Nvidia, Asus put the clamps on GTX 590 voltage

Apologies for being slow on the uptake with this news; I was out of the office on family business late last week. The fallout from the kinda-exploding GeForce GTX 590 cards continues, with both Nvidia and Asus taking additional steps to prevent the catastrophic release of the magic smoke sealed inside each and every card.

The main cause of the problem, as we’ve noted, appeared to be excessive overvolting during overclocking attempts and the failure of the card’s onboard power-protection circuitry to prevent the worst. We first understood this problem to be confined to an early driver set which shipped out to reviewers—and potentially to end-users, I believe—on a CD enclosed with the card. We used slightly later drivers in our testing, prior to the GTX 590’s release, and the card correctly slowed down clock speeds during our too-aggressive overclocking and overvolting attempts, saving us from fiery grief.

We now understand part of the potent cocktail of smoky death may have also been a modified video BIOS shipped on Asus cards that allowed more extreme overvolting than the 1.025V peak we attempted. (Since our review unit was a reference model directly from Nvidia, we weren’t able to venture much beyond that point.)

Now, Nvidia and Asus have made further adjustments in order to prevent problems.

The biggest news on this front, in our view, is that Nvidia has put the clamps on GTX 590 voltage in its drivers. We shot down a rumor last week that the public-release 267.91 drivers somehow reduced the performance of a stock-clocked GTX 590. They simply do not. However, we’ve since learned Nvidia has capped the voltage at the default 938 mV on GTX 590 cards starting in release 267.91. We’ve confirmed that the latest 270.51 betas will not allow us to raise our GTX 590’s voltage at all using MSI’s Afterburner utility.

This driver-based cap means that even fairly modest overclocking may no longer be feasible with the GTX 590. Our so-called WICKED config only took the GPU core to 690MHz, and getting there required a bump to 963 mV. With Nvidia’s latest drivers, a modest overclocked config like WICKED is not viable. The clock speed slider is willing, but the voltage is weak. GTX 590 owners will have to settle for whatever minor frequency increases they can achieve at stock voltages.

Given the way our GTX 590 card tolerated a reasonable bump up in voltage, we think this zero-tolerance voltage cap is probably a PR-driven overreaction. Then again, our WICKED config admittedly drew quite a bit of power—enough to make us uncomfortable. Either way, unfortunately, this change means GTX 590 owners will be robbed of what little headroom their cards may have in store.

We’ll call it the Way-Extreme Amperage Kap, or WEAK.

Meanwhile, Asus has released a new video BIOS for its GTX 590 cards that apparently limits its VoltageTweak overvolting feature to more manageable levels. Some folks have taken the release of this BIOS as confirmation that all GTX 590 cards have a problem, but Nvidia tells us it has not issued an official BIOS update for the GTX 590. Fuzdilla got an official statement saying the same. Asus appears to be the sole source of this updated video BIOS.

Incidentally, one of the juicier rumors we’ve heard in this whole affair is about that Sweclockers video of a GTX 590 going poof. Some of you have wondered how the site happened to be filming that GTX 590 exactly at the point of its death. We wondered the same, until a well-placed source hinted that the card in the video was not the first to die at the hands of that particular reviewer. We hear Sweclockers may have sacrificed its second review unit, a replacement card, in order to obtain that dramatic video footage, recreating the conditions of the first card’s failure with the camera rolling.

If true, wow.

One more bit of info for you. We’ve run the GTX 590 (at stock clocks) through the bulk of our GPU test suite using the new 270.51 beta drivers to see whether there are any notable performance gains, and the answer is no. Nvidia has cited some frame rate gains in specific games at specific settings for several cards, including the GeForce GTX 560 and 580 in single-card and SLI configurations, but none of those scenarios overlap directly with the games and settings used in our test suite. Outside of those scenarios, pretty much nothing has changed, as far as we can tell. You’ll still want to update your drivers in order to take advantage of the Dragon Age 2 optimizations, but don’t expect too much beyond that.

