Industry Watchers Suggest NVIDIA May Launch New GeForce Series Late Q3

[Guest_Jim_*]

In a Digitimes article covering the gaming notebook sector, the news outlet cites industry watches expecting NVIDIA to launch its next GPU series late in the third quarter. There have been a lot of rumors and speculation concerning when these highly expected GPUs will finally launch, with a number of suggested launch and announcement dates already behind us. This new expectation is based on the belief there will be significant increase in demand for gaming notebooks in H2 from gamers wanting to upgrade their systems following the release of the new GeForce cards.

Technically this is still a rumor and WCCFtech does point out one scenario that is probably not what any gamer would want to see; a staggered launch. It is possible, if highly unlikely, for NVIDIA to launch these expected mobile GPUs first with the desktop parts following later. Also something to consider is the reporting that NVIDIA has an oversupply issue, with one Taiwan OEM actually sending back 300,000 GPUs. NVIDIA would likely not want to diminish the interest in these unsold parts with a new series launching, but with demand too low to sell them and significant interest in the next series, that might be exactly what it needs to do.

Source: Digitimes and WCCFtech

Back to original news post

[Braegnok]

I think both Intel and NVidia may have underestimated AMD this time, and they are both in for a huge loss of market share on all fronts. 

 

AMD has a strong roadmap for Q3, H2,.. with Predator Helios 500 notebook, second-gen Threadripper, and 7nm Radeon Instinct Vega just to name a few.

Edited June 23, 2018 by Braegnok

[Guest_Jim_*]

Intel, I think, definitely underestimated AMD and perhaps even what is technologically possible with how the company settled on incremental improvements than advanced innovation. NVIDIA though, I'm less sure if it underestimated AMD or perhaps overestimated itself. What I mean is, the story about the unsold GPUs with only a decrease in demand and not an actual crash implies NVIDIA thought the demand for its products was much greater than it actually was, with only inflated pricing holding back sales. AMD's roadmap is strong, compared to what is has been at least, but NVIDIA has unprecedented mindshare. A significant enough stumble could really weaken the company's position. The GPP controversy I think already turned some people from NVIDIA, but even if that sticks the effect might not be felt until they choose to upgrade. I do think there is a correction point to this mindshare, where once enough people try and talk about what AMD offers, it would build and lead others to better consider the products.

For example, AMD recently put out videos about Enhanced Sync, and in at least one corner of the Internet I have seen people dismiss it because they only compare it to NVIDIA's Fast Sync and recognize that feature as a weak and poor technology. It just bounces your framerate between multiples of your monitor's refresh rate, but when I tested that and Enhanced Sync in Serious Statistics Pt. 2: The Sync Encounter my data and my experience indicate it is in fact a driver-side triple buffering. It achieves the goal of sending the newest completed frame to the monitor but without any restrictions on the engine rendering. It is the superior technology, but because people only identify it with Fast Sync and assume it is the same, Enhanced Sync is talked down and dismissed.

[Braegnok]

For a vary long time, Intel had a process advantage at any given node over AMD, and were also typically a node or more ahead of AMD. These days, a major shift has taken place: everybody is on essentially the same node (Intel 10 nm and GF/TSMC/Samsung 7 nm being in rough parity, density-wise), but the production readiness of Intel versus the world is binary.

Intel has seen several problems at 10 nm, with a horrible trade-off between clock speed and leakage, which will not resolve anytime soon. TSMC, GF, and Samsung have spread there 7 nm development investment across many markets and customers, Intel has but one (themselves). The foundries appear to have 7 nm working quite well, and GF has the added advantage of having subsumed IBM's 7 nm high performance expertice and process knowledge. 

Another vary important theme is that of large monolithic dies (Intel) versus small, high-yield, interconnected dies (AMD). 

Intel is stuck for at least a year or two with large monolithic dies; AMD has pioneered small interconnected die technology (as they pioneered multi-core, 64 bit x86, etc). This is a major cost/yield advantage for AMD. It is hard to understand how Intel can respond to AMD's 2019 high server core count and expected 12 core desktop CPUs  with a monolithic, large die approach. Intel will not innovate there way out of their problems anytime soon.

 

2017 was the year of AMD's launches; 2018 will be the year of the ramps, and 2019 will be the year of performance and cost domination.

 

"The future's so bright for AMD, I need to wear shades".