Curious to see what kind of results I can squeeze out of my 3080ti FTW3 Ultra. I have excellent air flow and case temps. My gpu usually stays just under 70 degrees most of the time with an ambient room temp of 75 degrees Fahrenheit.

I've gotten the core up to +170 but it's only stable in 15% of the games I tested although I haven't played with voltage or power much (no manual curve adjustments).

I've got the memory set to +600. I haven't tested with any other values but I'm sure I can raise it higher.

Yesterday I managed to get a pretty crap Patriot ram kit that's 3200mt CL16-20-20-20-40 to OC to 3800mt CL16-16-16-36 (and many more changes) at 1.4v without using the dram calculator from 1usmus. Before I buy a new kit of better memory I'm going to see if I can get any better out of this using the app.

So today is officially OC day for me as I prepare for Forza Horizon 5 tomorrow

Ive gotten +108 core / +500 mem. Could go higher on memory but heard reports of poorer performance at higher memory settings. Using Matlab and performace data points, I found maybe a 1% gain with the memory overclock in games. The 108 core did make a noticible improvemnet. I put an ek waterblock on my FTW3 3080Ti and card runs stable at 2040-2055 mhz in all games. If the temps are kept in check, youll see a larger perfomance than just craming in more core clock. 

Mgilz8
Ive gotten +108 core / +500 mem. Could go higher on memory but heard reports of poorer performance at higher memory settings. Using Matlab and performace data points, I found maybe a 1% gain with the memory overclock in games. The 108 core did make a noticible improvemnet. I put an ek waterblock on my FTW3 3080Ti and card runs stable at 2040-2055 mhz in all games. If the temps are kept in check, youll see a larger perfomance than just craming in more core clock. 

That's because the memory is ECC or Error correcting code. So what basically happens is if you overcook it higher than the chips can handle errors will occur which generally crashes the game but as long as there aren't to many errors the ram can correct them. It does this by having more chips on the dimm that hold parity bits... Think Raid 5 or Quickpar.

When it corrects an error it adds latency which slows things down (obviously) so with these cards it's super important to properly overclock the memory and test it.

I got my memory to +600 and I played a game with it at that setting along with +165 core and perform was fantastic. It was Forza horizon 5 at 4k max settings and I was between 85 and 115fps depending on what area I was in.

I still need to see if I can squeeze more performance out of the 3080ti FTW3 Ultra but I really wanted to play the game and I had spent the last 2 days overclocking and lowering the timings on my ram. At 1.4v (oh boy) I got a 2x16 kit of 3200mt CL16 to 3800mt CL18-19-19-38 1T.

So far I've gotten mine stable in games at..
+125 core +1000 mem

That's with power at 114% and voltage maxed. I'm considering flashing the EVGA more power bios and see if I can get my core any higher. Thermals certainly aren't an issue at my current settings. Not even close to being a consideration when changing settings so I think using the other bios wouldn't be an issue.

Pretty sure I'd get crashes (assuming I can clock the core higher) before thermals became an issue even with the extra potential power use. EVGA's air cooler on the FTW3 Ultra series is top notch. I had a 2080ti FTW3 Ultra that ran stable at +170 core +1000 mem on regular bios and also had amazing thermals. Basically the same as the 3080ti.

I've also upgraded system ram to Samsung B Die. So now I'm at 4200mt CL15, dual rank, dual channel.

2 weeks (most of it waiting for and OC'ing sys mem) and I've improved performance quite a bit.

No 70 is fine for an air card run.

Don't know if you were replying to me or someone else but even after an hour of full load I'm usually at about 62-64 degrees. My case/fan setup has really good airflow plus I actually installed a way to grab cold air. I'm warmer months I can get air from the ac directed at my front intakes and the bottom fan feeds air directly to the gpu. In winter I can pull cold air from outside.

The temps above are without using the cold air so if I was to use it then things would definitely be cooler. It's like 40 degrees F outside.

If I find I need more cooling after the bios update and assuming I get enough extra performance out of my core to make it worth it then I'll probably 3d print a water block at work. I'm working on a system to run the liquid through my cooling system so my loop will never actually normalize, it will stay cool.

Stepped up to this 3080ti ftw3 ultra.
 
14500 Port Royal I would link but I'm a new member since 2017.
+180 Core Clock
+1800 Memory
On Air
 
19347 Time Spy (GPU 21394 CPU 12547) 10850k
 
Is this considered a decent sample.  I'm undervolting it for some games.  

I would think this is absolutely above average for any 30 series on-air, most will not go above 150. All you need to do is copy the link to your show run page.

Here are some numbers I just tested this morning, I am water cooled as well temps around 42c. I did not touch the voltage at all.I had the numbers backwards doh on my part.
 
Time Spy                  140/1200    Overall 18705 GPU  21609
                            
 
Fire Strike Extreme    140/1200  Overall 21645 GPU  25025
                             
 
Port Royal                140/1200  14523
                              

rmorse27
Here are some numbers I just tested this morning, I am water cooled as well temps around 42c. I did not touch the voltage at all.
 
Time Spy                  120/1000    Overall 21639 GPU  18784
                                120/1200    Overall 21501 GPU 18684
 
Fire Strike Extreme 120/1000  Overall 21707 GPU  24945
                             120/1200  Overall 25011  GPU 21949
 
Port Royal               120/1000  14324
                              120/1200 14430

interesting, with this level of PortRoyal score, your GPU score for Time Spy should be around 21000 yet you only have 18700