Comments closed
    • clone
    • 11 years ago

    you don’t see much at all Ryu.

    you wanted clock throttling on Coppermine and Socket A, you wanted shims on coppermine and socket A, I didn’t and was content with the absence of both features.

    see my expectations were met, no complaints, I then provided the reasons why I didn’t like the options, you see this as reinforcing your point because you are silly and only see what you want to see.

    when clock throttling became standard, when shims became standard I had no expectations I did what I always do, I worked to limit their effects on my efforts whether it be remove the shims or shave the heatsinks to compensate or switch to water cooling to control the throttling, that isn’t complaining that’s doing, no whining or demands just dealing with the issue as required.

    Nvidia’s choice to lockout the BIOS is because the VRM management is that weak, how the public was made aware of it really isn’t something to get upset about so much as to be thankful for and I thank Asus for revealing that weakness on launch day before I made the mistake of buying a GTX 590.

    superficially I guess my expectations that the GTX 590 would be equal in quality to the HD 6990 were not met and to be honest that is surprising given I thought the assumption that it would be was not unreasonable. Nvidia has done what it’s done and their is nothing I plan on doing about it because it’s not a concern, if it was I’d work beyond it like I always do in this case buying an HD 6990 because it doesn’t have the problem.

    • Ryu Connor
    • 11 years ago

    All I see are more expectations and complaints about how certain things didn’t meet them. If you goal was to reinforce my point you succeeded.

    As for the 590 it has headroom:

    [url<]http://www.hardocp.com/article/2011/04/03/asus_geforce_gtx_590_overclocking_followup[/url<] NVIDIA has decided not to allow that headroom as a business choice and you can blame a stupid vendor and people for it essentially.

    • clone
    • 11 years ago

    I have to disagree, having installed 300+ copper’s and Socket A’s none burned up or broken despite clocking 100+ of them and overvolting I found most issues overrated but times have since changed.

    I never wanted socket A athlons to throttle and hated P4’s clock throttling as it prolonged the time taken to establish an overclock limit.

    I also never once wanted Coppermine or Thunderbird to have shim caps because they affected cooling, shim caps made P4 installs a breeze but slowed cooling response, I pulled the caps off 3 Intel P4 3000 mhz cpu’s so that I could better cool/overclock them, 1 broke apart, 1 hit 4200mhz stable and the last hit a mundane 3700mhz stable all had water cooling.

    that “computer parts have limits” is a platitude.

    lastly I fully agree with RagingDragon given HD 6990’s aren’t exploding I expect the same from Nvidia and more importantly I expect my flagship card to have some headroom.

    • dragosmp
    • 11 years ago

    It didn’t seem to me he gave a bad review to the GTX 590 in itself, but he pointed out the 6990 has the components that do allow that kind of extreme overclocking while the 590 doesn’t. If absolute extreme OCing and max performance is the purpose, then the 590 looks bad. If the 6990 didn’t exist, the 590 would look a lot better as one may argue it’s close to impossible to deliver 400W+ on a graphic card PCB… but AMD did it.

    • Firestarter
    • 11 years ago

    Back then, up to 10% more voltage was considered somewhat safe. Anything beyond that and you started cooking the Northwoods!

    In that Sweclocker video, the 590 is running at 1.2v, or 28% more voltage than stock. That’s way beyond what any overclocker should use for regular overclocking with the stock HSF, that’s LN2 territory!

    • Waco
    • 11 years ago

    …except that AMD does not warrant the cards if you use the AUSUM switch. 🙂 To date, only XFX does.

    • Ryu Connor
    • 11 years ago

    Yes, expectations.

    Sort of like how I expected Athlon TBirds to have throttling so they wouldn’t burn up under poor cooling conditions.

    Or how I expected Coppermine or TBird chips to have a shim or cap to prevent cracking the core.

    There are lots of things I have expected from computers over the years. Very rarely do we necessarily get them when we want them.

    Computer parts have had limits and will continue to have limits.

    • RagingDragon
    • 11 years ago

    I expect a $700 part to be overbuilt. Some enthusiasts will use that headroom to overclock for increased performance, everyone else benefits from increased reliability and a longer lasting product when running at stock speeds.

    • RagingDragon
    • 11 years ago

    Subtlety is inefficient and thus has been optimized out of our lives.

    • l33t-g4m3r
    • 11 years ago

    It isn’t just the OC community, Nvidia itself is constantly hyping the overclock capability of their cards, when there is no proof the average consumer will ever hit those clockspeeds, or if it’s dangerous, or if it violates the warranty. EG: the 460. The funny thing was that even if you did overclock it enough to reach the speed of the 470, it ended up using more power and still wasn’t a better card. I guess the hidden agenda is that after you burn out your card in a year, you need to go out and buy another 200$ nvidia card. Of course that logic only works on die hard fanboys, as anyone with a shred of common sense will just switch brands.

    • MrJP
    • 11 years ago

    Surely you’ve now got to draw a much stronger distinction between AUSUM and WICKED in the original review? AUSUM may still require some user intevention, but at least the manufacturer doesn’t actively stop you from using it.

    • Ryu Connor
    • 11 years ago

    The enthusiast community has wide stripes.

    I’m an enthusiast, but I gave up on overclocking years ago. It always felt like a money pit for very little gain. By the time you spent the money for the quality heatsinks, fan, case work, or even something more extreme you begin to wonder if it wasn’t just easier to buy the bigger chip.

    I also couldn’t ever get past the hang up of wondering if every application crash might have been caused by the overclock. Given my recent penchant for mATX and SFF (something I’ve noticed on HardOCP has attracted some other dual GPU buyers) my thermal leeway for overclocks has naturally diminished as well.

    So in a sense I’ve managed to walk myself right out of the OC side of the house. Doesn’t mean I don’t live a bit vicariously through others though.

    I still like seeing what people can do. I love seeing the extreme stuff with LN2, I love how water cooling has grown into a truely accessible market, and so geeky weird that I think heatsinks have a certain visual flair to them. Especially these superbly machined and polished copper, aluminum, or even nickel that one could use in place of a mirror.

    So I appreciate what this guy did. Those two giant HSF he applied to that 590 were awesome sauce and I get his desires. I’m not bagging him for wanting to be able to do extremes. I’m more so bagging him for being so negative that the product didn’t meet his extreme. Not every product is going to have oodles of headroom and I think deep down the community has known for months (I’ve seen the random comments) that dual GPU Fermi on a single card would be break some fundamental constant of the universe. Well, guess what, it does. It has apparently met the big bang theory head on in the hands of some.

    To be fair to the reviewer he does mention in the article that even he fundamentally understood that his desires are cost prohibitive. He mentions that if they built the card the way he wanted it would have had more PCB layers, bigger voltage parts, and had *three* PCI-E power connectors – more specifically an 8, 8, 6 configuration.

    Such a card would be a beast and somewhat amusingly would also probably bring about a rash of complaints because it kills poorly made power supplies. 🙂

    It also wouldn’t be $700.

    I definitely get someone, somewhere out there bought a 590 and intended to OC the snot out of it. I’m not gonna screw them over and say they should be happy about this situation. They should be unahppy and I hope they don’t end up too much out of pocket on the return or ebay (whichever they choose).

    OC’ing with voltage tweaks on the 590 may at some point return, but Asus kinda ruined it for everyone. The drivers allow you to adjust the cards voltage and clock based on values read from the BIOS. The Asus BIOS allowed for extremes that could break the card. NVIDIA puckered up and pulled a soup nazi on everyone because of the Asus BIOS. Perhaps at some point they’ll ease back on all this.

