FurMark GPU Killer?

[Guest]

So I got a PM from someone a while ago saying that the stress testing program/benchmark program FurMark can actually kill GPUs and that sounded pretty strange to me and I was just wondering if that was true or not. If so, please source. 

[themctipers]

It can because it stresses the GPU way, way more than normal usage (like prime 95 for CPUs, post AVX update)

[Guest]

3 hours ago, themctipers said:

It can because it stresses the GPU way, way more than normal usage (like prime 95 for CPUs)

Is Prime95 a CPU killer? 

[Mr.Meerkat]

http://wccftech.com/nvidia-gtx-1080-evga-catches-fire-video/

 

I know it's wccftech as the source of choice but everyone knows the EVGA 1080 VRM issue with furmark israel  

[themctipers]

Just now, IAmLamp said:

Is Prime95 a CPU killer? 

Well, AVX with P95 makes the CPU go up insane temps. You can try it  but keep looking at the temps 

[Aphexxis]

The GTX 480 egg baking application, you can cook sausages almost if you run it with 8xMSAA.

[8-Bit Ninja]

Just now, IAmLamp said:

Is Prime95 a CPU killer? 

No, as i said in the other one unless you do it for days its fine. The reason they exist is that they give you the worst possible result for temps, so you can know hand on heart that your cpu + gpu temps will always be safe 

Just now, themctipers said:

It can because it stresses the GPU way, way more than normal usage (like prime 95 for CPUs, post AVX update)

Yes, but unless your stressing the gpu for days without stopping its fine

[Guest]

3 hours ago, 8-Bit Ninja said:

No, as i said in the other one unless you do it for days its fine.

???????

[Guest]

3 hours ago, Mr.Meerkat said:

http://wccftech.com/nvidia-gtx-1080-evga-catches-fire-video/

 

I know it's wccftech as the source of choice but everyone knows the EVGA 1080 VRM issue with furmark israel  

How is that relevant? Isn't that just the guy turning on his PC? Where does it say he's running FurMark? 

[Guest]

3 hours ago, themctipers said:

It can because it stresses the GPU way, way more than normal usage (like prime 95 for CPUs, post AVX update)

Any other stress testing/benchmarking programs/applications that can kill GPUs? I want to stay away from them and make notes of which ones they are. 

[manikyath]

furmark is designed to put a beyond realistic amount of stress on your GPU, and it makes potential hidden issues come up (like the EVGA issue), but if a GPU is all happy, it wont magically kill them.

 

its like P95 wont kill a happy cpu, but a lot of people *think* their cpu is happy at a certain OC, and then P95 comes in and wrecks stuff

[Mr.Meerkat]

7 minutes ago, IAmLamp said:

How is that relevant? Isn't that just the guy turning on his PC? Where does it say he's running FurMark? 

Ah fack, wrong link  

Ah well...

[Guest]

3 hours ago, manikyath said:

furmark is designed to put a beyond realistic amount of stress on your GPU, and it makes potential hidden issues come up (like the EVGA issue), but if a GPU is all happy, it wont magically kill them.

 

its like P95 wont kill a happy cpu, but a lot of people *think* their cpu is happy at a certain OC, and then P95 comes in and wrecks stuff

So Prime95 is a potential CPU killer? 

[manikyath]

2 minutes ago, IAmLamp said:

So Prime95 is a potential CPU killer? 

if your cpu is potentially not running quite that happy already.

[Drak3]

9 minutes ago, IAmLamp said:

So Prime95 is a potential CPU killer? 

Only if it's an old CPU ( > 10 years at least), or malfunctional. There are safeguards that will force system shutdown or force system crash before before Prime95 could kill a CPU. You would actually have to try to kill a CPU with Prime95, with quite an extensive effort to do so, or put a high power draw CPU on a a 3 phase VRM mainboard (9590 anyone?).

[Guest]

3 hours ago, Drak3 said:

Only if it's an old CPU ( > 10 years at least), or malfunctional. There are safeguards that will force system shutdown or force system crash before before Prime95 could kill a CPU. You would actually have to try to kill a CPU with Prime95, with quite an extensive effort to do so, or put a high power draw CPU on a a 3 phase VRM mainboard (9590 anyone?).

Could you give an example of some of those safeguards? 

[Drak3]

Just now, IAmLamp said:

Could you give an example of some of those safeguards? 

The first is one we all should be familiar with: Thermal throttling. Lowering the speed until heat can be dissipated fast enough that a CPU won't be damaged. Some chips can actually clock to below 1GHz, and run like that with just an empty water block as a heat sink.

 

Then you've got the BIOS and the microcode on the CPU that will force system crash/shutdown if unsafe temperatures are maintained for a set duration (a matter of minutes, but I don't know how long exactly).

[Guest]

3 hours ago, Drak3 said:

The first is one we all should be familiar with: Thermal throttling. Lowering the speed until heat can be dissipated fast enough that a CPU won't be damaged. Some chips can actually clock to below 1GHz, and run like that with just an empty water block as a heat sink.

 

Then you've got the BIOS and the microcode on the CPU that will force system crash/shutdown if unsafe temperatures are maintained for a set duration (a matter of minutes, but I don't know how long exactly).

