SweClockers: Geforce GTX 590 burns @ 772MHz & 1.025V

[pcm81]

I disagree, 2 GTX580 will cool a LOT better than a GTX 590 ever could. If you give it ridiculous VRMs and have it pull 600 watts for that single card it will be limited by heat.
In fact, the card is already hitting max temps (90C) on a "mere" 375 watts.


Fair enough. I agree that this would be a valid use for it. But what I said was why use GTX590 vs 2x GTX580 or 2x GTX570. Not why use 2xGTX590 vs 3x GTX580 or 3x GTX570

That's why I said it will need water cooling at decent clocks...

I run my OS off of RevoDrive, and I would like to add a 2nd revodrive but dont have slots for it, so there is always need for more slots, even on high end MOBOs

If a user has avaiable slots and does not need more than 2 GPUs, there is no reason to use 590 over 2x580s other than cost, unless it is priced the same...

 

[taltamir]

That's why I said it will need water cooling at decent clocks...

I thought you only said VRMs, power draw, and clocks. But still, the problem is that physics (the laws of nature, not physX ) means that "proper cooling" is impossible.

Dissipating 600 watts on a single card is a ridiculous task. The GTX590 altready has a smaller back vent, and one GPU vents into the case while the other out the back vent. That is just going from 300 to 375 watt card.

Having two seperate cards of 300 watt each is so much more practical to cool than a single 600 watt card.

I do not think it is plausible to do without exotic (and very expensive) cooling.

I run my OS off of RevoDrive, and I would like to add a 2nd revodrive but dont have slots for it, so there is always need for more slots, even on high end MOBOs

If a user has avaiable slots and does not need more than 2 GPUs, there is no reason to use 590 over 2x580s other than cost, unless it is priced the same...

Now that is extreme right there .
Fair enough assertions.

 

[AtenRa]

You are thinking that voltage is what is killing the VRMs, bit it's not. It's the current that's killing them. By reducing the clock speed of the GPU you reduce the amount of current it draws.

That’s correct but I believe that lowering only 50MHz and keeping the same voltage will not decrease the current more than 0.1-0.3 Amps per 12V line.

Look how the current is much more dependable to the Voltage and not to the Frequency.

http://www.xbitlabs.com/articles/video/display/gpu-power-consumption-2010_4.html

One more thing,
All the pictures of the blow up cards (Default settings) don’t show a VRM apparatus (Voltage Controller, MOSFET, Inductor) to have failed except W1zzards photo and that’s with 1.2V.

I still believe that the VRM implementation is fine for default usage and the blow ups happened from a different cause and not from the VRMs.

 

[pcm81]

@AtenRa
Higher clocks do not draw more current. But the chip becomes unstable, because there is not enough time for the gates to react (fill with electrons) at what ever voltage was originally. As the result to establish stability of the chip at higher clocks one must increase the voltage, which increases the current as well.

@taltamir
I never said 590 is a good thermal design, in fact it is pushing the limits even at lower clocks. I strongly believe that "properly designed" gtx590 will REQUIRE water cooling. 2x580 on the other hand can survive on air because of 2x lower power density.

 

[Phynaz]

Higher clocks do not draw more current.

Yes, they do.

IDC has even posted the formula to calculate the additional power draw.

 

[AtenRa]

@ taltamir

Yes the GTX570 SLI is cheaper than GTX590, just show it on newegg
I believe GTX570 also throttle down in futuremark, any way noise is close between the three.

@pcm81

Yes I was talking about the rumor with the 50MHz decrease in frequency.

 

[AtenRa]

Yes, they do.

IDC has even posted the formula to calculate the additional power draw.

Can you post that plz

 

[Phynaz]

http://forums.anandtech.com/showpost.php?p=31477365&postcount=191

 

[AtenRa]

thx

 

[JaussS]

Some people say it's safe to buy as long as you run it at stock speeds and use the latest available drivers from nvidia.

