Deaths of many GTX570s

[W1zzard]

Power throttling is a necessity. It's a manifestation of evolution in GPU designs.

look at gpu power consumption over the last years ... there is no significant increase

 

[streetfighter 2]

But I wonder if power density in the VRMs has increased then? My intuition says no.
So the quality of the VRM components has decreased? My intuition says no on this as well.
Perhaps then the VRM components are actually higher quality and provide better sourcing, but are more fragile at higher wattages? My intuition says, "tired... need alcohol".

Then whence cometh fire?

 

[robal]

look at gpu power consumption over the last years ... there is no significant increase

I agree.
What I meant is that a design evolution to have a chip that is capable (shader count, etc), but particular SKU enforces limits based on... well.. SKU spec., VRM design, cooling => price
...rather than a chip that is designed not to be able to exceed desired TDP.

It's a bit like having a car with speed limiter, instead of car with engine that won't make the car go past desired max speed.

 

[newtekie1]

Well you could also say the 6950 overclocks better with the voltage above stock, something people may have a good reason to fear doing on a 570. As much as a prefer nvidia I wouldn't buy or even keep a gifted 570, there's plenty of options out there that you don't have to worry about the durability on.

I disagree, besides the fact that the GTX570 at stock is a fair bit better than the HD6950, you can also overclock and even overvolt the GTX570 without fear as long as it is done within reason.

A few cards dying when run under extreme conditions isn't cuase to be alarmed quite yet.

 

[erixx]

Furmark? hehe Just don't leave it running and go walking the dog!!! Watch those crazy temps get higher and higher and higher

 

[dj-electric]

Seriously, people need to stop using Furmark.

 

[OneMoar]

depending the on level of scorching and you're ability with a soldering iron you might be able to repair the damage

 

[Gzero]

I agree.
What I meant is that a design evolution to have a chip that is capable (shader count, etc), but particular SKU enforces limits based on... well.. SKU spec., VRM design, cooling => price
...rather than a chip that is designed not to be able to exceed desired TDP.

It's a bit like having a car with speed limiter, instead of car with engine that won't make the car go past desired max speed.

Err your analogy is gibberish. Speed limiter set to x, and a car that only goes to x speed are not bothered by what desired y is because the modding needed is completely different to make the car go faster in comparison to overclocking.

The power throttling is simply there as a handicap to get around the current technical limitations of shrinking die sizes.

 

[Gzero]

I disagree, besides the fact that the GTX570 at stock is a fair bit better than the HD6950, you can also overclock and even overvolt the GTX570 without fear as long as it is done within reason.

A few cards dying when run under extreme conditions isn't cuase to be alarmed quite yet.

"The sky is falling, Starcraft 2 burnt my Nvidia card!"

Shame crysis's menu wasn't frame limited.

 

[yogurt_21]

Yeah, and it seems most of the people having problems are over-volting the cards to at least 1.1v, some further than that. I don't think I'd push them past 1.05v, or 1.08v like the max on the GTX400 cards.

yup extreme voltage needs extreme cooling which should be a common sense. and people should know by now that more than the core needs to be cooled. trouble is by the time you add in an extreme cooling method to the price of the card you might as well have gotten a 580.

 

[OneMoar]

Seriously, people need to stop using Furmark.

./agree
the only thing fumark does is melt stuff
the gpu's are not made to handle 100% load sustained

 

[erocker]

The power throttling is simply there as a handicap to get around the current technical limitations of shrinking die sizes.

Or, it could be there because the power componentry isn't sufficient enought to provide the rest of the components with power on a higher voltage setting. This definitely seems to be the case since we are clearly seeing vregs blowing up. The GPU can handle it, the vregs cannot.

correct why add "extra VRM" and increase cost when you can lock the power usage down and
sell it at a lower price point

Exactly.

 

[OneMoar]

I agree.
What I meant is that a design evolution to have a chip that is capable (shader count, etc), but particular SKU enforces limits based on... well.. SKU spec., VRM design, cooling => price
...rather than a chip that is designed not to be able to exceed desired TDP.

It's a bit like having a car with speed limiter, instead of car with engine that won't make the car go past desired max speed.

correct why add "extra VRM" and increase cost when you can lock the power usage down and
sell it at a lower price point
if there was ever a case that could be used to prove why they spec a MAX TDP then this is it
I suppose if you where 31337 enough you could mod the card with extra mofsets and maby a extra pcie plug should't be to .. terribly hard
personally I think the power override should be removed from GPU-Z NV already blocked it once maby they put that in for a reason (hint hint)
after looking at pics of the mofsets on some of these baked cards I am starting to see a pattern
looks like the first phase is going black first causing a cascade effect burning up the rest of them

 

[Easy Rhino]

seems to me furmark is the problem not the 570s. after all, it is possible to create software that when executed completely destroys your hardware. so would you blame the hardware or would you blame the software...

 

[erocker]

seems to me furmark is the problem not the 570s. after all, it is possible to create software that when executed completely destroys your hardware. so would you blame the hardware or would you blame the software...

If furmark was the problem wouldn't it cook all cards? Nvidia cheaped out on the power delivery, that's why they are failing. They had to do it to keep the cards competetively priced. Big GPU's cost money.

 

[Easy Rhino]

If furmark was the problem wouldn't it cook all cards? Nvidia cheaped out on the power delivery, that's why they are failing. They had to do it to keep the cards competetively priced. Big GPU's cost money.

not necessarily. and i would hardly blame nvidia. if you overvolt your card you are taking huge risks. furmark is a program designed to stress your card so you know when enough is enough. clearly the 570 is a fine card when you don't overvolt it too much.

 

[overclocking101]

wow that looks like mosfet is making contact to the heatsink and causing it to short. one guy said his died at 1.05v whts the stock 570 voltage?

 

[LAN_deRf_HA]

If nvidia doesn't want people to cook their cards then they themselves need to be providing the be all end all of stability testing apps. If you overclock at all you need to test stability. Doing it game by game is absurd. I had a instability on my 260 that wouldn't show up until about 4 hours of UT3. Until they make such a program, they should be supplying sufficient power circuitry or locking the cards from all overclocking.

 

[newtekie1]

seems to me furmark is the problem not the 570s. after all, it is possible to create software that when executed completely destroys your hardware. so would you blame the hardware or would you blame the software...

Where do you get that furmark is the problem? Cards have died doing simple 3DMark runs as well, which means they probably would have died in normal gaming situations.

 

[Easy Rhino]

If nvidia doesn't want people to cook their cards then they themselves need to be providing the be all end all of stability testing apps. If you overclock at all you need to test stability. Doing it game by game is absurd. I had a instability on my 260 that wouldn't show up until about 4 hours of UT3. Until they make such a program, they should be supplying sufficient power circuitry or locking the cards from all overclocking.

nvidia doesnt want people to cook their cards. that is why they express the dangers of doing so. it is at your own risk.

Where do you get that furmark is the problem? Cards have died doing simple 3DMark runs as well, which means they probably would have died in normal gaming situations.

i was told only cards were failing when overvolted and tested running furmark

 

[newtekie1]

i was told only cards were failing when overvolted and tested running furmark

And who told you this, or better yet why can't you just read the thread that was linked to in the first post that started all of this? On the first page there are at least two people that had cards fail while running 3Dmark. In fact the very first poster, the person that brought up the topic to begin with, hard the card die running Vantage.

 

[micropage7]

wow that looks like mosfet is making contact to the heatsink and causing it to short. one guy said his died at 1.05v whts the stock 570 voltage?

maybe its not like that, it looks the burnt from the under of chip then flow upward then leave burned track , its (maybe) caused by high voltage, even the cooling is better i guess the chip will be the same

 

[Easy Rhino]

And who told you this, or better yet why can't you just read the thread that was linked to in the first post that started all of this? On the first page there are at least two people that had cards fail while running 3Dmark.

u mad? i read the link. 2 people! my god, do a recall immediately! my point was if it was only happening with furmark then it was probably furmark. now that i know it is happening with other software then it is probably that some of the 570s dont overvolt well. big deal.

 

[LAN_deRf_HA]

If wiz had a card die with throttling doesn't that add to the notion that the vrm isn't sufficient? I could imagine this killing weaker cards after a year of dicking around in crysis. That seems to be what ended up killing my 260, just gaming wear and tear. It'd be interesting if we had a way of monitoring rma numbers after crysis 2 hits, as the multilayer might actually hold people's attention this time.

 

[phanbuey]

And who told you this, or better yet why can't you just read the thread that was linked to in the first post that started all of this? On the first page there are at least two people that had cards fail while running 3Dmark. In fact the very first poster, the person that brought up the topic to begin with, hard the card die running Vantage.

Ok yeah but those cards were REALLY overclocked... they were gonna pop sooner or later...

Nvm... picture didnt make sense....