Pages:
Author

Topic: Manual overclocking makes no sense on nVidia cards... - page 3. (Read 3296 times)

hero member
Activity: 1274
Merit: 556
Stop doing so if you have no clue how to do proper OCing while mining.
Let it be to the pro's but don't post bullshit based on your inabilities.

You statement already shows your inability and wrong assumption -> generalize your fault to all others is a certificate every moron likes to get.
100% nobrainer stamp granted to dragonmike

Haha, I would ignore you if you weren't so clueless!

My friend, I'm happy to report I'm on the sixth spot for Time Spy scores on HWbot for water-chilled scores. First if you take my exact hardware combo. Here, see for yourself: http://hwbot.org/benchmark/3dmark_-_time_spy/rankings?cores=1#start=0#interval=20#coolingType=5

I've power modded my card with liquid metal, chilled it with a Hailea chiller, and benched that run with the water at 4C.

I know damn well what the fuck I am doing, you retard.


So, now that we've got this over with, let's get back to topic.

I'm basically agreeing with Nebell on this. It's a fine-tuning exercise, depending on what coin/algo you're going to mine indeed. But for those (the majority on nVidia - except for Equihash mostly) that are core-clock dependent, you'll find yourself power-throttling in most of the cases.

I'm not running my fans at 100%, no, I try to keep the GPU temps in the 60's though (with limit at 79), so fan speed will usually sit around 60-75% depending on ambient.

I've increased voltage limit, so I can be sure there's no problem there, but I've indeed left TDP at 100%. Increasing it will increase clocks a little, but not really worth the while at it'll have a negative effect on power consumption and heat.

As I'm using an auto-switching miner (Nemo's), I found it to be the best short-term compromise. I would probably be able to fine-tune it to be a little more efficient TDP-wise, but tinkering with clocks actually yields nothing as the power limit will kick in anyway.
full member
Activity: 406
Merit: 104
OCing on Nvidia hardware is fairly easy.

Higher TDP > higher voltage > higher top OC > higher temps
Lower TDP > lower voltage > lower top OC > lower temps
Lower temps > higher stable boost clock
Don't crack up your fans to 100%, it's not good for their life span. I run mine at 70% which makes them rather quiet (depends on the manufacturer).
You shouldn't go above 100% TPD on your cards. Don't touch the voltage. If you go above 74C the cards will start throttling down. Nvidia cards are very temp sensitive.

OC only memory if you mine ETH, lower TDP to 50-60% (check your hashrate, 50% TDP is too low for my cards but 55% is enough).
OC both memory and GPU if you mine a coin which can benefit from both.
Experiment, try this, try that, see what's the best.

I have been working with hardware for over 20 years. It still took me 2 weeks to find the perfect OC/TPD/temps for my 2 rigs.
hero member
Activity: 1274
Merit: 556
Yep, they're right up there. Which basically seems to support my assumption.
sr. member
Activity: 518
Merit: 250
What are your hashrates? Have you compared them with crypto benchmarks sites?
hero member
Activity: 1274
Merit: 556
Yeah I've got plenty of overhead on the PSU (1000W Gold powering 3 of the cards, 750W gold the other 2).
I've steepened the fan curves so the temps are in the 60's under full load.

What I'm saying is it pointless to do a manual OC. The card will boost as high as the power limit will allow anyway.

Unless I'm missing something.
sr. member
Activity: 518
Merit: 250
Have you got enough powerful PSU? OC in general is very efficient.
sp_
legendary
Activity: 2954
Merit: 1087
Team Black developer
I'm wondering.
One of my rigs is a 5x1080 (EVGA Superclocked).

Whatever I do in terms of overclocking seems pointless and leads to the same clocks because the cards are power-throttling.

Whether I add a fix offset (eg +100), or raise the curve, I have the feeling it makes no difference at the end of the day because the cards will only boost as far as the (ridiculously low) power limit will allow at the current GPU temperature.

You need to find the reason for the trottleing. Sometimes it can help to reduce the tdp of the card to f.eks 80%. Then each card will draw less power and generate less heat, and can be clocked higher.
hero member
Activity: 1274
Merit: 556
I'm wondering.
One of my rigs is a 5x1080 (EVGA Superclocked).

Whatever I do in terms of overclocking seems pointless and leads to the same clocks because the cards are power-throttling.

Whether I add a fix offset (eg +100), or raise the curve, I have the feeling it makes no difference at the end of the day because the cards will only boost as far as the (ridiculously low) power limit will allow at the current GPU temperature.
Pages:
Jump to: