What do you think of nVidia locking down voltage?

[utahraptor]

I just will be holding onto my 1350 MHz triple unlocked MSI 680 Lightning with a death grip.

 

[Keromyaou]

The second comment in the 'brightsideofnews' site ( http://www.brightsideofnews.com/new...proving-quality-or-strangling-innovation.aspx) is very interesting. The guy said, 'The problem is that while the manufacturer can provide its own warranty, there is an inherent danger of getting the allocation cut'. This suggests that although Nvidia pretended like giving AIBs a way to do overvolting, in fact there was no choice for AIBs but giving up overvolting. It seems to carry too much risk for AIBs to choose the option to keep the overvolting while providing their own warranty. Something very sneaky seems to be going on with Nvidia.

 

[blackened23]

The terms nvidia provides isn't a choice. The AIB maker HAS to do it, they can't do business without the warranty.

 

[3DVagabond]

Just posting this as reminder of the recent past.

Many models even specialty models did not support voltage adjustment during the 58xx series run. All the reference boards did. Not so many of the non-reference.

I'm not sure why you are bringing this into the conversation? The AIB's deciding to go with a cheaper voltage controller on some cards is not the subject of this thread.

nVidia's hardware is fully capable of voltage control. It's being locked at the software level.

 

[3DVagabond]

I just will be holding onto my 1350 MHz triple unlocked MSI 680 Lightning with a death grip.

If they come out with an updated BIOS look the other way.

 

[Keromyaou]

Are there any truths about this guy's comment (http://videocardz.com/35287/evga-drops-evbot-support-for-geforce-gtx-680-classified-edition)?

He said, 'We support overvoltaging up to a limit on our products, but have a maximum reliability spec that is intended to protect the life of the product. We don’t want to see customers disappointed when their card dies in a year or two because the voltage was raised too high. ~Nvidia. What I got out of this is that Kepler cannot safely support voltages above 1.175V or NV will not warranty the claim made by AIB for failed Kepler chips that were overvolted beyond this spec. Fair enough. But this also says how flaky GK104 is. Can't even take anything above 1.175V? Talk about NV pushing this chip to the limits.'

If this is the real reason why Nvidia is fanatically trying to stop overvoltaging, this Nvidia's move might be limited to this generation's cards??? Overvoltaging could come back with gtx780???

 

[blastingcap]

Does anyone remember the rumors of gtx 680/70 degradation back when they launched? I wonder if NVidia's new policies have anything to do with those rumors.

http://www.xbitlabs.com/news/graphi...e_GTX_600_Due_to_Performance_Degradation.html

 

[Rifter]

Just realized no one posted this yet.

This is how i feel about it

 

[sontin]

Thanks, any info on the lower limit?

The lower limit is the vcore for the base clock. And you can't adjust anything. If you want to "undervolt" you must increase the base clock and reduce the power target.

 

[Anarchist420]

Are you ready for this...

I understood. Jen Hsuang acts like a govt, so all he is going to care about is getting money for those closest to himself and he isn't thinking long term because even if the market gets free enough to eat his institution alive (and it will) he will have enough money for life and he can just start a new institution for the gullible. I didn't want to believe it, but pretty much every institution now has been corrupted their buddy, the state. Fewer and fewer new ones are propping up and big firms like EA, nv, and AMD are kept alive by little more than patents and subsidies in the form of contracts. EA and Activision have been the statists of the video game industry since the beginning. Atari never should've been bailed out by the govt in the 70s, because they sucked compared to what the Japanese could do... the Japanese have bended over backwards for us, since time immemorial, and look at what we've done to them. The French developers like Ubi Soft have been kept alive by their own govt... that's going to go on until France and the French govt (including Ubi Soft) crash in its own way since they got rid of the logical loon and replaced him with the emotional idiot. The world is on its way to global fascism and a division between two worlds.

I hope I'll be around to enjoy the peace and change after the chaos settles, but I'll either be killed in it or maybe too unhappy if I lose loved ones. If I could just live the rest of my life in a world of confederalism (which the world was meant to be by the Laws of Nature) it would be happy and peaceful for all. Romney needs to just have me murdered in cold blood and maybe his buddies at nvidia can do it because I'm sure they're going to be unhappy that I woke up and helped to reveal the truth.

