Originally Posted by ZealotKi11er
If this was 3080, its would amazing but Nvidia is not in that kind of business. I mean sure it can be 3080 that used to be a 2080 Ti in the past. It might cost between 700-1200.
I've been waiting for this topic to catch on.
It's not about NVIDIA being in any kind of business.
The best I can tell from leaks and industry sources:
1) NVIDIA was considering using Samsung as an industry partner for next-gen parts. They continued down this path until it became apparent that Samsung couldn't quite deliver what they were looking for.
2) While that was going on, AMD was in the process of rolling out multiple generations of GPUs on 7nm. At this point AMD has 7nm down to a science. (keep this in mind as we move forward).
3) NVIDIA finally decided, at minimum, to use TSMC 7nm for high end and possibly midrange parts. At this point they are behind schedule.
4) NVIDIA spends a ton of time and money porting their designs to TSMC 7nm. The results end up being somewhat underwhelming.
Yes, this is an engineering sample. Yes things could get better before launch. However, unless someone was testing a really OLD engineering sample, this is close to being representative of the final product. I've been predicting the top TI card to be 30% faster than the 2080ti, and thus far my predictions are looking like they are going to be pretty spot on TBH.
The reason I bring up AMD is that I know for a fact they've been able to hit 2.5 GHz boost clocks on an RDNA 2 part. I've also heard from a third party they may even be at 2.66 or 2.8 GHz for some of their silicon. A lot of this boils down to new power management features and to a lesser extent, other architectural improvements.
What I'm getting at is that NVIDIA slipped this gen. We'll see if they are able to recover, but it looks like AMD might be the one with the faster part this time around. That hasn't happened in a few years now, so it'll be nice.
One other juicy tidbit I hesitated bringing up: AMD apparently is working very hard at optimizing power consumption, including the ability to shut down multiple CUs and boost the remaining CUs to much higher clocks. Based on this information, 3 GHz+ or even 4 GHz+ from future GPUs (not Navi2X, however) are not out of the question.