Nope... but I know that 2080ti in SLI gets FPS somewhere between a 3080 and 3090 - depending on the game. Problem with SLI now is that not every game supports it... It's a dead technology.

SLI is only dead to the knobs on reddit.  It's still fairly popular on overclock.net and for serious OC benching.  One 2080 Ti with good cooling and 2,100+ mhz OC is about equivalent to a stock setting 3080 reference card.  Major drawback of 2080 Ti is that it's only HDMI 2.0, so if you want 4K 120Hz 4:4:4 you need to do it over display port, but HDMI 2.0 4:2:0 chroma for gaming isn't bad.
 
Properly set up and OC'd 2080 Ti should be about 25% stronger than a GOOD 3090 overclocked.

kevinc313
SLI is only dead to the knobs on reddit.  It's still fairly popular on overclock.net and for serious OC benching.  One 2080 Ti with good cooling and 2,100+ mhz OC is about equivalent to a stock setting 3080 reference card.  Major drawback of 2080 Ti is that it's only HDMI 2.0, so if you want 4K 120Hz 4:4:4 you need to do it over display port, but HDMI 2.0 4:2:0 chroma for gaming isn't bad.
 
Properly set up and OC'd 2080 Ti should be about 25% stronger than a GOOD 3090 overclocked.

right, for benchmarking.  For modern gaming not so much.  the last game I played that was actually well optimized for SLI was overwatch.  We ran a pair of 2080tis in SLI and it was very disappointing to say the least.  I come from the era of tri and quad SLI and i kind o miss it but then i kind of don't.  We moved onto 3090s and will never look back.  SLI is actually dead on modern gaming and has been for years.  
 
https://www.gamingscan.com/best-games-that-support-sli/
 
 

haha 2x2080ti's VS 3080.     may as well wait for a used 3080 or the new 3080ti !     single powerful card makes everything so much easier.

WaitingInLine

Dyezak
SLI has a place in Overclocking, AI, Neural Nets, and Deep Learning as well

Rendering as well, I presume.

Rendering usually uses something like OpenCL. NVLINK with the titans that allowed you to share the memory space between cards effectively doubling your memory space... which was a benefit. But most OpenCL workloads that I know of don't care about the SLI connection, since it's just compute and the cards don't need to be able to communicate to each other.  I know in some other OpenCL compute workloads, SLI is discouraged.

WaitingInLine

q5sys
Rendering usually uses something like OpenCL. NVLINK with the titans that allowed you to share the memory space between cards effectively doubling your memory space... which was a benefit. But most OpenCL workloads that I know of don't care about the SLI connection, since it's just compute and the cards don't need to be able to communicate to each other.  I know in some other OpenCL compute workloads, SLI is discouraged.

 
I see. But you can still use them 'separately' at the same time to cut the workload in half, essentially?

Yes, as long as the program is properly written to take advantage of multiple GPUs, it will scale excellently, and it wont run into diminishing returns that gaming with SLI does.  Because each card will be assigned work that's not dependent on what the other cards are doing at the moment.  My Photo RAW editor (Capture One) can utilize multiple cards, and what it'll do is if there's 100 images to be processed, during the final render it'll task out one image to each card as they need work.  So if you have 2 cards, it'll be twice as fast... if you have four cards, it'll be close to 4x as fast... etc.

SLI is dead. Most games don't support it, so I'd advise to hold off on SLI.