RTX 3080 causing crashes

The important part of that story was that EVGA engineers discovered it before the cards were ever delivered to customers and delayed the affected parts. Unfortunately, a lot of other companies weren't as on the ball or didn't care to deal with it before they sold them. That's why EVGA rocks!

"EVGA's statement says that engineers were doing quality control on the new cards when they found that a specific 6-capacitor configuration "cannot pass the real world applications testing." The replacement of some of the capacitors, called POSCAPS, with a larger number of alternate MLCC capacitors, fixed the problem. "This is why the EVGA GeForce RTX 3080 FTW3 series was delayed at launch," said EVGA."


By the way, if you look at the pics for the EVGA products, you can see that they don't use the 6 cap array, they have either 4 or 5 depending on it it's a FTW or an XC variant. Either way, the EVGA cards should not have this issue.

EVGA GeForce RTX 3080 FTW3 ULTRA GAMING https://www.evga.com/products/product.aspx?pn=10G-P5-3897-KR&associatecode=XX95PJN1T76IONP

1601231314749.png
 
Last edited:
I am an EVGA customer. They are far from perfect. Yes, they caught it before customers had the product. The only way this is "good on them" is if they do not have control of the supply chain (what parts are chosen to be used and therefore QA) and specs where the cards are manufactured. I think they do. My guess this was an attempt (in good faith) to cut costs and not sacrifice performance and it backfired.

Saying they are soo good because of the error, is overlooking the fact that someone f'd up and how that actually impacts them as a business.
 
Last edited:
I think you may be missing the point here. A number of manufacturer's cards are having this problem when they are boosted beyond a certain point. EVGA was the first to OFFICIALLY talk about it publicly and the focus of the article was that they picked it up while pretesting the design, identified the cause, and delayed selling the cards until they improved the capacitors responsible for the power delivery system.

The design specs for AIB's are provided by Nvidia. It looks like the NVIDIA FE cards, EVGA, and some Asus TUF ones do NOT suffer from the issue, but other companies like Zotac, MSI, Gigabyte etc are the ones that are crashing. The partners are free to choose what components they use as long as they follow the basic design that is supplied by Nvidia. That's why I said 'good on EVGA" for catching it before they put them up for sale. I don't recall ever saying "they are soooo good because of the error" so not sure where you're going with that.

Here's a good read if you want to learn more with pictures of the components and a good explanation of the timeline and how other companies are responding so far.

https://videocardz.com/newz/manufacturers-respond-to-geforce-rtx-3080-3090-crash-to-desktop-issues
 
Last edited:
I had a couple of bad EVGA cards about 7 years ago.
 
Last edited:
I am of the opinion there isn't an issue outside of a lack of limits. These cards boost way past their advertised speeds which is great. But if the answer to the problem is that better components allow better performance... that's always true and all of a sudden every card ever made is a potentially faulty piece of shit. Which isn't the case. Power quality has always been a consideration for overclocking and instability is a risk when you give a boost algorithm the authority to do whatever within X temp or X amperage. It's crazy that implementations so far have done as well as they do.

A reasonable upper limit determined by the OEM on a per SKU basis with the option to turn it off for enthusiasts and Nvidia having decent drivers for testing would have turned this into an entirely different story. The moment you say "It got unstable after I turned the limits off" people have an entirely different attitude about it. A quick response to appease crazy rabid customers isn't something that impresses me much. EVGA is a decent company but with everyone getting caught up in this it's hard to say anyone did anything further wrong.
 
ok this is my take on what i have seen....i maybe right, i maybe crazy or i maybe flout out wrong....but no matter what this is going to be good and all of us are going to have to wait to see how it plays out

NgreedyA done fucked up BIG....on a few sites (now taken down) base level tech drawings of the 3080's lay was shown (these drawings are from files sent by NgreedyA to 3rd party makers to build their PCB's)....these drawings showed NgreedyA OK'ed the designs that used a 6 cluster of CAPs for the main 6 rails feeding the GPU directly, that have been reported so far as the problem when taking the GPU clock speed over 2ghz.....so far i have seen posts of MSI, ASUS TUF, EVGA pre-release, Zotac and even starting to see a few reports of NgreedyA's own GPU's showing the issue but none have been confirmed outside of EVGA at this point to my place in reading everything.....the only thing every post says is that the problem happens when the GPU goes over 2ghz in clock speed ether from over clocking or from factory boost set clocks pushing that speed....rumors are already flying that some cards will end up with firmware updates that limit speeds to a level that keep the card stable with a lot of things being said that 3rd party makers are not going to replace or fix effected cards (if this is true i am so fucking happy the bots got all the cards on the first run)


i know this will take time to figure out but IF we have more then 1 maker truly having problems with 1 type of power delivery design, then it must be placed on that and where the design came from....NGREEDYA

so IF this^^^^ is all true be happy you did NOT get a first run card (never mind the fact we have leaks showing higher Vramed versions of all cards coming soon)
but there is going to be a lot to take in on this as the weeks go by IF more versions of the these cards show up with the problem from other makers....
STAY TUNED KIDS and make sure you have popcorn

on the EVGA part i have had good and bad cards from them over the years along with the same in power supplies BUT for the most part they have been good over all for my usage....my current 1070 is made by them and never had a issue in 1000's of hours of game time never mind the years i put on my 560ti 2gb back in the day ....or my 260 216core.....but 2 460's i installed in friend's builds have fuck tons of issues and where nightmares
 
At base level, this just comes down to facilitating digital programming. The solution can only be total logistical paradigm shifts. We need a more blue-sky approach to 'Outside the box' modular contingencies. This is no time to bite the bullet with our integrated reciprocal projections. The consultants recommend functional modular mobility.
 
At base level, this just comes down to facilitating digital programming. The solution can only be total logistical paradigm shifts. We need a more blue-sky approach to 'Outside the box' modular contingencies. This is no time to bite the bullet with our integrated reciprocal projections. The consultants recommend functional modular mobility.

This guys running out of room over here. Quick we need some vertical integration so he can aggregate seamless mindshare of innovative convergence.
 
So I haven't watched half of the videos on the issue but Der8auer put up a new video 5 hours ago. 200Mhz or so overclock on stock. Maybe 300mhz after changing the caps. It's not going to work any miracles but with AIBs riding that tipping point... Perhaps this will have some influence on the industry standard practice of not having fucking drivers... Oh who am I kidding, next time they won't release drivers until cards are on shelves for a month.
 
Back
Top