If it was me I'd wait it out awhile longer and see what developes.

so many crazy pc people everyone talk about card 590 only reply. i read xtremesystems many people scary of gtx 590. £600 too much to poker with i think thank you. think 580 gtx or maybe try 6990 £500 so is less but i think just get 580 gtx and be safe. gtx 590 scary me. thanks

 

[Keysplayr]

Higher clocks do not draw more current.

It certainly does.

 

[Idontcare]

Can you post that plz

Phynaz posted the link you requested so I won't go to the effort of repeating it here but suffice it to say power consumption scales with operating frequency.

There is static power-consumption (leakage when at idle) and dynamic power consumption (scales linearly with frequency).

Increasing clockspeed does increase the power consumption as it increases the contributions of dynamic power consumption even when holding all else constant (voltage, temps, etc).

 

[pcm81]

It certainly does.

You right, long day at work...

 

[AtenRa]

Phynaz posted the link you requested so I won't go to the effort of repeating it here but suffice it to say power consumption scales with operating frequency.

There is static power-consumption (leakage when at idle) and dynamic power consumption (scales linearly with frequency).

Increasing clockspeed does increase the power consumption as it increases the contributions of dynamic power consumption even when holding all else constant (voltage, temps, etc).

Well I didn’t say power doesn’t scale with frequency, I was trying to say that 50MHz less will not make the VRMs work a lot easier without lowering the Voltage.

 

[Silverforce11]

If clock speed affects power use and current, if you drop clock speed by 10%, you don't think it will affect the VRMs? Come on dude. Logic.

 

[Idontcare]

Well I didn’t say power doesn’t scale with frequency, I was trying to say that 50MHz less will not make the VRMs work a lot easier without lowering the Voltage.

It does decrease the amps though (P=IV).

There is a reason the VRM's are spec'ed by the amperage they can reliably operate at, right?

Too many amps is bad, less amps is better, lower and lower amps means longer and longer operating lifetime for the component.

This is probably universally true of all electrical components owing to the fact that fundamentally the degradation mechanisms involve the same basic factors - electric field gradient, thermal energy, electrons, and atoms.

 

[Keysplayr]

If clock speed affects power use and current, if you drop clock speed by 10%, you don't think it will affect the VRMs? Come on dude. Logic.

Clock speed, whether increased or reduced, will either increase or reduce power draw. And I would think that means more, or less, stress on the power circuitry. It makes sense.

 

[wahdangun]

It does decrease the amps though (P=IV).

There is a reason the VRM's are spec'ed by the amperage they can reliably operate at, right?

Too many amps is bad, less amps is better, lower and lower amps means longer and longer operating lifetime for the component.

This is probably universally true of all electrical components owing to the fact that fundamentally the degradation mechanisms involve the same basic factors - electric field gradient, thermal energy, electrons, and atoms.

so its mean gtx 590 VRM degrade faster since its pushing its limit ?

 

[Silverforce11]

As with everything, components aren't all the same. Some are weaker than others. Most would handle beyond the "rated specs" fine, while some won't and die. Essentially by running at the limit you expose the flawed ones relatively quickly, IMO.

 

[Idontcare]

so its mean gtx 590 VRM degrade faster since its pushing its limit ?

This goes without saying, right?

Your car probably redlines around 6000rpm, give or take 1000rpm.

Your car's engine will wear out in time regardless whether you keep the rpm's below 1000 its entire life, or if you run it at 4000 rpms its entire life.

The only thing you change is the time-to-death.

Run it at redline and time-to-death will come sooner than if you ran it at 50% redline until it died.

Very few things, mechanical or solid-state, in life are not like this. Its a universally true wear-and-tear fact of life.

VRM's are no different in this regard. The only question is "how long will it last when ran that hard?"

Does it matter how hard the VRM's are being pushed if they are still expected to last 10yrs even on a 590? Probably doesn't matter in that case.

 

[pcm81]

This goes without saying, right?

Your car probably redlines around 6000rpm, give or take 1000rpm.

Your car's engine will wear out in time regardless whether you keep the rpm's below 1000 its entire life, or if you run it at 4000 rpms its entire life.

The only thing you change is the time-to-death.

Run it at redline and time-to-death will come sooner than if you ran it at 50% redline until it died.

