Why is the clock on the gtx 10xx series so fucking unstable? even when watercooled

Why is the clock on the gtx 10xx series so fucking unstable? even when watercooled

Other urls found in this thread:

youtube.com/watch?v=w-3fi1ovAP0
twitter.com/SFWRedditVideos

It's running exactly two days late!

...

GPU boost 3.0 is finicky. It has discreet temps when it tries to boost further creating more heat, leading it to pull back on the boost. Your temps must be on the edge of a boost point.

what? it's watercooled ffs, look at the temp, it's around 55c

are you retarded? it's literally depends on load + power + temp + volt limits. boost 3.0 IS MADE TO CONSTANTLY TRY TO MAXIMIZE THE FREQUENCY IF NO LIMITS ARE REACHED *unless* GPU usage/load is low. Use fucking GPU-Z.1.10.0 to check what limiter you're hitting it's not rocket science.
imbeciles like you should stick to consoles 2bh

nice meme gaymer site graph btw, really informative on what GPU was running

>my temps are so low mom why doesn't it show 2000MHz+ constantly!?!?! Moooom
at least do some research on what you buy

2x 8pin power, watercooled, 100% load.

>2x 8pin power
the fact that you're mentioning this is indicator that you're tech-illiterate
>100% load
clearly with some crap that stops at or lags / loops at equal intervals

1 - initial max boost frequency, after that GPU will slow down if factors mentioned before are affecting it
2 - GPU is trying to stay close to 1 frequency but limiters are in play and temp of all components is rising
3 - all temps are constant (not only fucking CORE temp) and GPU finally stabilizes at constant 2000MHz while staying within limiters that are defined by a) BIOS b) user set stuff like power/temp target and volts

once again judging by the graphic they were running some poo in loo on it

>3
>calling that stable
fuck off nvidia babby

It isn't on custom cards with higher power limits. My 1070 Gaming X maintained a solid 2050MHz boost clock in-game and the 1060 SC I replaced it with maintains a solid 2088MHz.

>2x 8pin power
Except that's bullshit, the hybrid card tested in the OP has a single 8-pin, because it's just a cut up reference card.

youtube.com/watch?v=w-3fi1ovAP0

you're retarded beyond help. the only reason i bothered responding was because I'm doing the night shift at work. buy amd next time loser

those spikes happen every 2.9 minutes so it's bench looping. kys loser

Boost 3.0 exists solely to help regulate the piece of shit that is the reference pascal pcb. Nobody should ever by any pascal gpu that uses a reference pcb (regardless of cooling).

This - amusingly - is why the pascal titan has virtually no headroom left. Then again boost 3.0 is the latest step in Nvidia trying to lock out manual overclocking.

>It isn't on custom cards with higher power limits.
the pic in OP is a custom card with higher power

You sound like a moron

I think you have trouble understanding that graph.

Talking down to someone when they're asking a question is fucking retarded. It's even more retarded when done on an anonymous message board.

Next time just answer the question, dickhead.

Pascal drops the clock rate to be power efficient when it has a light load. It's retarded.

>Light load
op's pic is from 100% load benchmark.

Can someone explain why my GPU throttles past a certain point even though the temps are fine?

Power draw target.

Never happens with my card, unless HWinfo etc doesnt upgrade the figures fast enough to show this. If I run benchmarks the only thing that might happen is throttling by 14Mhz from 2114 to 2100 core at like 65+ degrees.

What benchmark is that? I have 2 1080s and they never drop to ~1.5GHz unless the load drops, so I can only assume that it's not some game, but some sort of bench loop with the card dropping clocks in between runs.

Possibly due to its power draw target, but temp can be a limit too. Pascal will lower boost before it hits the actual temp limit. Basically, say you have a temp limit of 90C, a Pascal GPU will run at a higher clock at 50C than it does at 80C, at the exact same settings in terms of core clock, power limit and voltage.

they do drop by a lot, but only for a fraction of a second. it's a known pascal issue