gsrcrxsi

TrekCZ
There is something seriously wrong there. XC3 ultra has TDP 350W and should be as founders edition, with increased power limit it should reach over 400W
You can try MSI kombustor to ensure gpu load, but port royal should reach almost TDP too, especially with water cooling.
 
There might be issue also with CPU, can you run Cinebench R23 and report single core and multi core score, please.
 

I see the same kind of 300W limit on Kombustor too maybe only slightly higher draw. I made a whole thread about this yesterday. search my profile and you'll find it.
 
this XC3 card only allows 105% power limit increase. it can only go up to 366W. which I don't have a problem with really. but it doesn't ever get there. stays around 300W, so increasing power limit in software doesnt change anything now.
 
will report back on R23

Here is your thread on it. Hope you get this figured out.
 
https://forums.evga.com/3080-Ti-XC3-Ultra-Hydro-Copper-low-power-consumption-hidden-power-cap-m3408367.aspx

Single core is 10% lower than it should be. Load bios defaults please and try port royal.

on stock settings, it boosted to 4.7-4.8GHz, 1600 score, as expected.
 
Port Royal with CPU @ default (ram still at 3600CL14 XMP, 1800MHz FCLK) = 12780. exactly the same. still the same 300W limit.
 
hopefully that puts to rest any suggestion that it's a problem with the CPU or that the CPU is in any way involved. it's a GPU limit.

TrekCZ
Can you run Aida memory benchmark? And dram calculator has membench benchmark, you should have membench score 100 with 3600 cl14

I don't know why you expect to see some smoking gun with these random tests lol. these things have nothing to do with the power limit enforced on the GPU. GPUz reports that it's power limited. PerfCap reason is PWR. which means the GPU is throttling itself. it's not waiting around for data from the CPU.
 
but I'll humor you. AIDA64 memory test was unremarkable. all the values in-line with what I would expect. ~54GB/s read speed, 61ns latency yada yada. matches with a quick search online of people running 3600CL14 (i havent tweaked all the little sub timings and I'm not going to. this is the default XMP profile)
 
result from DRAM calculated membench puts it "current result" at 224, which seems in line with what other people report on Ryzen 5000 with similar speeds.

 
it's as it should be for this card. but the limits are being enforced too early. that's the whole point.

TrekCZ
Ok, you could also open monitoring in afterburner, so you have a chart with power, so it would show values when running Port Royal, maybe sometimes it is over 300W or there is hard line at 300W? Also GPU-Z sensors tab to show max value of Board Power and chart with perfcap reason.

I've done that. Afterburner shows a mix of values between PX1 and GPUz
 
it shows ~300W that agrees with GPUz
it shows ~100% PWR which agrees with PX1
PX1 does not show raw value, only %
 
so how is 300W equal to 100% when the power limit is 350?
 
I feel like you're not reading my posts completely. I have mentioned that GPUz shows perfcap reason as PWR. at the same time it shows power consumption at 85% TDP (300/350 = 85%). so why is it power limiting at 300W?

gsrcrxsi

TrekCZ
Ok, you could also open monitoring in afterburner, so you have a chart with power, so it would show values when running Port Royal, maybe sometimes it is over 300W or there is hard line at 300W? Also GPU-Z sensors tab to show max value of Board Power and chart with perfcap reason.

I've done that. Afterburner shows a mix of values between PX1 and GPUz
 
it shows ~300W that agrees with GPUz
it shows ~100% PWR which agrees with PX1
PX1 does not show raw value, only %
 
so how is 300W equal to 100% when the power limit is 350?
 
I feel like you're not reading my posts completely. I have mentioned that GPUz shows perfcap reason as PWR. at the same time it shows power consumption at 85% TDP (300/350 = 85%). so why is it power limiting at 300W?

I would suggest re-flashing or trying an entirely different vBios and if that fails then you may want to contact support for RMA.

talon951
Flashing a different bios is a long shot at best. I'd be trying to get an RMA. No way it's performing normally when it won't even reach the default TDP or at least get close. Not sure how to convince EVGA though.

I've read reports of others having same issue and re-flashing bios or flashing a different one, re-installing drivers etc solved their issue, hence the suggestion. Can't hurt to at least try it before going down the RMA path, assuming there are bios' available for 3080Ti's.
 
I have the same issue on my Kingpin on the LN2 bios but luckily the XOC bios is a thing for KPE's and works fine for me.

talon951

degenerate

talon951
Flashing a different bios is a long shot at best. I'd be trying to get an RMA. No way it's performing normally when it won't even reach the default TDP or at least get close. Not sure how to convince EVGA though.

I've read reports of others having same issue and re-flashing bios or flashing a different one, re-installing drivers etc solved their issue, hence the suggestion. Can't hurt to at least try it before going down the RMA path, assuming there are bios' available for 3080Ti's.
 
I have the same issue on my Kingpin on the LN2 bios but luckily the XOC bios is a thing for KPE's and works fine for me.

XOC bios is a thing for all 3090's. :-)

Yes lol, the point was a different bios fixed the issue so it's probably worth trying before sending a brand-new card in for RMA and being without a GPU for 2-4 weeks or more. Usually worth troubleshooting when something is behaving strangely before jumping straight to RMA. Takes like 15mins to test so why not?

I do want to try a new BIOS. The ideal solution is that EVGA isolates the problem and issues a BIOS update for this exact card. They did something similar when everyone was having problems with the 3080s from what I remember.

I’m hesitant to try a BIOS from another card. There’s the difference of 3x8-pin to 2x8-pin on my card and I think a non-HC card will mess up the RGB lighting scheme.

13970 For me
http://www.3dmark.com/pr/1067259

I think I can get it a tad higher. If I can get it to boost appropriately. If I set it to +195 to core it max clocks to 2165ish but if I set it to +210 it will nuke itself to 2220 and crash. Weird behaviour.

14866:

14801- 3080 ti ftw3 and 5600x on water
http://www.3dmark.com/pr/1103732

13,055 with the 3080 Ti FTW 3 on the OC BIOS but otherwise stock.  Haven't modified any mem or power settings for the GPU and didn't modify the fan curve.  I won't run the fans at 100% thanks to the noise on my daily gaming so won't do that just to get a higher score either.  5900X is set to PBO and was bouncing in between 4.625-5 GHz.

If you want some real comparisons
 
Port royal on a 1080Ti 2272 https://www.3dmark.com/pr/1098924
Time Spy on a 1080 Ti 8093 - https://www.3dmark.com/spy/21153130
 
These are recent runs in anticipation of a new card soon lol