2080ti 2080 Ownership Club

I have an RTX 2080ti, but my benchmarks are low. I'm getting 7886 in Port Royal. My specs are in the sig below.
 
Did you do any tweaking at all yet? Power limit? Fan profile?

I've been using MSI afterburner, and the highest I can clock the core is 82MHz. I haven't tried to increase the voltage, due to the fact I don't want to harm the card. I've also run the fans at 88% and the highest temp I recorded during any benchmark was 76C.

I have increased the power limit to 111, but it doesn't seem to do much of anything.
 
OCing a little (2115/+1000 mem) got me to 9k Port Royal. I've never maxed out my memory since that would require installing PrecisionX rather than Afterburner... but I am happy with this.

https://www.3dmark.com/pr/16703

What's interesting is in all cases I've seen so far DXR features drop power usage compared to straight up rasterized. I wonder if there will be a case one day where that is not true. Maybe no RT, just DLSS? I've never tried the Final Fantasy benchmark.
 
OCing a little (2115/+1000 mem) got me to 9k Port Royal. I've never maxed out my memory since that would require installing PrecisionX rather than Afterburner... but I am happy with this.

https://www.3dmark.com/pr/16703

What's interesting is in all cases I've seen so far DXR features drop power usage compared to straight up rasterized. I wonder if there will be a case one day where that is not true. Maybe no RT, just DLSS? I've never tried the Final Fantasy benchmark.


What is your power target at? 110-120%?
 
What is your power target at? 110-120%?

I flashed the Galax bios. So with +126% it’s 380W.

I also shunt modded but it didn’t gain me anything over the Bios flash... yet. So far it seems ray tracing is such a bottleneck it holds back the CUDA cores where power usage drops a little.
 
I flashed the Galax bios. So with +126% it’s 380W.

I also shunt modded but it didn’t gain me anything over the Bios flash... yet. So far it seems ray tracing is such a bottleneck it holds back the CUDA cores where power usage drops a little.

Yea that's what I was afraid of. I really don't want to void my warranty. Good job on your overclock and score. I think the last time i tested my score it was 8300 ish 2085/800. I can't wait to run that test when my 2nd card + nvlink arrives on tuesday.
 
Nice! I could probably get on the board if I tried. At 90xx and it starts at 92xx. But it’d be at settings I wouldn’t necessarily run 24/7 lol.

Hey Dayaks.. Can you message me information on how to flash GPU bios? I'm having second thoughts.. I might give that a try. My card is in my signature.
 
Hey Dayaks.. Can you message me information on how to flash GPU bios? I'm having second thoughts.. I might give that a try. My card is in my signature.

There is a custom release of NVflash that's floating around which will allow your to cross flash between vendors on using the FE PCB (nvflash_5.527.0_modified).
 
There is a custom release of NVflash that's floating around which will allow your to cross flash between vendors on using the FE PCB (nvflash_5.527.0_modified).

I will research this and see If i can come up with something. I figure there's someone that has done it before.
 
I should point out however that I'm not sure if the Black is running the "A" silicon which would determine how stable it would be using the XC/XC Ultra vBios.
 
I should point out however that I'm not sure if the Black is running the "A" silicon which would determine how stable it would be using the XC/XC Ultra vBios.

Honestly, I don't expect it to at $999.99 per card it was a great value when compared to other 2080Ti prices. I may not get the most insane clocks but my performance has been more than enough to push games using my Acer X34 predator 3440x1440p with max settings within 100-120 fps. Thanks for going out of your way to let me know the options and what to look out for. I really appreciate it.
 
I should point out however that I'm not sure if the Black is running the "A" silicon which would determine how stable it would be using the XC/XC Ultra vBios.

It is not using the "A" binned chip and will not work with a bios made for the "A" binned chips. It cannot be flashed with a higher power limit bios because they are all for "A" chips.
 
Hey Dayaks.. Can you message me information on how to flash GPU bios? I'm having second thoughts.. I might give that a try. My card is in my signature.

I sent you the place to go. The 380W galax bios worked great on my ref pcb asus dual.
 
It is not using the "A" binned chip and will not work with a bios made for the "A" binned chips. It cannot be flashed with a higher power limit bios because they are all for "A" chips.

Thanks for doing the leg work on the research. I will keep this in mind down the road when I pick up the next pair of Ti's. Much appreciated.

I sent you the place to go. The 380W galax bios worked great on my ref pcb asus dual.

Thanks a ton for the information. Unfortunately it looks like it wont work on my cards, but I will use the information as a reference. Thank you kindly.
 
Thanks for doing the leg work on the research. I will keep this in mind down the road when I pick up the next pair of Ti's. Much appreciated.



Thanks a ton for the information. Unfortunately it looks like it wont work on my cards, but I will use the information as a reference. Thank you kindly.

At worst you might be missing out on 5% perf. When you waterblock the cards don’t increase voltage, just max the power limit and increase freq first.

Also when I ran sli with 1080s if they throttled at all my frametimes went to hell. Watch out for that.
 
At worst you might be missing out on 5% perf. When you waterblock the cards don’t increase voltage, just max the power limit and increase freq first.

Also when I ran sli with 1080s if they throttled at all my frametimes went to hell. Watch out for that.

