Techreport: A note on rumors about gtx 590 issues.

[pcm81]

This is a quote from your link

While overclocking the HD 6990 we set the switch to the position 1 "AUSUM" setting to make use of the additional power. At this setting we were able to push the Cayman GPUs upward of 1000MHz, which was well beyond our exceptions.

so they didn't just flip the switch

This marginally concerns me... The 1000MHz OC is irrelevant as long as they did not increase voltage beyond those set by the swich change. Increasing clock speed can make the card to "crash" ie yield invalid results, fail CRC checks, the internal ones to the cpu/gpu etc, but pumping up Hz without increasing voltage can not burn up the card, cpu or what ever else...

 

[hdfxst]

I would think that running a 6990 @ 1000 on stock cooling would be about the same as running a 590 @1.2 volts

 

[cusideabelincoln]

If they clock down on older games when not being stressed, to me that is actually a good thing. Not only will it prolong the life of the product, but it will also present relief to your electrical bill.

I still play many older games. It would be nice to have a card automatically downclock itself when reaching a certain FPS threshold.

On the other hand, if I were the type of person who liked to overclock, this card would not be under my radar at all. I don;t overclock at all and I would still be kinda weary. I would wait until the reports of meltdowns go away.

I think it's simply a matter of time before this feature is implemented into the drivers. And hopefully it can be a user controlled feature.

I thought of this idea late last year, and AMD's Powertune was the first step in the direction I think we might be going. How I envision this feature is you plug in the desired framerate cap in the drivers, and the card will throttle clockspeeds and voltages on the fly. What would make this difference than individual games implementing a framerate cap is that drivers can adjust the voltage to further reduce power consumption. With this green movement, I would expect things like these to be implemented.

 

[nitromullet]

Relevance: Cards can die if pushed hard enough.

Well, of course... this is known.

Another question: why are you posting a bunch of third party links? I mean, you have one of these cards, don't you?

You've known me for a while, and you know I've owned both AMD and NV cards at the same time in the past. The reason for this is because it gives me the luxury of ignoring reviews, and basing my opinion on firsthand knowledge.

So, whatda say..? Pump 1.2mV through that thing and see if OCP kicks in...? How confident are you in NVIDIA's driver team?

 

[Ryan Smith]

I dont see those listed in Ryan's review all he has listed are

NVIDIA ForceWare 262.99
NVIDIA ForceWare 266.56 Beta
NVIDIA ForceWare 266.58
NVIDIA ForceWare 267.71

Did AT not receive launch drivers?

The press drivers for the GTX 590 were 267.71. These were made available to us on March 16th.

Also, did his card die?

No, obviously it did not - otherwise we would have said something. In fact normally I don't overvolt launch cards due to a lack of time and a need for them to be in working order for years to come. However in this case I did since we were looking at a manufacturer sample instead of a reference card, so I needed to in order to look at EVGA's ELEET. But for this reason we kept the voltage fairly low at 0.987v; this was already a .075v increase over stock, which is quite a bit for a dual GPU card.

 

[waffleironhead]

The press drivers for the GTX 590 were 267.71. These were made available to us on March 16th.

thanks ryan

 

[LiuKangBakinPie]

Cards don't blow at boot even if you oc the crap out of them coz their performance tables and voltage tables sets it to a low clock and a low voltage so the card can always boot. Only way to make a oc stick at boot is to edit the bios.

 

[Keysplayr]

Um, do you not understand what the rest of that sentence was? Are you claiming there was never any problem at all? Because, regardless of it being the driver that was killing the cards, the fact is, there are dead cards. Now, maybe the problem has been fixed, maybe it has not. One site's results does not mean you can extrapolate that across the board.

This is what I said:



And that is accurate. There was a problem. There might still be. I'm certainly not going to take your word for it. Its good to see some evidence that it is fixed, but you're being every bit as ridiculous by acting like that proves irrefutably that its resolved.

Its clear that this is being turned into some big fanboy BS, but please stop whining about things and then turn around and do them yourself. I don't enjoy you willfully ignoring my comment because you have an agenda.