Very few things, mechanical or solid-state, in life are not like this. Its a universally true wear-and-tear fact of life.

VRM's are no different in this regard. The only question is "how long will it last when ran that hard?"

Does it matter how hard the VRM's are being pushed if they are still expected to last 10yrs even on a 590? Probably doesn't matter in that case.

Looks like based on currently available data of failed VRMs we can conclude that they will last on average a week? As more fail later rather than sooner this number will gradually grow to a month or higher. The VRMs which have not failed so far are not part of the data set until they do fail.

EDIT: Any 1 know exact model number of VRMs used on 590s? The manufacturers spec sheet should include the SN curve for the part at misc voltage/current levels.

 

[wahdangun]

This goes without saying, right?

Your car probably redlines around 6000rpm, give or take 1000rpm.

Your car's engine will wear out in time regardless whether you keep the rpm's below 1000 its entire life, or if you run it at 4000 rpms its entire life.

The only thing you change is the time-to-death.

Run it at redline and time-to-death will come sooner than if you ran it at 50% redline until it died.

Very few things, mechanical or solid-state, in life are not like this. Its a universally true wear-and-tear fact of life.

VRM's are no different in this regard. The only question is "how long will it last when ran that hard?"

Does it matter how hard the VRM's are being pushed if they are still expected to last 10yrs even on a 590? Probably doesn't matter in that case.

so is there a formula to measure time-to-death ? It maybe handy, because you will know the limit, when overclocking

 

[Seero]

Looks like based on currently available data of failed VRMs we can conclude that they will last on average a week? As more fail later rather than sooner this number will gradually grow to a month or higher. The VRMs which have not failed so far are not part of the data set until they do fail.

EDIT: Any 1 know exact model number of VRMs used on 590s? The manufacturers spec sheet should include the SN curve for the part at misc voltage/current levels.

If i flip a coin 10 times and they all ended up heads, can I conclude that is only head on the coin?

If i roll a 6 face dice 10 times and all outcome is greater than 5, can I conclude that the chance of getting 5 or 6 is far greater than any other number?

It seems you concluded that VRM is to be blamed for the failure, but not to the improper OC. Why?

It seems that you don't believe that running something under spec is far safer than running something off the spec. 375 Watt is a magic number for a single slot pci express card. It can draw up to 75 Watt from the PCIe slot, 150Watt from each 8 pin connector. What does it mean?

What will happen if you are to draw electricity off a power line? Will anything heats up? That is, the unit which either converts AC to DC, DC to DC and the power line/wire/pin itself. What will happen if you are to draw electricity beyond its limit? Will anything heats up beyond specification?

I know you like better VRM, but have you consider the fact that the parts that supply electricity to those VRMs ain't designed to push out > 375 Watt in total worth of electricity? What does that mean?

With better VRM, the video card may be protected, but not the rest of your system if you push it beyond 375 Watt! Adding 1 more connector, be it 6 pin or 8 pin allows > 375 Watt output, but it does not pass the pci express standard. What does that mean?

 

[taltamir]

If i flip a coin 10 times and they all ended up heads, can I conclude that is only head on the coin?

is it one of these coins? http://www.google.com/search?hl=en&...q=double+sided+coin&spell=1&biw=1920&bih=1075

 

[AtenRa]

Looks like based on currently available data of failed VRMs we can conclude that they will last on average a week? As more fail later rather than sooner this number will gradually grow to a month or higher. The VRMs which have not failed so far are not part of the data set until they do fail.

EDIT: Any 1 know exact model number of VRMs used on 590s? The manufacturers spec sheet should include the SN curve for the part at misc voltage/current levels.

Voltage Controller (PWM) = CHil CHL8266
http://www.chilsemi.com/products/multiphase-digital-power-controllers/chl8266/

Inductor = A101 1050F no specs found

MOSFET (DrMos) = Infineon TDA21211
http://www.infineon.com/dgdl/PB-TDA...e6cb4&fileId=db3a304323b87bc201240b3c2da847b1

Photos
http://translate.google.gr/translat...dia-geforce-gtx-590-studiu-de-overclocking/12