Thanks for the advice. I will remember it when i get my ek parts in february. I am also looking forward to the cpu cooling it will provide. the 8700K is a hot cpu and on air im getting 70-85C at 5ghz. Something that I know will be much lower with a velocity ekwb.


EDIT:

You know what Dayaks I completely forgot about the damn evga step up program. I have 71 days left on this card and ill have a full 90 on the card that comes in tomorrow. Im going to mull it over but I may just step both of them up to XCs so that I can get into the much higher power bands.
 
Last edited:
Thanks to my RMA, EVGA has sent me both a red and a black trim kit bringing my total to red, white or black for accents. I feared shutting off my system to replace them but settled on black. I may go red later if I turn on the LED of my mobo as it has a red glow theme to it.
 
Thanks to my RMA, EVGA has sent me both a red and a black trim kit bringing my total to red, white or black for accents. I feared shutting off my system to replace them but settled on black. I may go red later if I turn on the LED of my mobo as it has a red glow theme to it.

Careful, the red trim may slow performance down by a good 4-5% as the card may get slightly confused thinking it's an AMD product. :p

1st world problems... How your $1350 card looks and which trim accent it should sport for max performance. :D
 
Alright so im comparing another run ( a little lower score ) of TimeSpy Default (2560x1440 etc) with another guys with very similar specs here

https://www.3dmark.com/compare/spy/4494965/spy/5789536#

and I cant for the life of me figure why he is pulling 18.5-22.8% higher in the GPU. If you look at the clock speeds. For some reason mine is showing up as 1/4 of the frequency (known issue) so multiplying mine by 4 and you get 7800 which is true my mem clock is 7800 and his was only 7650mhz. My temps are only 61C. His core clock speed is also lower at 2050. So what is happening?

Lastly, looking at his CPU his is OCd way higher which I will be doing an OC on my proc in feb when i get the custom liquid setup going etc.
Menthol is using a much older driver version than you are (411.63 vs. 417.35). Menthol's score was also done on Windows 10 1803 while yours is with 1809. Yes, 3DMark scores can vary wildly between driver and OS versions.
 
Did some video editing over the weekend, nvenc render with a 2080 ti is pretty nice.

Edit: takes 1 min 28 sec to render a 3 min 20 second file in mp4 hevc using the 1080p nvenc preset in vegas video. HEVC is stupidly slow rendering via CPU only so quite happy with that. This card isn't just for gaming.
 
Last edited:
No formal benchmarks to report, but for my deep learning project the 2080Ti is about 25% faster in FP32 work.

I did some gaming last night, PUBG for about 2 and a half hours, 4K Medium AA, High Shadows, Ultra everything else and it's incredibly smooth.

The experience was on par with my 1080Ti at 4K with low / medium settings.

I don't have the NVLink bridge yet, so no SLI results to report, but single card is insanely fast.
 
Menthol is using a much older driver version than you are (411.63 vs. 417.35). Menthol's score was also done on Windows 10 1803 while yours is with 1809. Yes, 3DMark scores can vary wildly between driver and OS versions.

I didn't even think about that. Makes sense but I am also sure there's a few things i can do to increase my score.

I don't have the NVLink bridge yet, so no SLI results to report, but single card is insanely fast.

I have my NVLink bridge coming tomorrow. Pretty excited.
 
No formal benchmarks to report, but for my deep learning project the 2080Ti is about 25% faster in FP32 work.

I did some gaming last night, PUBG for about 2 and a half hours, 4K Medium AA, High Shadows, Ultra everything else and it's incredibly smooth.

The experience was on par with my 1080Ti at 4K with low / medium settings.

I don't have the NVLink bridge yet, so no SLI results to report, but single card is insanely fast.

If you can take advantage of scaling your data to 16bit or even 8-bit range, you will see a big performance increase with the 2080ti for both inference and training aspects of deep learning.
 
If you can take advantage of scaling your data to 16bit or even 8-bit range, you will see a big performance increase with the 2080ti for both inference and training aspects of deep learning.

Yup, that's the idea. when trying to use fp16 backend I keep running into NaN loss (immediate) with a fully convolutional model in Keras. Debugging that to fully realize the gains that Turing has to offer.
 
Last edited:
EVGA HYBRID kits are in stock as of this posting. The ones in stock are for those using the reference board. FTW3 versions are still out of stock.

https://www.evga.com/products/productlist.aspx?type=18&family=Cooling&chipset=GPU+HYBRID+Cooler
upload_2019-1-18_16-14-13.png
 
If they had this earlier I probably would have got it instead of the two hydro copper water blocks but now I am fully committed to a custom liquid loop! Only have fans, fittings, and tubing left for the end of the month paycheck hype!
 
If they had this earlier I probably would have got it instead of the two hydro copper water blocks but now I am fully committed to a custom liquid loop! Only have fans, fittings, and tubing left for the end of the month paycheck hype!

While hybrids and AIO type products have their place/purpose, nothing is better than a good custom loop! I've found that once you put together and invest in a good setup, you'll be set for quite some time and will probably carry it forward as you upgrade CPU/GPU hardware. One thing for sure, you'll never have to sweat it as to running hot components - good cooling is like a "ticket to ride" when it comes to high performance overclocked hardware... without having to worry about heat issues / things getting cooked.
 
Back
Top