Yeah I'm confused as to how the components could be killed then because if I'm not mistaken, a system will also not boot if temperatures are not safe enough. So it sounds like it would have to be some pretty intense testing. 

[Drak3]

Just now, IAmLamp said:

Yeah I'm confused as to how the components could be killed then because if I'm not mistaken, a system will also not boot if temperatures are not safe enough. So it sounds like it would have to be some pretty intense testing. 

Components can be killed via overvoltage, and this is often the most common way that they die. Some components have no or inadequate failsafes (some boards will allow you to pump 1.5v to a CPU, but 1.5v would kill any compatible CPU), and others can only protect from user ignorance/stupidity.

 

However, come components (generally cheap mainboards) don't monitor VRM temperatures, or include poor safeguards if they do. This was the case with the EXGA GTX 10 cards with ACX 3.0(?), where VRM temperatures went unchecked and weren't dealt with properly in the beginning. It was possible to kill the card (not necessarily the GPU itself), but that was due to poor design on EVGA's part. Remember, what I said about Prime95 only applies to the CPU, not the VRMs feeding it with power.

[Aphexxis]

3 minutes ago, Drak3 said:

Components can be killed via overvoltage, and this is often the most common way that they die. Some components have no or inadequate failsafes (some boards will allow you to pump 1.5v to a CPU, but 1.5v would kill any compatible CPU), and others can only protect from user ignorance/stupidity.

 

However, come components (generally cheap mainboards) don't monitor VRM temperatures, or include poor safeguards if they do. This was the case with the EXGA GTX 10 cards with ACX 3.0(?), where VRM temperatures went unchecked and weren't dealt with properly in the beginning. It was possible to kill the card (not necessarily the GPU itself), but that was due to poor design on EVGA's part. Remember, what I said about Prime95 only applies to the CPU, not the VRMs feeding it with power.

1.5v through a Phenom II is nothing, Phenom II will take that all day long.

 

x58 CPU's on 45/32NM will handle 1.5v so long as you have insane cooling.

[Drak3]

Just now, Aphexxis said:

1.5v through a Phenom II is nothing, Phenom II will take that all day long.

 

x58 CPU's on 45/32NM will handle 1.5v so long as you have insane cooling.

Technically speaking, I could pump 1.8v through Ryzen on LN2. But that isn't the point. The point is that one can have a mainboard that will allow a user to pump voltage that would kill a CPU under any consumer scenario, without addressing it at all. My board will technically let me push 2v, but it'll take quite a bit of effort as it won't push above 1.45v without reaching the dark nether regions of the board that joe average won't go.

Those records you see, where people get 5.5+ GHz at voltages like 1.8v, take quite a bit of work to get the board to feed that voltage, if they're using a decent one that can handle it.

[Aphexxis]

Just now, Drak3 said:

Technically speaking, I could pump 1.8v through Ryzen on LN2. But that isn't the point. The point is that one can have a mainboard that will allow a user to pump voltage that would kill a CPU under any consumer scenario, without addressing it at all. My board will technically let me push 2v, but it'll take quite a bit of effort as it won't push above 1.45v without reaching the dark nether regions of the board that joe average won't go.

Those records you see, where people get 5.5+ GHz at voltages like 1.8v, take quite a bit of work to get the board to feed that voltage, if they're using a decent one that can handle it.

Phenom II can handle 1.55V as a maximum on high end air cooling lol.

 

First gen Core i7's will handle 1.4-1.5v, 1.5v usually on water.

[Drak3]

Just now, Aphexxis said:

Phenom II can handle 1.55V as a maximum on high end air cooling lol.

 

First gen Core i7's will handle 1.4-1.5v, 1.5v usually on water.

No inclination to believe you, or trust that those CPUs aren't being killed off slowly.

[Aphexxis]

2 minutes ago, Drak3 said:

No inclination to believe you, or trust that those CPUs aren't being killed off slowly.

Ran a Phenom II at 4.15ghz 1.49v on a Hyper 212 for 6 years on a 790FX motherboard.

 

My CPU right now the Xeon 32nm Westmere is sat at 4.2ghz on 1.32v and only hit's 62c on maximum load on a small air cooler, i have way more voltage headroom.

 

You seem to forget that the shrinkage of later CPU's drops voltage handling.

[Drak3]

3 minutes ago, Aphexxis said:

Ran a Phenom II at 1.49v on a Hyper 212 for 6 years on a 790FX motherboard.

 

My CPU right now the Xeon 32nm Westmere is sat at 4.2ghz on 1.32v and only hit's 62c on maximum load on a small air cooler, i have way more voltage headroom.

Again, no inclination to believe you. Especially as the second scenario isn't exactly unique, that voltage can be maintained on a 14nm 7700K. Also, 62C on air tells me you haven't stressed that CPU much, especially as my 5930K hits 80C @ 4GHz/1.19v with a 12 thread Prime95 load, under a 240mm rad running 2200RPM Static Pressure fans that are running nearly 100%.

 

7 minutes ago, Aphexxis said:

You seem to forget that the shrinkage of later CPU's drops voltage handling.

Not drastically, and not to the point that 1.5v would be considered safe on air, or lower end water cooling, for any period of time over a meager few hours, from a thermal aspect.