise prices will prevent the situation that occurred in 2018, where cheap GTX 1080’s selling for $500 blew holes in Nvidia’s attempted value proposition around Turing.
High prices on the RTX 2070 Super, 2080 Super, and 2080 Ti will definitely make Ampere cards look more attractive by comparison, but only to a point. If AMD is telling the truth about its ability to deliver a further 1.5x performance-per-watt improvement for RDNA2 over RDNA — and I want to say up-front that’s a very large gain without a process node shrink to support it — then RDNA2 should be at least in Ampere’s ballpark. It’s been 6-7 years since AMD matched Nvidia’s power efficiency at the same process node, so even being within 10 percent of Ampere would be a real step forward for Team Red.
Now Read: