Overclocking - can you klil your card

A

a3venom

Gawd
Joined
Mar 3, 2015
Messages
695
so i have a ref r9 290, i keep messing around with voltages upto +300mv sometimes, clocks not so much but memory 1600mhz or so just out of curiousity.

Normally i run my card at stock to +19mv on 1000/1500, but i sometimes like to benchmark with 100% fan speed with retarded clocks and voltage.
i have done benchmarks with artifacts (started artifacting in the start but i still let it finish) to see how clock speed affects.
I mostly do unigine benches.

My question is, can this break something? i've made it crash over a 100 times due to too high mem on too low volts - pretty sure.

or can overclocking / messing with volts break cards at all? aren't there limits imposed so that no matter how much u play with your clocks, card will just shut down and be ok.

EDIT : i also have flashed different bioses 100+ times.
 
Last edited:
1. Don't artifact. Artifacting is bad. Artifacting may not kill the card, but it definitely means it's been pushed past it's limits.
2. Don't push so much voltage +300 is way too much.
3. 1100/1600 is a good OC. You may not need any extra voltage to get the mem to 1600.
 
yeah i usually keep it undervolted even, because i dont wanna go above 50% fan speed, ref 290 cooler sucks but i got this for like $160 US($200 CAD) so it was hard to pass on last year.

but my curiosity is, if there is no way you can kill your card by using a software, why are people so afraid to OC or ask others to be careful, increase in steps etc..

this card has to be 2 years old btw, isnt it a machine that will take all the beating you throw at it and continue to run as its supposed to at stock no matter how much u screw with it?
 
In an ideal world, yes, things are supposed to shut down or throttle if pushed too hard. But that doesn't always happen how it's supposed to, and when you play with fire you're gonna get burned (or in this case, neat artifacts you can't get rid of without RMA)...

Best approach to GPU OC, just like with overclocking anything else, is to push in small increments and test after each modification. Very time consuming, but less so than waiting for a new card because yours has been shipped to the manufacturer :p
 
I pushed too much voltage with my 7970 for a very long extended period of time. Replaced it with something better a year ago but just put it back in another systen and it artifacts like no other. There may have been other factors but the likely culprit is overvolting.
 
I am covered till next year i will get a r9 390 if it breaks.
i would love to break it, tbh - but i was just curious about the volt thing, the post above me is pretty clear it fucked his card in the long run
 
Its unlikely that just upping the clocks will break anything unless the card overheats, it just causes it to artifact and become unstable. But upping the volts to push the instability and artifacting threshold further does. It puts the chips under more stress and increases the temperature a lot. Every chip has the maximum voltage it can stand (which can vary from chip to chip batches) before the stress starts to degrade them fast and eventually cause it to fail. The stock voltage are usually way below that threshold but what the maximum exactly is is unknown because, as said, it varies from card to card. You can find some information on the internet and do an educated guess what the generally dangerous voltage range is on that GPU model. There are a lot of people who have a habit in pushing the cards too far and make them melt down. Stay way below them.

When overclocking first up the core clocks until you become unstable. Then increase the volts until the instability stops. There is no point increase the volts further since volts itself do not increase performance, just add more stress and heat. Repeat this process until you are satisfied, run out of guts or get too close to the supposed dangerous voltage levels. Then increase the memory clocks until you they become unstable and then drop them below that level. No need to touch volts for memory, it does nothing for them.
 
I roasted the memory on my HD 7950 by juicing it and running it at 1800+ speeds. Card still works, but artifacts constantly. So yes, you can mess things up. As soon as I see artifacts now I drop to a more reasonable OC.

With that said, I slapped a H80i GT and Corsair HG10 GPU bracket on my reference R9 290. Runs 24/7 @1125 / 1500 with the core hovering at 42c idle and 56c under extreme load. +200 mv on the core and 0 on the memory. Seems to be fine.
 
Consider what ages the card and how fast.
Heat and voltage both age silicon faster and can cause other components to fail as well.
Individually, the aging is exponential as they raise, both together is even worse!

+300mV is way too high, keep doing that and the GPU will expire.
That is the territory of supercooled cards.
Running high voltage at high temp rapidly accelerates wear.
Over +200mV sustained with the stock cooler is too high imo, you are lucky the card is still working.

Once the card hits max temp it will throttle.
Higher voltage will do nothing except cause damage.
 
+300mV, holy crap... You're lucky your card is still kicking at all.
 
The general idea with most people afraid OC graphics cards is that, they just want to make sure the card lasts as long as it can. They're usually people who buy a new build and use it for years and years until it can't do what they want anymore. At +300mv to the card, I am a firm believer that the only reason the card did not die on you, is that you at it locked to 100% fan speed, it was not overclocked so it may not have clocked itself higher to use more power, and it is just simply a well made card. You're lucky that you didn't damage your PSU or trip any of the protections given there are any..
 
I have never overclocked before owning this card, i treat it like shit because i have extended warranty.
I was surprised to see i can bench at 56 fps stock on valley to 64 fps oced with about 50-100mv added.
anything beyond that is very shitty gains for too much volts, but just exploring. i will probably not do it again.
 
cageymaru said:
I roasted the memory on my HD 7950 by juicing it and running it at 1800+ speeds. Card still works, but artifacts constantly. So yes, you can mess things up. As soon as I see artifacts now I drop to a more reasonable OC.

With that said, I slapped a H80i GT and Corsair HG10 GPU bracket on my reference R9 290. Runs 24/7 @1125 / 1500 with the core hovering at 42c idle and 56c under extreme load. +200 mv on the core and 0 on the memory. Seems to be fine.
Click to expand...

Sounds good but, what about the VRM's?
 
a3venom said:
I have never overclocked before owning this card, i treat it like shit because i have extended warranty.
I was surprised to see i can bench at 56 fps stock on valley to 64 fps oced with about 50-100mv added.
anything beyond that is very shitty gains for too much volts, but just exploring. i will probably not do it again.
Click to expand...

If you are going to continue doing so, I recommend a higher capacity power supply. Something at least of the 750 Watt variety with a high 12V line amperage capacity would be a very good idea. Many years ago, I used to not go with a good enough power supply and blew two of them up. (Thankfully, nothing else was damaged.)

Just as a heads up, the heatsink fan on the reference edition is fine. However, you need a quieter case and proper airflow to make use of it. (I usually run it at stock clocks though since I have seen little difference in my games.) I have had to up the fan speed to 60% during the summer but, my case is a quiet one so that helps a lot.
 
Killing graphics cards via overclocking these days is a lot harder compared to what it was back in 2001-2006.

Overvolting is a different story (but your BIOS has limitations to whatever amount you can set).
 
In 20+ years of overclocking CPUs and GPUs, I've never once killed a chip. And that's not because I'm ultra safe or an expert, it's because it's actually very difficult to damage a chip.

And nowadays with temp and throttling controls, unless you're doing something extremely wrong meddling with a custom bios, it's virtually impossible to kill a chip.

What you may do if you're futzing with voltages is reduce the lifespan of a chip. However, we're talking about the difference between 5 and 10 years in which time you will likely have changed your card/cpu out three times over.
 
cageymaru said:
The bracket covers them and the stock fan that comes with the video card is repurposed by Corsair's design to blow across the card and thus directly on them. The VRMs with the HG10 installed is the last thing I would worry about. :)

Video
https://www.youtube.com/watch?v=c97OgnuK5V8
Click to expand...

Good. I know that some of these aftermarket coolers ignore the VRM's and they are the primary reason our 290's get so hot. Glad it is working for you. :)
 
The only thing I don't like about it is that the way that the H80i GT is designed; the tubing comes out of the pump's top instead of the sides. So this makes my card into a 3 slot card. I didn't exactly intend on running CrossfireX, but it would be nice to be able to. The modder in me says that some 90 degree pipes should take care of it, but the H80i just came back from Corsair RMA last week. I'm just not into voiding warranties that fast. ;)
 
I wouldn't say 5-10 years. I've seen as little as 2-3 years depending on how you're volting the card. It's not so much the actual chip that goes bad but the solder joints from the heat/cool cycles that fails.
 
You must log in or register to reply here.
Back
Top