 

[cmdrdredd]

Wow...what was that rant about?

 

[thilanliyan]

I understood. Jen Hsuang acts like a govt, so all he is going to care about is getting money for those closest to himself

Actually, I'm pretty sure he cares very much about nVidia as a company, rather than just making money for himself. While I don't always admire nV's practices, JHH is a good CEO.

 

[RussianSensation]

Does anyone remember the rumors of gtx 680/70 degradation back when they launched? I wonder if NVidia's new policies have anything to do with those rumors.

http://www.xbitlabs.com/news/graphi...e_GTX_600_Due_to_Performance_Degradation.html

Just found this Link posted on another forum:

It appears NV confirmed with certainty GK104's voltage was set to the absolute maximum and the chip can fail if you increase voltage beyond stock voltage spec:

"We love to see our chips run faster and we understand that our customers want to squeeze as much performance as possible out of their GPUs. However, there is a physical limit to the amount of voltage that can be applied to a GPU before the silicon begins to degrade through electromigration. Essentially, excessive voltages on transistors can over time "evaporate" the metal in a key spot, destroying or degrading the performance of the chip. Unfortunately, since the process happens over time, it's not always immediately obvious when it's happening. Overvoltaging above our max spec does exactly this. It raises the operating voltage beyond our rated max and can erode the GPU silicon over time.

'In contrast, GPU Boost always keeps the voltage below our max spec, even as it is raising and lowering the voltage dynamically. That way you get great performance and a guaranteed lifetime. So our policy is pretty simple: We encourage users to go have fun with our GPUs. They are completely guaranteed and will perform great within the predefined limits. We also recommend that our board partners don’t build in mechanisms that raise voltages beyond our max spec. We set it as high as possible within long term reliability limits.

They're also leaving a bad taste in board partners' mouths: where in previous generations each company has been able to push its own cards to the limit in order to beat the competition, under Nvidia's alleged new rules all GTX 680 boards will be more or less identical in performance and features."

Now we have 100% confirmation that Kepler's GPU voltage was red-lined from the factory to the absolute safest max allowed. That means NV just officially confirmed that anyone increasing it above this level is playing electromigration lottery with GK104.

There is more to this than meets the eye. As I have hypothesized, NV also did this to prevent slower GPUs from having the ability to overclock beyond the faster offerings (i.e., 670 beating a 680).

"We've been told that the secretive restrictions on board partners go yet further: 'They [Nvidia] also threaten allocation if you make a [GTX 680] card faster than the GTX 690.'"

 

[sontin]

And your problem with that is?!

Maybe you should ask AMD how much more voltage you can give their 28nm chips over 1,25V.

Oh and i saw your edit: Can you show me a AIB 7950 card which is faster than a 7970? I mean there must be hundreds of them because AMD is not limiting their partners...

 

[RussianSensation]

And your problem with that is?!

Maybe you should ask AMD how much more voltage you can give their 28nm chips over 1,25V.

I don't have any problem. I think it's new information that NV confirmed this week that GK104's max safe operating voltage is stock voltage. That hasn't been the case for NV or AMD/ATI GPUs in the past. NV never came out and said: "Oh, and btw, if you overvolt GTX480 to 1.087V, it's basically toast long-term." With Kepler they are saying that any increase above stock voltage is most likely going to damage the chip even within the next 2 years.

Tahiti XT original stock voltage is 1.175V, not 1.25V.

AMD launched Tahiti with 1.175V and later expanded it to 1.25V safely, which means they left a lot of headroom above 1.175V stock. The 1.25V BIOSes were later available for free to all users for 7950/7970 cards. Now we just got a confirmation that NV pushed the voltage to the absolute max. Nothing wrong with that but NV should have said from the very beginning the reason voltage control was not allowed was because they pushed it to the max. Instead it took more than half a year for this to leak out after NV blocked warranty on after-market chips that support overvoltage. Do you think AIBs would spend $$ launching $600 GPUs with voltage control if they knew right away NV wouldn't offer warranty for those chips and later they decided, Jeez I changed my mind? Why did NV allow voltage control on Fermi cards but it's absolutely prohibited on Kepler? When was the last time a GPU/CPU company publicly came out and said that if you increase voltage even a bit above stock voltage, your chip is likely to fail? This is the first time I've heard of such a case.

Oh and i saw your edit: Can you show me a AIB 7950 card which is faster than a 7970? I mean there must be hundreds of them because AMD is not limiting their partners...

What edit? I was editing the main body. You can easily take a 7950, add more voltage say above 0.98-1.06 of MSI TF3 to 1.16V and take it above 1100mhz. At those speeds it's much faster than a 7970. You do not have this flexibility on most NV cards (well officially on none of them). In other words, you can overclock a 7950 way past 7970 speeds and still be under 1.175V stock voltage of Tahiti. If you have watercooling, you have the option to say take that MSI TF3 card to 1.3V and you may hit 1250-1300mhz. If you buy a Sapphire HD7970 DX for example, you get voltage control up to 1.3V from the Sapphire Trixx program. Looks like AMD is confident enough to take 7970 to 1.3V then since they are obviously letting Sapphire ship this software in the box. Surely if EVGA and MSI won't accept the lack of warranty from NV then Sapphire likely wouldn't do so either. That means AMD will warranty chips up to 1.3V from Sapphire's failed GPUs.

NV removing voltage control for future GPU generations will make it even harder for budget overclocking (i.e., for mid-range cards to overclock to top SKUs).

 

[sontin]

Tahiti XT original stock voltage is 1.175V, not 1.25V.

Stock voltage is between 1,0x and 1,175V. 1,25V is the vcore for the boost edition. You know these cards which using more than 250 watt...

AMD launched Tahiti with 1.175V and later expanded it to 1.25V safely, which means they left a lot of headroom above 1.175V stock. They releases those performance BIOSes for free to all users for 7950/7970 cards.

AMD never released the bios for the user. Go find it on their homepage...
They gave it to Reviewer so that these people must published the bios. And the bios is hurting the warranty of the cards.

Now we just got a confirmation that NV pushed the BIOS to the absolute max. Nothing wrong with that but NV should have said from the very beginning the reason voltage control was not allowed was because they pushed it to the max. Do you think AIBs would spend $$ launching $600 GPUs with voltage control if they knew right away NV wouldn't offer warranty for those chips and later they decided, Jeez I changed my mind?

nVidia is warranty a 17,5% increase in the vcore. AMD is not doing this. Maybe you should start to get your facts right.

What edit? I was editing the main body. You can easily take a 7950, add more voltage say above 0.98-1.06 of MSI TF3 to 1.16V and take it above 1100mhz. At those speeds it's much faster than a 7970. You cannot do stuff like that on NV cards since you have no voltage control. If you have watercooling, take that MSI TF3 card to 1.3V and you may hit 1250-1300mhz.

I can take my Gigabyte GTX670 - which is out of the box as fast as a GTX680 - clock it @ 1300MHz, increase memory to 3660MHz and i am 22% faster than a standard GTX680 without overvolting my card.

BTW: nVidia is allowing Gigabyte to bring such a GTX670 to the market: Nearly as fast as a GTX680 which a much better cooler, platine and $100 cheaper.
Can you show me 7950 which is out of the box nearly as fast as the 7970? I hope so. Because i really starting to thing that you are working for AMD...

NV removing voltage control for future GPU generations will make it even harder for budget overclocking (i.e., for mid-range cards to overclock to top SKUs).

Sure. Have i wrote that my "budget GPU" is 22% faster than a reference GTX680?

 

[RussianSensation]

Stock voltage is between 1,0x and 1,175V.

Max stock operating voltage for Tahiti XT is 1.175V, which was later extended to 1.25V (implying the original stock voltage of 1.175V was a conservative number). Stop discussing semantics that some chips ship undervolted. Max stock voltage for a product line is not a variance, but a fixed number. Stock voltage for a particular product you buy in the store varies like VID varies for CPUs. However, SB and IVB CPUs do have a max operating safe stock voltage and it's not the stock voltage you get when you plug the CPU into your motherboard. If you pick up 5 GPUs and they all have different stock voltage, that has nothing to do with the official stock voltage allowed for Tahiti XT chip, which is generally much higher.

1,25V is the vcore for the boost edition.

The 1.25V BIOS can be safely applied to all 7950/7970 cards if you want. Some may be stable, others may not but it will not kill the chip from electromigration. Thus the chip supports additional voltage increase from 1.175 to 1.25V without failure, or otherwise AMD would never have allowed reviewers to release it as a downloadable link. AMD does not guarantee that your 7970 chip will work with 100% certainty at 1050mhz with the 7970 GE bios, but they guarantee that if you flash the card to 1.25V, it won't destroy the chip from overvoltage. Otherwise they would never release a 1.25V BIOS.

You know these cards which using more than 250 watt...

I see, derailing the thread already into AMD vs. NV that you love doing at any opportunity. You and many people here have been proven wrong 100x already that 7970 GE after-market cards outside of Sapphire TOXIC 1200mhz 6GB do not use 250W of power but you keep repeating this like Obama talking about 5 trillion tax cuts, over and over.

HIS 1180mhz 7970 X = 224W of power at load

Still uses less power than a GTX580. It has also been shown ad-nauseum that HD7970 at 1150mhz @ 1.175V uses about 225-238W of power but this has nothing to do with this thread.

AMD never released the bios for the user. Go find it on their homepage...

That's not what AMD said. They shared the BIOS through review websites:

"We fully expect that for the class of gamer that uses a 7970 or 7950, they’re very savvy gamers. They’re guys that build their own systems or upgrade on a fairly regular basis and have the capability to flash a BIOS regularly and probably read the forums to know the BIOSes are available." ~ Source

Oh look here you go PCPerspective included the full 7950 B Bios if gamers want to use it. "AMD is allowing us to share the FW updater with you."

They gave it to Reviewer so that these people must published the bios. And the bios is hurting the warranty of the cards.

So does overclocking but AMD still went out of their way to provide this option. NV removed that option completely because they maxed out Kepler voltage from the factory but kept quiet about it.

is warranty a 17,5% increase in the vcore.

Kepler's stock voltage is 1.175V as far as I am aware. NV does not allow any voltage adjustment above this level. Therefore, NV warranties 0% increase in voltage above stock levels, outside of whatever bump occurs during dynamic boost (I believe up to 1.212V).

is not doing this. Maybe you should start to get your facts right.

I have my facts straight. My card allows voltage up to 1.3V from 1.174V. That option to 1.3V ships with software in the box. You are not getting it. AMD never said, look if you overclock beyond 1.175V, your card is toast. After they released 1.25V BIOS for 7950/7970 chips, they never said, look if you overclock beyond 1.25V, your card is toast. Don't turn this around into AMD vs. NV trying to justify NV's actions. :whiste:

I can take my Gigabyte GTX670 - which is out of the box as fast as a GTX680 - clock it @ 1300MHz and increase memory to 3660MHz and i am 22% faster than a standard GTX680 without overvolting my card.

That doesn't address the main issues - no voltage control, not warrantying voltage control for AICs which allowed them to differentiate their cards in the past, and what this means for overclocking of NV GPUs in the future if this status quo is maintained. It's a step back for the consumer.

BTW: nVidia is allowing Gigabyte to bring such a GTX670 to the market: Nearly as fast as a GTX680 which a much better cooler, platine and $100 cheaper.

See above.

can you show me 7950 which is out of the box nearly as fast as the 7970? I hope so.

I never said 7950 is nearly as fast as a 7970 out of the box. You brought that out of nowhere. I specifically stated that voltage control often allowed someone to buy a lower end SKU and overclock it much higher. GTX670 is a $400 card, not a $280 card. Can you overclock 660Ti to $500 GTX680 speeds? No, you cannot. You can with 7950 and voltage control is a huge reason for it. Go ask 7950 owners. How about GTX460/470 overclocks without voltage control? Much worse.

Because i really starting to thing that you ware working for AMD...

Not even close. I just don't share the view of loyal customers of a company defending anything their precious company does, even if it is a step back for the hobby. I have no allegiance to NV or AMD. My principles are price/performance and overclocking and NV for the most part disappointed on both this generation after having a clear lead for 1 quarter. If NV allows voltage control next generation and has great price/performance, I'll gladly recommend their cards over AMD's. Alternatively, if GTX600 series of cards fall in price to reasonable levels, I'll start recommending them again.

