KitGuru KitGuru.net – Tech News | Hardware News | Hardware Reviews | IOS | Mobile | Gaming | Graphics Cards
It has been impossible to ignore over the last week the problems surrounding Nvidia's new dual GPU card, the GTX590. After the high profile video capture of a sample failing on video there have been threads on many of the leading forums from people who have had their brand new £600 investment fail.
KitGuru heard recently that partners were rushing to get their bioses updated, we had assumed at the time that Nvidia had stepped in to release a new bios which would help ensure that thermal limitations were set firmly in place.
After we published our story, Nvidia took the time to reply to Fudzilla and have said “NVIDIA's reference GTX 590 VBIOS on the production boards has not changed, and we have not provided a new VBIOS to AICs or end users for GTX 590. The original GTX 590 VBIOS that NVIDIA shipped is good for customers to use and we stand fully behind the reliability of the hardware with that VBIOS.
We did hear of one AIC partner who issued a custom VBIOS update specific to their board and we are checking in with them on the details of this.”
KitGuru has actually heard of more than one partner who are making changes to their BIOS configuration, but we need to ask the tough question. Is there actually really anything wrong with the GTX590, or are nVidia being unfairly treated?
Nvidia originally issued this statement : “A few press reports on GTX 590 boards dying were caused by unsafe overvoltaging (as high as 1.2V vs. default voltage of 0.91 to 0.96V), & using older drivers that have less overcurrent protection. Rest assured that GTX 590 operates reliably at default voltages, and our 267.84 launch drivers provide additional levels of protection. See our knowledge base for more info (http://nvidia.custhelp.com/cgi-bin/nvidia.cfg/php/enduser/std_adp.php?p_faqid=2947)”
We thought that was the end of the problem. The updated driver would cure the issue and we could all move on. As the graphics hardware gets more powerful, the heat rises and greater cooling power is needed. This is logical.
When AMD released their HD6990 they issued a statement that they weren't supporting the higher bios setting. KitGuru published reports on this, and we were the first to get the official statement. This caused a fair amount of flak for them, and when XFX stepped up to say that they would support their end users, we all felt it was a very good move.
This brings us back to the GTX 590. While Nvidia updated the drivers to stop the cards failing, other users were reporting that their GTX590 had died even when running at reference clocks with default voltage. Some have blamed poor VRM design on the failures, but could it be that the card is having to deal with too much heat? Are the bios changes that some partners are making to help with enhanced thermal limiting?
Testing the temperatures of a video card is a difficult task, many publications use simple driver readings, and others can use diodes which can be attached to various positions of the PCB and cooler. They are never totally accurate. It is worth bearing in mind that often software related thermal readings can also be wrong.
One KitGuru reader brought it to our attention that Hardware.fr use a thermograph to accurately measure video card temperatures. Their GTX590 testing showed a reading of almost 112c under load.
I don't think anyone could argue that this is scientific and without repute. The more important question to answer now will be the reliability over the long term. I know when ATI released their dual GPU 4870 X2 a few years ago that the temperatures were often reading over 100c, and while they would never admit it they suffered from their fair share of problems.
If you are buying a GTX 590 today and spending a small fortune, how will this investment hold up in 9 months time, especially if the 112c readings from Hardware.fr are going to be maintained for months while gaming? Obviously getting a straight answer from any PR representative is going to prove difficult, and it is rare in history that any company will admit problems with their hardware. Is the GTX 590 perfectly fine now that the drivers have been updated? Or are there more serious problems with the design and long term reliability?
Kitguru says: Time will tell.
Those are some serious concerns with the design, I dont think it can be ignored……
I had a 4870X2 ! it died in 3 months. I noticed it would sit around 95c quite often, so id love to have seen a thermo reading of it. I know ATI said at the time it was fine, but we can all tell when a card is just running too hot.
GTX590 looks to be a long term RMA disaster, im glad im not a partner building those cards.
I havent had any of these problems, ive oc’ed and it never hits these temps. Im getting 80*c MAX under heavy load. This card is far cooler than the 6990.