    • eitje
    • 11 years ago

    As a TR regular, I think it would be awesome to get videos of systems while running, just for the hell of it. Put it right in the noise/power consumption part of a review!

    • dpaus
    • 11 years ago

    Likely, although he should actually be grateful that they saved Nvidia from untold millions in warranty claims and/or (even worse) bad publicity.

    • kilkennycat
    • 11 years ago

    I personally fully expected the GTX590 would have power-density problems, given the expected power-dissipation and the stated board size.. And the water-cooled versions of the GTX590 are now going to have the same artificial limitations embedded in the new drivers; the drivers look at the GPU chip-IDs, not the circuit-board variant. What a mess…………

    What single-card graphics-performance chasers with limited budgets really need is a dual-GF114 (GTX560Ti) card USING THE SAME VRMs and HEATSINKS/cooling as the current GTX590.
    In this case, there would definitely be sufficient power/cooling margin for a decent 900-950MHz overclock of the dual-GF114, a la the current set of factory-overclocked single-GTX560Ti.

    However, compared to a single-GTX580, the heat sinking and fan-size/airflow-capacity of all of the the currently-available single-GTX560Ti cards has been cheese-paired for the sake of a few bucks. Some of the so-called factory-overclocked versions of the GTX560Ti could have had GTX580-size ECBs and installed a variant of the GTX580 heat-sinking… they didn’t. I have some experience with the eVGA factory-overclocked GTX460 and 560Ti variants and well acquainted with the the fan yeowling like a mini-banshee when pushing performance in a single-card non-SLI setup. Fortunately I wear closed-ear headphones when playing high-performance games so audibly-weeping video cards don’t bother me too much.

    I fear that (for maybe $10-$20 parts-cost savings) that the upcoming dual-GF114 (dual GTX560Ti) card will again have inadequate VRMs, inadequate fan-cooling and inadequate heat-sinking, will be clocked at not more than 850MHz (when a single- GF114 with decent air-cooling will readily clock at 950MHz) and will again be subjected to a variant of the ridiculous voltage/power limitations of the GTX590 to artificially restrict any user-overclocking.

    • FuturePastNow
    • 11 years ago

    Good luck to Sweclockers in getting any more Nvidia review samples. They’re on Jen-Hsun’s **** list now.

    • flip-mode
    • 11 years ago

    Dunno, what I gather from the article is that the electrical components used are just barely sufficient to run the thing at stock voltage. To put it another way, the thing has no electrical headroom. Now, does that sound like an appropriate gesture on a elite-enthusiast class product? Dunno about you but I kinda feel like the answer is way-the-hell no. Jeez, I’m tempted to reply to your post with your own words:

    [quote<]Statements like that make me wonder what happened to the hardware enthusiast community.[/quote<] Can't say I really care as these cards are way out of my league, but if I ever did spend $700 on a video card that is aimed at the most extreme users, and then was told - "by the way, you best avoid overclocking" - I'd be pissed. I just don't think there's any excuse to put barely sufficient electrical components on a card of this class. Nvidia CUDA done better.

    • dpaus
    • 11 years ago

    Engineers: “Subtlety !R Us”

    • Ryu Connor
    • 11 years ago

    Hilarious (if crude) analogy. 🙂

    • dpaus
    • 11 years ago

    Oh, I agree: [i<]expecting[/i<] a 39% OC as some sort of birthright is ridiculous. But OC'ing is a bit like virginity; once you're told it's OK to start, it doesn't matter how far you go. And when review sites include OC'd cards as part of the manufacturer's 'standard/approved' offerings, then people expect to be able to OC them. "How far' isn't the issue; that fact that the GTX 590 can't be (effectively) OC'd at all will be a major P.O. to many potential customers.

    • Ryu Connor
    • 11 years ago

    I guess so. I’m begining to think my question is showing my age. I remember a time when a 39% OC or OV required the luck of a cherry chip and/or extreme cooling. Perhaps 39% or more has become a pedestrian matter.