The thread title is "What do you think of nVidia locking down voltage?" I think it's a step back for the consumer and hurts overclockers. Your view seems to be the opposite I imagine since you are trying to argue against me? I linked at least 1 real reason that NV provided for why voltage control above stock is not allowed on NV cards - they could fail because stock voltage is max voltage for GTX600 series. It seems you don't like that response from NV or are upset I linked it. Again, this has nothing to do with the thread but you keep turning it personal and not focusing on the subject itself.

Sure. Have i wrote that my "budget GPU" is 22% faster than a reference GTX680?

Except Gigabyte GTX670 costs $400. Please continue defending why NV removing voltage control is great for the consumers.

 

[blackened23]

How did AMD enter this topic again? I'm confused

 

[RussianSensation]

I have no idea either, or how 7970 GE's power consumption is related, or how AMD never released a BIOS update with 1.25V increase that was safe but they did to reviewers who then released it to the public.

blackened23, can you please explain to me how Kepler's stock voltage has now become 1.0V and suddenly NV provides warranty for an additional 17.5% overvoltage to 1.175V?

Not sure why this is so complicated. I provided a link where NV confirmed themselves that going above stock voltage of Kepler could degrade the chip by eroding the silicon over time. This was their primary reasoning for not allowing voltage control. If we believe them at face value, then it is perfectly reasonable why they never allowed any voltage control on GTX600 series. However, I think they should have been more transparent about this since launch. Also, it seems odd to say the least that MSi and EVGA would develop GPUs with voltage control full well knowing that NV would deny them warranty.

If MSI and EVGA knew that NV would deny them warranty and still went ahead and made these cards, that means NV forced them to block voltage control later anyway for marketing reasons.
If MSI and EVGA were promised by NV behind the scenes that the failed chips would be covered by warranty, but later on NV changed its stance, then it also explains the situation but falls back on NV for why they flip-flopped.

Perhaps NV wants to launch GTX685 style update before GTX700? Perhaps NV wants to reduce RMA costs as low as possible. I don't remember NV taking such a hard stance against voltage control towards AIBs. It's basically telling AIBs that other than changing the cooler and I/O layouts and VRAM configuration, there really isn't much more value they can add to the GPU. That dilutes the differentiation factor among AIBs in my eyes.

 

[SirPauly]

As I have hypothesized, NV also did this to prevent slower GPUs from having the ability to overclock beyond the faster offerings (i.e., 670 beating a 680).

Many factory over-clocked GTX 670 sku's offer more performance than a default GTX 680 out-of-the-box over-all. This is what made the GTX 670 sku so compelling when first released.

You feel that over-clocked GTX 670's would defeat over-clocked GTX 680's -- and nVidia put the stops on volts?

I don't get what you're saying.

 

[blastingcap]

Just found this Link posted on another forum:

It appears NV confirmed with certainty GK104's voltage was set to the absolute maximum and the chip can fail if you increase voltage beyond stock voltage spec:

"We love to see our chips run faster and we understand that our customers want to squeeze as much performance as possible out of their GPUs. However, there is a physical limit to the amount of voltage that can be applied to a GPU before the silicon begins to degrade through electromigration. Essentially, excessive voltages on transistors can over time "evaporate" the metal in a key spot, destroying or degrading the performance of the chip. Unfortunately, since the process happens over time, it's not always immediately obvious when it's happening. Overvoltaging above our max spec does exactly this. It raises the operating voltage beyond our rated max and can erode the GPU silicon over time.

'In contrast, GPU Boost always keeps the voltage below our max spec, even as it is raising and lowering the voltage dynamically. That way you get great performance and a guaranteed lifetime. So our policy is pretty simple: We encourage users to go have fun with our GPUs. They are completely guaranteed and will perform great within the predefined limits. We also recommend that our board partners don’t build in mechanisms that raise voltages beyond our max spec. We set it as high as possible within long term reliability limits.

They're also leaving a bad taste in board partners' mouths: where in previous generations each company has been able to push its own cards to the limit in order to beat the competition, under Nvidia's alleged new rules all GTX 680 boards will be more or less identical in performance and features."