@ A. you are getting what the driver is telling you. read the article first see what its all about. thermal readings dont lie. hardware.fr are great for this.
Yeah its like mobo bioses. they underread and read off all the time. it depends how its been set up. thats the point. the card will read whatever nvidia tell it to read.
Thats why the hardware.fr article is really making such an impact. but you never know, according to nvidia fan boys, hardware.fr might have put a fire under the card before taking the image reading. just to make nvidia look bad? lol
While I think the heat is quite excessive, this is ALL clearly a result of a poorly designed power delivery system.
No matter how many different drivers Nvidia releases to limit power draw (read: performance), no matter how many BIOSes they release to further confuse readers (and apparently reviewers as well), the facts will not change. A 4+1 VRM design (for each GPU) is just not enough for this card if the mosfets are only rated for 35A each.
Thanks Zardon. I appreciate it.
After reading of the many problems that are happening to these cards it is clear that Nvidia under built this card or worse rushed it to market before it was ready. It seems like they did not test it fully in different setups to make sure the card would not over heat & then fail. Did they want to try to reclaim the king of the hill rep again that they would release such a poorly designed card like this. Sure it is fast but I will take a slower & more reliable card over this ting any day.
As for the 4870×2 having problems I still have one of those in one of my setups & it has never failed me yet & I have had it since almost launch day & it has many many hours on it of hard core gaming. Yes it runs hot but the card was designed to take the heat now if your stupid & over clock the reference cards yes they may fail after 6 months of extreme overclocking & hard core gaming. In the ATI/AMD control center there is a thing called fan control turn it up a bit & the card never hits anything over 79 degrees.
This also makes the AMD 5970 & 6990 like like real champs since we all know AMD over built them so they could be run at faster speed rates more so for the 6990. Before anyone says I’m a AMD fan boy nope but I do like a quality built product. I also own a Nvidia GTX 470 in one of my gaming systems & it has been a very solid card as well but seems to be slower than the 4870×2 by a small amount.
just my take on all of this.
http://www.overclock.net/nvidia/929152-have-you-killed-570-a.html
Honestly, there’s no surprsie to this. The GTX 590 has a total of 12 VRM’s, that’s 6 per GPU. If we quantify that, you’re assigning 4 to the main GPU, 1 to the memory controller, 1 (dual phase) to the RAMs. This is pretty inadequate since full GTX 580’s are usually complemented with 6+2 power phases, here in the 590, Nvidia gave only 4+2. It’s no surprise why the card fails under such treatment.
VRM insufficiencies affect Motherboards as well.
http://www.overclock.net/amd-cpus/943109-why-vrms-big-issue-why-choose.html
GTX 590? Wait until the Asus, MSI or Gainward/Palit non-reference, then we’d be talking power.
There’s no question on whether or not this is caused by the VRMs. The VRMs are what are failing. They’re underspecc’d for the job, at 35A each on the GPU/IMC phases, and only four phases per GPU (plus one phase for the IMC).
The 112*C temps you’re seeing are a side effect/contributing factor to the failures, not the cause. The DrMOS chips are being run near their maximum peak forward current rating with a high duty cycle, when the card is being stressed. At these load levels the DrMOS chips become very inefficient (being optimized for efficiency at low loads, which is why they shouldn’t be used for VRMs for high power electronics). That inefficiency means more waste heat. Add to that that the VRMs are placed partially under the fan’s dead spot, and it’s no wonder they’re getting that hot.
And the greater heat lowers the practical current rating of the DrMOS. Then you increase the clock speed (increasing peak current draw) and the voltage (increasing mosfet duty cycle), and ***BANG***. No more GTX590.
Or you just put the fan ratio higher than stock. Those GPU’s are marked at around 90C~100C to only then kick in the fan hardcore. But if you just put the fan 10%~20% higher already it is bound to run much cooler and when you are gaming (headset or good speakerset) you shouldn’t notice the fan noise anyway. 🙂