Is Nvidia’s GP106 Pascal based GT1060 a real card or just a cynical spoiler for the press? If you read SemiAccurate’s earlier reporting on the two versions, you probably have a good idea as to where it lies.
A couple of days before the unveiling of AMD’s RX 480 GPU, Nvidia delivered a doorstop to reviewers as a spoiler. By doorstop we mean a card without drivers, docs, MSRP, release dates, or performance numbers, they couldn’t even turn it on. If this sounds like a last-minute PR game to spoil a competitor’s real launch, you might be on to something. The lack of drivers is especially disconcerting considering that the card is simply a fused off GP104, not an actual GP106 die.
This is where the problems begin for Nvidia though, with a GP104 based ‘GP106’ card, it won’t be hard for them to beat the RX 480 in whatever benchmark they choose. They just have to fuse off enough shaders and memory while dialing the clocks back to meet whatever performance and power consumption numbers they choose. It is a great PR stunt but only a PR stunt. Don’t expect this part to ever hit the shelves in anything more than a PR approved quantity if that.
Why? First of all the current GP104 die based cards are in short supply because of this. Officially Nvidia is blaming demand rather then production issues but either way there are nowhere near enough GP104 dies for the GTX1070/1080 parts the moment. If Nvidia really launches the GP104 based ‘GP106’ in July as some are claiming, how is a higher volume part competing for the same silicon going to help the situation?
Don’t expect this GT1060 to ever be a real part, just a press spoiler. SemiAccurate told you about the real GP106 last week, when that comes out expect it to ‘replace’ the GP104 based version while offering much lower performance and actual availability. Nvidia has been doing this for generations, expect a GT1065 or GT1055 to launch with much new fanfare and the old one to disappear in short order if it was ever actually real.
That is the good news, when you look at the details things get really ugly.
Note: The following is analysis for professional level subscribers only.
Disclosures: Charlie Demerjian and Stone Arch Networking Services, Inc. have no consulting relationships, investment relationships, or hold any investment positions with any of the companies mentioned in this report.
Latest posts by Charlie Demerjian (see all)
- HyperX ships it’s 60 millionth enthusiast memory module - Oct 15, 2018
- Bittware/Nallatech water cools 300W of Xilinx FPGA - Oct 12, 2018
- More on Intel’s 10nm process problems - Sep 17, 2018
- Intel puts out another 14nm 2020 server platform - Sep 11, 2018
- Why Can’t Intel Supply Enough 14nm Xeons? - Sep 10, 2018