Now we have 100% confirmation that Kepler's GPU voltage was red-lined from the factory to the absolute safest max allowed. That means NV just officially confirmed that anyone increasing it above this level is playing electromigration lottery with GK104.

There is more to this than meets the eye. As I have hypothesized, NV also did this to prevent slower GPUs from having the ability to overclock beyond the faster offerings (i.e., 670 beating a 680).

"We've been told that the secretive restrictions on board partners go yet further: 'They [Nvidia] also threaten allocation if you make a [GTX 680] card faster than the GTX 690.'"

This post deserves a sticky.

I would also emphasize that last sentence about threatening to withhold GPUs from non-complying partners. This is not some benign recommendation.

Is there a reason why Sapphire (AMD-only) seems fine with going over 1.175v on AMD GPUs even though they are built at the same fab (TSMC)? I vaguely remember one AT forumer who said he asked Sapphire what the max reliable voltage was and Sapphire said "below 1.25v" which implies something very close to 1.25v (like, say, 1.24v), even if it technically could mean some other threshold. Did AMD build their GPUs with more robust structures or something; or did AMD not do their homework and those running >1.175v on their AMD GPUs will suffer from electromigration early death; or is there some other reason for the difference between 1.175v (NV) and ~1.25v (AMD)?

Personally I think this is the result of NV pressing what was supposed to be their midrange GPU into service as a high-end GPU. They had to really push that chip's limits, leaving little to no overvoltage breathing room. This is the GPU we're talking about, not other components, so other than tighter voltage regulation, I don't think there is much that the AIBs could do even if they were told to push the chip further. And there are probably negligible gains to be had from tighter voltage regulation. In other words, the GK104 is "redlined" as RS put it.

 

[cmdrdredd]

Stock voltage is 1.175v. Always has been. When I fire up a game the cards I have will jump to 1.175v even at stock levels with no overclock or power target increase. That is stock voltage. At idle it's much lower but nobody compares idle levels to full load. That's absurd and Intel and AMD would laugh at you if you said "you guys are doing factory overclocking when you use Turbo frequencies and drop down at idle".

Also LOL @ threaten allocation if you make a 680 faster than a 690. What does that say about their $1000 GTX 690? haha

This is too funny guys. What's the next step in limiting your partners? Tell them that if they build AMD cards they have to underclock them to make the Nvidia counterpart look better?

 

[SirPauly]

They already said that here as well:

We support overvoltaging up to a limit on our products, but have a maximum reliability spec that is intended to protect the life of the product. We don’t want to see customers disappointed when their card dies in a year or two because the voltage was raised too high.

 

[Keromyaou]

The most important part of the article posted by Russiansensation is the alleged statment by Nvidia's spokesperson who admitted that 1) more than maximum voltage of Kepler will degrade it fast, 2) the max safe voltage is already applied, and 3) they understand the importance of voltage control among customers. Nvidia seems to fear many RMAs for Kepler cards might happen in the next two years if they allow voltage control for even custom models (Since the degradation happens in gpu chips it doesn't matter wether they use quality parts outside gpus for their custom models or not). They seem to understand the negative impacts for prohibiting overvoltage on their customers, which is a good news for customers since they might bring back overvoltage in the next generation gpus. Bad news is that Kepler cards might be already pushing the limit so that even without overvoltage the lifespan of the cards could be a bit shorter than let's say Fermi cards. Of course we don't know for sure that this is the case. However it surely makes you worried a bit if the cards operate at the maximum safety limit all the time (Who knows if many Kepler gpus might start dying suddenly two or three years later even without overvoltage? I don't trust what company representatives say at the face value). If you are a savvy customer, this is how you might think about the quality of Kepler gpus. If Nvidia was very explicit about the issue from the beginning many customers might have decided to hold off buying Kepler gpus and wait for 700 series especially the performance is not that different between gtx580 and gtx680. In this sense I can imagine why Nvidia hasn't been translucent about this issue for a long time.

 

[SirPauly]

is the alleged statment by Nvidia's spokesperson who admitted that 1) more than maximum voltage of Kepler will degrade it fast,

I can't seem to find that part.