Nvidia's increase wasn't really an increase. It was simply recovering what it lost in Q2 when they transitioned from the Maxell to Pascal. The report above further demonstrates that. No one was claiming pc gaming was dead last quarter because Nvidia's shipments tanked. What this does show is that the $300+ dgpu market is a much more significant percentage of the market than even a year ago. AMD sure made a brilliant business decision ignoring the highend to focus where the "bulk of the market" was with the Polaris release schedule.
Increase wasnt an increase? Mind blown!!!
/joking
Had a bunch of funny things to say but I fear that my lighthearted humor might be mistaken. I think things are taken to heart too easily these days and that places of companionship for like-minded PC geeks have shrunk down and down to become this tiny line that we must tip toe on. Like tight rope walking.
But reminiscing aside (and I apologize for being an old school from back in the day), I do think that there are many things we could talk about in the data. More than concluding that Nvidia's increase wasnt an increase, which joking aside, I think I can kind of see what you saying...if you are speaking of volume in individual units.
I think that there was a transition and this can be seen in the data over the last couple quarters. The volume in Q2 and Q3 seems to tell a story for both nvidia and AMD.
From a marketshare perspective, its often concluded that the quarterly chip shipments to AiBs and such exactly represents the real market shift to the end users individual purchases. Which, it doesn't. So for example, if jon peddie reports an increase of 10% marketshare for Q2, this is a percentage based on total shipments and not sales of individual gpus. The shipments become stock to be sold, like walmart or Sears ordering televisions to be sold. Of course, you don't order tons of truck loads of something unless there is the expectation it will be sold. Many people know already how it this works but I wanted to clarify what the numbers really mean to anyone who doesnt.
So what I will suggest is my opinion on what the data means for the past few quarters. I believe it's very true that nvidias transition from maxwell to pascal is visibly apparent in the Q2 and Q3 results. The Q2 dip in market share was due to the fact that nvidia had only one pascal variant ready and it came with obvious volume constraints. I believe had nvidia been able to produce more chips, they would have been able to move more chips. Unfortunately for all the people trashing their 16nm kick off, there is enough data out now for me to say that it appears nvidia had managed to move a lot of pascal chips. So while demand outweighed their ability to produce, its hard to argue they were shipping gm204 in tiny quantities. I would say that it went extremely well for nvidia and that demand was very high and not that there was no supply. chips were coming from nvidia.
Its increasingly difficult to ramp up a new architecture on a new node. The volume always starts off slow and increases over time. It seems though, both nvidia and amd was able to roll out the products they launched in decent quantities, this could have been a huge issue. It often is.
I would say that amd Q2 market share jump was do to polaris chips sold to AiBs to fill the channels and the shelves. This was a high volume product and I believe that there was a large segment waiting and ready for a new product. Channels were ready to be filled = Amd Q2
The following months, Nvidia started replacing all of their maxwell products with pascal counter parts. The channels start filling up for their new products and will eventually level out as they saturate. AMD has done the same with the slots polaris can fill.
The marketshare levels out at some point, it may be now. Which would mean that we are very close to where we were all along, and leveled off after Q2. Or we could still be seeing an nvidia surge as their pascal line fills up the channel. This would mean that their market share will dip down some when they saturate and level off.
If you know what the data is then its clear that reading too much off one quarter is wrong. It takes a few to determine a real uptick or dip. There are other things in play, such as capability. This can have a big effect one quarter and then cause a huge bounce the next, one that might even over compensate as the projected sales fail to meet expectations.
So, take the long view. Then there is so much more to gather from the data. And more value can be had in the conversation