    Still, a 39% overvolt does seem like one is asking to fry your hardware. My understanding that 570 and 580s will also gladly die in the 1.2 and 1.3v range and those are just single chip cards.

    • dpaus
    • 12 years ago

    I don’t think he said (or even implied) that it did. While poorly structured, the sentance says that even in the WICKED OC configuration, the card wasn’t faster than an AUSUM’d 6990, and that – here’s the unclear part – [i<]the 6990[/i<] is designed to be AUSUM'd, has full warranty support when AUSUM'd, and does not burst into flames when AUSUM'd.

    • dpaus
    • 12 years ago

    When sites (including, but certainly not limited to, TR) began including overclocked cards in testing reviews as a matter of course?

    • flip-mode
    • 12 years ago

    OMG BBQ: [quote<]Nvidia, who may not know that a very strong VRM can not be replaced by CUDA[/quote<] That's the best line in the article right there!

    • Ryu Connor
    • 12 years ago

    Would be more interesting if the guy was somewhat reasonable. He’s upset he can’t get 1300mv out of a card designed to run at 938mv. We’re in a strange world where 39% overvolts are apparently required in order to not get negative reviews. Not sure when that came became a written rule.

    • flip-mode
    • 12 years ago

    Winning?

    • SomeOtherGeek
    • 12 years ago

    Indeed, very nice read. Thanks, dragosmp!

    • Sargent Duck
    • 12 years ago

    Thanks. fixed.

    • derFunkenstein
    • 12 years ago

    There’s at least one factual error in your rant – the “WICKED” OC did not produce any flaming cards.

    • ew
    • 12 years ago

    Thanks for the link! I’ve never seen a review site talk about the VRM when reviewing a graphics card before and would have never found this.

    • sweatshopking
    • 12 years ago

    either way, it’s still more.

    • sweatshopking
    • 12 years ago

    Standard?

    • xeridea
    • 12 years ago

    ROFL

    • xeridea
    • 12 years ago

    So the Nvidia marketing video about the GTX 590 stating it is “really a technological marvel”, or whatever that guy said, is not really true. The “technological marvel”, is slower than the the competition, doesn’t OC well as I expected because the the power hungry Fermi architecture, and their supposed power management doesn’t do crap. The VRMs not even being able to reliably handle anything above stock is a bit risky in itself, and doesn’t bode well for the long term reliability of the cards. Perhaps if they didn’t need to cheap out because of the titanic GPU dies and need to be competitively priced, for an inferior product, they could find space to have proper hardware. Even with the super optimistic WICKED overclocks, it is slower on average than the 6990 with the AUSUM mode, which the card is designed to handle, is warranted when used, and has produced no flaming cards.

    • continuum
    • 12 years ago

    More GTX 590’s in existence, or did you mean more [i<]working[/i<] GTX 590's?

    • dragosmp
    • 12 years ago

    In the article linked below is an in-depth analysis of the VRM section of the GTX 590 (google-translated from Romanian amazingly well). The conclusions are pretty clear-cut, the GTX 590 has VRM sized so that it can barely support the power consumption st stock clocks and some mild OCing. An interesting afternoon read.

    Original:
    [url<]http://lab501.ro/placi-video/nvidia-geforce-gtx-590-studiu-de-overclocking/12[/url<] English: [url<]http://translate.google.com/translate?js=n&prev=_t&hl=en&ie=UTF-8&layout=2&eotf=1&sl=auto&tl=en&u=http%3A%2F%2Flab501.ro%2Fplaci-video%2Fnvidia-geforce-gtx-590-studiu-de-overclocking%2F12[/url<]

    • Sargent Duck
    • 12 years ago

    The golden standard

    • OneArmedScissor
    • 12 years ago

    I think there are more articles about GTX 590s now than there are GTX 590s in existence.

Pin It on Pinterest

Share This

Share this post with your friends!