Warning on MSI cards?

[mehta23]

I just read this article on how MSI are secretly overvolting their cards ... which can reduce the life of it ...

http://translate.google.com/translate?sl=auto&tl=en&js=n&prev=_t&hl=en&ie=UTF-8&layout=2&eotf=1&u=http%3A%2F%2Fwww.tomshardware.de%2FMSI-GTX660Ti-Overvolting%2Ctestberichte-241108.html&act=url


I didnt really understand a lot of it , but I have 2 x MSI 660Ti, and it only comes with a 2 year warranty..

I'm thinking now - should I swap them for a Asus/Evga/Gigabyte 670 ?

 

[nforce4max]

Look at pics of the bare pcb without the cooler installed the build quality is much higher than normal models for the 660 ti. The only real concern would be if the core voltage was being pushed to high for the 28nm process. With voltage comes faster oxide gate degradation and reduced life on top of higher power consumption. Personally I would pull the face plate and swap out the stock fans for some high flow 80mm fans if they would fit. 60cfm and higher would keep the card cool even in a hot summer without ac

 

[monu_08]

dont overclock ur cards u are fine with them on stock one neither need an another cooler because gpu already overclock by factory msi

 

[mehta23]

I am pretty much a noob...But isn't the problem that there may be too much strain on the PWM ? I don't even know what that is.

 

[nforce4max]

The power vrm can take a lot of abuse but it has to be kept cool. The faster each mosfet has to work the hotter they get and as heat builds up their performance suffer. The worst part is life span if they remain at high temps for to long and they do have limits when it comes to how much current they can provide. If loaded to much they will burn out. I have seen a stock reference model 680 burn out this way. The mid plate is a lot better than there being nothing but if it had ribs or needles for better heat dissipation would be a huge improvement over a flat surface.

 

[monu_08]

i am not too much expert in hardware point of view but consider this will be help full for u http://en.wikipedia.org/wiki/Pulse-width_modulation
and u can not predict u r card how long that will work u just dont worry about it simple words in this world nothing can be remain permanent just chill and happy with 2x msi gtx 660 ti

 

[Maxx_Power]

I just read this article on how MSI are secretly overvolting their cards ... which can reduce the life of it ...

http://translate.google.com/translate?sl=auto&tl=en&js=n&prev=_t&hl=en&ie=UTF-8&layout=2&eotf=1&u=http%3A%2F%2Fwww.tomshardware.de%2FMSI-GTX660Ti-Overvolting%2Ctestberichte-241108.html&act=url


I didnt really understand a lot of it , but I have 2 x MSI 660Ti, and it only comes with a 2 year warranty..

I'm thinking now - should I swap them for a Asus/Evga/Gigabyte 670 ?

What they are doing that is potentially DANGEROUS is tinkering with the voltage reference values that is supplied to the PWM controller.

Essentially the PWM controller needs a reference voltage internally, against which the output is compared by using a negative feedback to change the duty cycle/frequency (usually just duty cycle). To raise the output voltage, you can reprogram the PWM controller if it supports software programming, or use a different reference voltage. What MSI has done that is despicable is that the voltage reference value for the PWM controller is jacked up to a whopping 8.3 volts under load conditions and startup surge. That chip as per manufacturer's specifications only has a recommended reference voltage of 5 volts, and an ABSOLUTE MAX upper limit of 7 volts. By feeding it 8.3 volts as a voltage reference, the PWM controller will be very shortly lived compared to its normal life span. No amount of cooling will compensate for this, since this reference voltage is not used to derive power, only comparison purposes, so jacking up the voltage does not increase internal heat generation significantly, but does shorten (I'm guessing the Vref is tied to a leg of a transistor or opamp internally) the lifespan of that part, in a non-thermal fashion. This startup surge also reacts with certain computers (mainly PSUs) in such a way as to prevent the card from properly POSTING on startup, and non-specific hard crashes when the GPU clock changes (gaming or otherwise).

To make matters worse, MSI did this in a sneaky way. Instead of tinkering with programming or using a different set of resistors to set the reference value at the TL431 regulator that feeds into the PWM controller, they used an inconspicuous capacitor to isolate the regulator from the ground, which it normally connects to. By isolating it from the ground, the regulator isn't really doing any regulation anymore, and essentially, the resistor divider just before the regulator completely sets the voltage. AND, from an engineering point of view, if you just casually examine the board, it LOOKS as if the capacitor is used to bypass the regulator OUTPUT, which is the NORM to stabilize the regulator. THat's why it is sneaky.

The article talks/speculates about the reasons why MSI (having a long history of successful engineering) would do something like this, and do it in the sneaky way. It has to do with Nvidia's TIGHT restrictions on Kepler vcore adjustments. And just this morning, Nvidia officially slammed the door on EVGA's custom voltage manipulations solution, see:

http://www.techpowerup.com/173110/NVIDIA-Forces-EVGA-to-Pull-EVBot-Support-from-GTX-680-Classified.html

 

[mehta23]

So , am I doing the right thing by returning my MSI 660Ti's?

 

[Maxx_Power]

So , am I doing the right thing by returning my MSI 660Ti's?

Double check the back of your card, to see if yours match the images posted on the German tomshardware site.

If yours looks exactly like this:

and zoomed in:

With that TINY capacitor next to the TL431 regulator, then YES. I would NOT hold onto, or recommend anyone holding onto a card that is designed with a part driven far, far out of spec. Even if it doesn't die, you can't be sure it will work in another computer with a different PSU (which is how this issue was discovered), so that there is practically NO resale value either because of this.

THis is just too bad, because I actually like MSI cards and their coolers.

 

[mehta23]

Damn! I have two , once can be returned , but the other needs to be sold - unopened...

What should I go for now?

 

[sarinaide]

Its the only way to get anything out of Nvidia's attitude to GPU's with Kepler, flawed architecture followed by almost non existent expansion, AMD threw the big breaking ball and while GCN evolves with drivers, Kepler is stagnant. Cue GK110, same architecture same flaws. nvidia don't want their cards clockable which is why most are volt limited where a mere .5 slide is enough to crash drivers. (avoid model name)

 

[ram1009]

Look at pics of the bare pcb without the cooler installed the build quality is much higher than normal models for the 660 ti. The only real concern would be if the core voltage was being pushed to high for the 28nm process. With voltage comes faster oxide gate degradation and reduced life on top of higher power consumption. Personally I would pull the face plate and swap out the stock fans for some high flow 80mm fans if they would fit. 60cfm and higher would keep the card cool even in a hot summer without ac

The relative "build quality" is NOT discernible from any pics I saw and isn't even an issue here. The quality of the components or workmanship are not at issue. They are clearly over-driving the chip by very significant amounts. I have one of these boards and am waiting to see if this gets resolved.

 

[ram1009]

Here is an e-mail I received from MSI tech support in response to my telephone inquiry about the over volted GPU:

Thank you for contacting MSI support, In reference to your inquiry about the N660Ti PE 2GD5/OC, MSI respects the result which is tested by Tom’s Hardware, but we have much confidence and believe our products would not cause any customer RMA concerns. Since MSI designs these custom products with overclocking in mind, we “supercharge” these cards because we’re anticipating enthusiast to overclock. Because of this design decision and the higher component quality, we’re able to provide more power to the board resulting in higher and longer GPU Boost operation without reducing the lifetime of the graphics card or warranty term. MSI’s all graphics cards including GTX 670 and GTX 660 Ti passed strict test and stand behind our products with a 3 year warranty.


Make of it what you will. I'm also told a supervisor will be calling me back today. I won't hold my breath.

EDIT: The supervisor never called.

 

[Maxx_Power]

Damn! I have two , once can be returned , but the other needs to be sold - unopened...

What should I go for now?

I can't say for SURE, but I am certain that MSI can arrange an exchange. OR, it is actually an easy DIY repair to solder bypass that pesky capacitor they put in. This is described on that article (need to translate it first).

Since you have warranty still, I'd go with the first route. See if you can talk to the store where you purchased it, and exchange it for a similar model, OR send an email to MSI. I can't imagine that MSI would say NO, because this is damaging enough to their reputation as it is. They sort of alluded to this in their response saying that (paraphrasing) the user can always make use of the warranty during the warranty period.

I saw that comment you posted just now from MSI. I wouldn't trust their "supercharging". To properly test a component part for accelerated testing, MSI would need months and months to see what the failure rates are of the PWM chip when driven far out of spec to understand what the failure rates will be in current situations.

YOU CAN always have it RMA'd on the account that these cards are prone to crashes and no-POST issues due to incompatibility issues with certain boards and PSUs. See the original article on German Tomshardware site for this, and if you google it, it is a fairly rampant problem.

Original article translated:

http://translate.google.com/translate?sl=auto&tl=en&js=n&prev=_t&hl=en&ie=UTF-8&layout=2&eotf=1&u=http%3A%2F%2Fwww.tomshardware.de%2FMSI-GTX660Ti-Overvolting%2Ctestberichte-241108.html&act=url

 

[nforce4max]

The relative "build quality" is NOT discernible from any pics I saw and isn't even an issue here. The quality of the components or workmanship are not at issue. They are clearly over-driving the chip by very significant amounts. I have one of these boards and am waiting to see if this gets resolved.

I haven't kept up to date 100% as I don't have 10-15 hours a day to burn on reading everything posted on the net when I have school and family to take into account. I decided over a month ago to not bother with owning a Kepler era card anyway. Kepler in general is crap and most cards are built worse than this card despite what MSI had done but as one poster had put it that it is an easy fix. Desolder the cap and spread some solder across the two points.

EDIT: If all else one can just add a turn pot to adjust voltages outside of software/bios.

 

[smorizio]

what msi does not want right now is someone starting a class action lawsuit.

 

[ram1009]

I haven't kept up to date 100% as I don't have 10-15 hours a day to burn on reading everything posted on the net when I have school and family to take into account. I decided over a month ago to not bother with owning a Kepler era card anyway. Kepler in general is crap and most cards are built worse than this card despite what MSI had done but as one poster had put it that it is an easy fix. Desolder the cap and spread some solder across the two points.

EDIT: If all else one can just add a turn pot to adjust voltages outside of software/bios.

It might well be an "easy fix" for those with the proper tools & skills to do micro soldering by hand however I doubt that description fits most of the people reading these posts. All these boards are still under warranty and most are still within the return period from the retailer provided the contents of the retail box were retained. IMHO, it's a very bad idea to suggest that modification should be attempted by the average reader of this forum.

 

[nforce4max]

It might well be an "easy fix" for those with the proper tools & skills to do micro soldering by hand however I doubt that description fits most of the people reading these posts. All these boards are still under warranty and most are still within the return period from the retailer provided the contents of the retail box were retained. IMHO, it's a very bad idea to suggest that modification should be attempted by the average reader of this forum.

I can fix it with a fire and a hot nail, practically stone age tech. If it be a fire and a piece of rock I could possible do it. I have done wire and tape before when I didn't have anything on hand.

 

[Maxx_Power]

I can fix it with a fire and a hot nail, practically stone age tech. If it be a fire and a piece of rock I could possible do it. I have done wire and tape before when I didn't have anything on hand.

"Fire and a hot nail" - awesome.

May I also suggest in addition to wire and tape, to use a pencil, ala pencil-mod ? Something dark and full of carbon will work just fine. Pencil the capacitor on each side and blow off the excess pencil carbon should do the trick nicely.

 

[mehta23]

"OR, it is actually an easy DIY repair to solder bypass that pesky capacitor they put in"

Would this void any warranty?

Also, I bought it from Ebuyer.com with only a two year warranty from them . THis applies to the first card, but not the second.

 

[nforce4max]

Just return it if you feel the need.

 

[mehta23]

I am from the UK , if this makes any difference to my position.

If I email MSI , telling them to upgrade me - what would I get ? A reference cooler 670?

 

[Mousemonkey]

It might well be an "easy fix" for those with the proper tools & skills to do micro soldering by hand however I doubt that description fits most of the people reading these posts. All these boards are still under warranty and most are still within the return period from the retailer provided the contents of the retail box were retained. IMHO, it's a very bad idea to suggest that modification should be attempted by the average reader of this forum.

So sad, there was a time when the average readers of these forums were the very people who would come up with and then try out these kinds of mods just for the fun of it. Thank you for highlighting what a shadow of its former self this forum has become.

 

[mehta23]

Double check the back of your card, to see if yours match the images posted on the German tomshardware site.

If yours looks exactly like this:

http://media.bestofmicro.com/G/J/354403/original/PCB 01.jpg

and zoomed in:

http://media.bestofmicro.com/G/U/354414/original/TL431 02.jpg

With that TINY capacitor next to the TL431 regulator, then YES. I would NOT hold onto, or recommend anyone holding onto a card that is designed with a part driven far, far out of spec. Even if it doesn't die, you can't be sure it will work in another computer with a different PSU (which is how this issue was discovered), so that there is practically NO resale value either because of this.

THis is just too bad, because I actually like MSI cards and their coolers.

I didnt really read this properly the first time , but is there a chance that my card wont have that capacitor?

 

[ram1009]

So sad, there was a time when the average readers of these forums were the very people who would come up with and then try out these kinds of mods just for the fun of it. Thank you for highlighting what a shadow of its former self this forum has become.

Unfortunately I couldn't agree more, however I find anybody suggesting rework on a surface mount board without proper tools and experience to be irresponsible. It will almost certainly lead to an unserviceable board that can't even be returned to the OEM. There's no way to disguise what was done once you do it.