The review, and most of the comments, fails to address some of the major issues why one would pursue a new Titan (or 1080 Ti).
1. The Titan X is the card for the user who renders by day and games by night, allowing a "best of both worlds" scenario instead of two separate (GTX + Quadro) builds. The article doesn't look at this.
2. The suggestion that "gamer only" purchasers were divided by class I think is less relevant than those measured by what the consumer has upstairs. Informed gamers didn't buy Titans, they bought the Ti. Those who just associate cost with being "better" were the Titan's "gamer only" market niche. An overclocked Ti was faster than a Titan.
3. The comparisons about 1080 to 1070 is this so 1080 Ti to 1080 will be that I think are off the mark. The 980 Ti provided fps increases above 30% when overclocked, the 980 was 25%, 970 was 17%. The difference between the 970 and 980 was very small before overclocking, the difference between the 980 Ti and the 980 was much bigger. The article, and comments so far, haven't addressed this.
4. As for SLI, I'm not in a position to make a judgement as yet. From 5xx thru 9xx, SLI dominated performance wise as two x70s, x60s or even sometimes x50s outperformed the top tier card for less money. The only inconvenience suffered over that time from SLI by our users was waiting two weeks for a BF3 Beta profile. The ROI for 3 and 4 way SLI made it a choice rarely taken so it is not surprising that nVidia dropped support for this option. But we have seen a reduction in scaling in 2-way SLI, and one has to wonder why ?
a) We do see scaling rise with resolution, substantially; so one has to wonder if we have finally reached the point where CPU / Memory performance is limiting SLI performance.
b) nVidia loses money when customers choose two lower cost cards instead of one top tier card and has been taking strides in recent generations to reduce the attractiveness of this option. Of course, they can only take this so far with AMD in the picture but if we look at the 960 for example, in SLI, it couldn't even catch the 970 let alone the 980. And yet two 970s substantially outperformed the 980. Now with the 10xx line up, the price structure, lack of competition from AMD, and lower scaling... the 1080 looks much more attractive over twin 1070s. Why no 1060 SLI while AMD has it on the 480 ? Because two 1060s cost more than the 1070 which edges twin 480s (153% to 147% over single 480) outta the box and that's before adding on the effect of the 1070s 18% to 8% OC capability.
https://www.techpowerup.com/reviews/AMD/RX_480_CrossFire/19.html
So... with no competition in the upper from AMD at this point in time, it is in nVidia's best interests to have poor scaling performance. Will this change when AMD drops their higher tier cards ?
c) Focusing on scaling alone doesn't address the issue if the question is say two 1070s versus single 1080. If scaling is say only 33% at 1440p including games that offer no significant scaling....at the same cost, is this not a batter choice than the 1080 which is only 21% ?. The 52% scaling at 4k certainly brings the Titan X into the realm of the ULMB feature of a G-Sync card and monitor for every game out today.
d) What impact will DX12 have on SLI / CF performance ? Too few games as yet available and too little time for drivers to adapt.
e) Given the above, and the growing prominence of console ports, will game devs decide that supporting SLI / CF is not worth their T & E ?
In short ... the 10xx series certainly make SLI a less attractive alternative for gamers at this point in time. Does this spell the death knell for SLI / CF in gaming ? I think it's too early to tell. Given the continued large ROI which the Titan X delivers on a workstation where a workstation and operator bill out at $180 an hour, $3600 extra in 4-way SLI still pays for itself after < 28 hours
5. Hopefully we'll see a follow up that not only addresses the power and temp issues, but also workstation performance and overclocking which will allow one to evaluate the potential of the card.