This has nothing to do with the 6990. Ditto about maybe not throwing rocks when you live in a glass house.

I'm sorry, I thought I was responding in a respectful way. Maybe I didn't because you didn't return the favor. Apologies.

 

[Keysplayr]

The press drivers for the GTX 590 were 267.71. These were made available to us on March 16th.

No, obviously it did not - otherwise we would have said something. In fact normally I don't overvolt launch cards due to a lack of time and a need for them to be in working order for years to come. However in this case I did since we were looking at a manufacturer sample instead of a reference card, so I needed to in order to look at EVGA's ELEET. But for this reason we kept the voltage fairly low at 0.987v; this was already a .075v increase over stock, which is quite a bit for a dual GPU card.

Thanks Ryan.

 

[Keysplayr]

Well, of course... this is known.

Another question: why are you posting a bunch of third party links? I mean, you have one of these cards, don't you?

You've known me for a while, and you know I've owned both AMD and NV cards at the same time in the past. The reason for this is because it gives me the luxury of ignoring reviews, and basing my opinion on firsthand knowledge.

So, whatda say..? Pump 1.2mV through that thing and see if OCP kicks in...? How confident are you in NVIDIA's driver team?

Pretty confident. But, I'd say I'm not an experienced "overvolting" overclocker. Is 1.2 up from .923 significant? Besides, the 267.91's disable overvoltage.

 

[LiuKangBakinPie]

Well, I'd say I'm not an experienced "overvolting" overclocker. Is 1.2 up from .923 significant? Besides, the 267.91's disable overvoltage.

how do you disable overvolting without changing the voltage table in the bios?
So that means the entry is still there the driver just don't scan for it?
As far as I knows when the it reach its threshold it switches to the thermal throttle performance entry in the bios as well as the matching voltage.
You got a 590 right? Run Rivatuners diagnostic report on it. It will show you performance and voltage table internals

 

[Grooveriding]

http://hardforum.com/showpost.php?p=1037046423&postcount=170

Another GTX 590 explodes for an end-user.

 

[Keysplayr]

http://hardforum.com/showpost.php?p=1037046423&postcount=170

Another GTX 590 explodes for an end-user.

Says user "AthlonXP". There isn't any proof there.

 

[AnandThenMan]

If the 590 had better over current and thermal protection, then upping the GPU voltage would end up killing one or both processors. But that's not what is happening here, the regulator section is exploding.
I think D. Bauman said it best:

I scratch my head why drivers are needed for this. I would have thought there would be be hardware CTF and regulator shutdowns in place.

The new protections put into the latest drivers are a band-aid, but can't fix the fundamental design flaws. Hopefully this is corrected by non reference designs.

 

[nitromullet]

Pretty confident. But, I'd say I'm not an experienced "overvolting" overclocker. Is 1.2 up from .923 significant? Besides, the 267.91's disable overvoltage.

Yes, 1200mV is significantly more than 923mV - most likely enough to damage the card. I posed it as a challenge because IMO you would be relying 100% on the drivers' ability to protect the card via OCP.

I did not realize they had completely disabled over volting entirely.

I assume that you realize this adds credibility to the notion that the card was designed with components that are somewhat subpar to what we normally expect from an enthusiast class product? I wouldn't go so far to say that it's a substandard product, just that it's a little under built for its target market.

 

[LiuKangBakinPie]

If the 590 had better over current and thermal protection, then upping the GPU voltage would end up killing one or both processors. But that's not what is happening here, the regulator section is exploding.
I think D. Bauman said it best:


The new protections put into the latest drivers are a band-aid, but can't fix the fundamental design flaws. Hopefully this is corrected by non reference designs.

but nvidia boards have GPIO pins and is using a bmp structure with a valid VID bitmask for each entry. The target voltage identifier is used just to allow the driver to pick up the corresponding VID entry from the table coz the driver knows nothing about the VID is just knows the target voltage picked up from the corresponding performance table entry. So when programming the voltage the driver just picks required voltage entry from the table by scanning all the voltage table entries, comparing target voltage identifier with the voltage identifier with each entry and selecting the closest entry.. When the entry is selected, the driver disassembles the Vid and programs the non masked bits via the GPIO register. So how do you overvolt it?

 

[Sind]

Says user "AthlonXP". There isn't any proof there.

The guy has 15K+ posts almost 10 years on that forum, and prior he was a 580GTX owner. Continue to spin however. Just adding that I also consider "launch drivers" to be whatever is in the box despite whatever might pop up on the web on the launch day or afterwards. I'd also take this article with a grain of salt and if you're in the market for a 590 to give it some time to see how other users find their product and if there actually exists some issue, in that time I'm sure more thorough testing and reviews of the new drivers will have been done.

 

[pcm81]

but nvidia boards have GPIO pins and is using a bmp structure with a valid VID bitmask for each entry. The target voltage identifier is used just to allow the driver to pick up the corresponding VID entry from the table coz the driver knows nothing about the VID is just knows the target voltage picked up from the corresponding performance table entry. So when programming the voltage the driver just picks required voltage entry from the table by scanning all the voltage table entries, comparing target voltage identifier with the voltage identifier with each entry and selecting the closest entry.. When the entry is selected, the driver disassembles the Vid and programs the non masked bits via the GPIO register. So how do you overvolt it?

With the Flux Capacitor... Duuhhh!!

 

[3DVagabond]

Says user "AthlonXP". There isn't any proof there.

So now we need the user name to be Jen-Hsun Huang for it to be credible?

 

[Arkadrel]

Says user "AthlonXP". There isn't any proof there.

That guy has been on that forum for 9.5years... so its not like its some made up random poster.

Also the last guy that claimed to have a 590 that blew up, people said same thing... "where is the proof", and the guy posted pictures the day after or so.

 

[thilanliyan]

It is a really nice card. Aint it?

Are you being sarcastic? There are much better options for the same price. This, like the 6990, is an extreme niche product. You're better off with XFire/SLI IMO (unless you have 1 PCI-e slot, which is highly unlikely for someone looking to spend this kind of money...heck I have a lowly 790GX motherboard and I even have 2 PCI-e slots).

They did, otherwise the card wouldn't run.

There's a difference between just making the cut, and going the extra mile. In both cases the card would run (at stock...debatable I suppose in this instance ).

but pumping up Hz without increasing voltage can not burn up the card, cpu or what ever else...

It can if the power delivery circuity is not up to snuff. It probably won't kill the GPU, but it could kill the CARD.

 

[Pantalaimon]

With all the discussion, I still haven't seen a good explanation from NVIDIA to why a number of reviewers' card died during the review process, when there's been no situation like this happening in recent memory with other cards, from AMD or NVIDIA.

I mean, you can't say that it's just business as usual that this many cards died during reviews, and in the hands of some reviewers that can hardly be called novices when it comes to overclocking. Somebody messed up somewhere in the process.

 

[Keysplayr]

I assume that you realize this adds credibility to the notion that the card was designed with components that are somewhat subpar to what we normally expect from an enthusiast class product? I wouldn't go so far to say that it's a substandard product, just that it's a little under built for its target market.

I think they added an extra measure of protection in light of what happened. You said yourself that 1.2v was substantial. So, maybe the parts are subpar for 1.2v. but just fine for the normal operation of the card.

 

[Keysplayr]

That guy has been on that forum for 9.5years... so its not like its some made up random poster.

Also the last guy that claimed to have a 590 that blew up, people said same thing... "where is the proof", and the guy posted pictures the day after or so.

I can take a graphics card, place it on a table and take a picture of it and say it's dead or died at stock speeds and voltage or whatever too. Then share it with you guys. Some of you will believe it, and some of you will not.

 

[Keysplayr]

So now we need the user name to be Jen-Hsun Huang for it to be credible?

But you do understand my position, don't you? Situation reversed and all?