RTX Space Invaders Wanted

The only thing that can be done is for the truth to come out.

I understand what you, personally, are citing. However, I'm going to trust what I know and the word of hardware legend Kyle Bennet over anything that's been reported by Nvidia at this point.

I have had 2 bad cards in a row and my third still exhibits issues that lead me to believe it will fail at some point. I must be part of your 3.5 percent then.. lol

But that is also my point, I have not seen this 20% figure anywhere. So all I have to go on are the 3.5%. But even at 20%, it doesn't really change things. As for the truth, I think we all want that, but it may be that Nvidia still doesn't really know what the truth is. They said there was a test escape, that may be the truth (doubtful that is the only reason, especially since they mentioned only FE cards). They may not know what happened exactly with it or how to fix it yet. So what truth would we really get at that point? This is why I am as anxious as you guys are to see what Kyle can find out.

As for your personal experiences, just because you have had a number of cards that have failed, that doesn't mean the 3.5% isn't legitimate. I went through 3 8800GTXs, and 4 Nvidia motherboards at one point. That didn't change the fact that the failure rate was still extremely low for those products. Also note that sometimes we can't always see the full problems when people aren't using the full extent of the cards either.

It's question that can't be answered until Nvidia either admits there is a problem or is forced to admit it. Unless of course you take their their BS about test escapees as admitting a problem. It's a shame that Kyle's investigation wlll be cut short because of his new job.

So we should blame Nvidia for apparently doing nothing even though there is no consensus on what should actually be done? Nor is there any evidence yet of what the specific problem is? And while using test escapes for FE specifically is BS, it still lends one to believe there is a problem with the manufacturing process, especially since we also see issues now with AIB cards.

Again, it isn't that Nvidia is not at fault here, but it still begs the question, What exactly do you want them to do? You want them to admit there is a problem? What problem? Do you think people will be satisfied just by them saying there is a problem with the cards? They have already said that at least once and you called it BS. So if they come out and say after further examination they have found there is a problem with the 2080Ti, and that's all they know right now, will that make you satisfied? What would be the point of that?

This is like people crying all over the place that the system is broken, but not knowing why it is broken or offering any solutions on what can be done.
 
The failures afaik are on 2070's, 2080's, 2080Ti's, RTX Titans and I don't have any figures on the 2060's but suspect they are suffering in similar fashion . The 1660 and 1660Ti are made from the same silicon... I don't have figures on those recent boards tho.

But what amount of failures on those others? All models will have some failures, is it under 5% on the others, or 20%? Seems like initially when these issues started appearing, I only recall 2080Ti users having the issues. What sources do we have regarding the other models and their failure rates?
 
But what amount of failures on those others? All models will have some failures, is it under 5% on the others, or 20%? Seems like initially when these issues started appearing, I only recall 2080Ti users having the issues. What sources do we have regarding the other models and their failure rates?
Sounds like a great time for you to become an investigator. Honestly, if you're so interested do some research. By that I mean, dig through these forums and Nvidia's and AIB partners and you will see the rest of us are talking about.

There is plenty of history here with issues. Plenty of viable info if you look hard enough. You're not going to get hard numbers because of the Nvidia NDA.

However, you might just trust what some of us have seen or perhaps what Kyle has already communicated.

There's way too much drama here about this and a lot of disbelief from a lot of people all too willing to blindly move forward accepting all the bullshit Nvidia officially serves out.

If this were AMD I would be saying the same thing, do your own research and don't accept what they say. Find your own truth.
 
<-- not an investigative journalist

Why do you think I will miss [H] so much...

Reading thru forums will not be accurate. I might find 20 posts from 2080 (non-TI) card owners who had issues, which might be 1% of all of those who had issues, so 2000 cards bad, inferred (guesstimate). But if they've sold 50,000 of them, not a smoking gun.

So no... I'm not going to 'guess' based on what I see from browsing a forum. I want data from an AIB or reseller. Forums are full of trolls and haters "OMG, my 2080 has been rma'd 3 times!" from a guy running a vega56 with an axe to grind. Fanbois are crazy like that...
 
<-- not an investigative journalist

Why do you think I will miss [H] so much...

Reading thru forums will not be accurate. I might find 20 posts from 2080 (non-TI) card owners who had issues, which might be 1% of all of those who had issues, so 2000 cards bad, inferred (guesstimate). But if they've sold 50,000 of them, not a smoking gun.

So no... I'm not going to 'guess' based on what I see from browsing a forum. I want data from an AIB or reseller. Forums are full of trolls and haters "OMG, my 2080 has been rma'd 3 times!" from a guy running a vega56 with an axe to grind. Fanbois are crazy like that...
My point was, you don't have to be an investigative journalist to reach an informed conclusion.

It's up to us, now, more than ever. We no longer have the luxury of being spoon fed. I will miss the old [H] terribly. However, if I learned one thing from Kyle it's that we need to find our own truth. All of us can do this. That's a big part of what made the [H] great, this place inspired me to search for answers and the truth on my own.
 
<-- not an investigative journalist

Why do you think I will miss [H] so much...

Reading thru forums will not be accurate. I might find 20 posts from 2080 (non-TI) card owners who had issues, which might be 1% of all of those who had issues, so 2000 cards bad, inferred (guesstimate). But if they've sold 50,000 of them, not a smoking gun.

So no... I'm not going to 'guess' based on what I see from browsing a forum. I want data from an AIB or reseller. Forums are full of trolls and haters "OMG, my 2080 has been rma'd 3 times!" from a guy running a vega56 with an axe to grind. Fanbois are crazy like that...

I literally had two cards die on me with receipts to prove it. I ended up picking up a used 1080 Ti Strix after giving up on two 2080s. I still prefer Nvidia over AMD but the 2080 seems risky.
 
If the 2080's are doing it too, then this 12nm TSMC process has issues... Sucks! Get moved to 7nm already.

I think nVidia should extend the warranties of these cards.
 
I literally had two cards die on me with receipts to prove it. I ended up picking up a used 1080 Ti Strix after giving up on two 2080s. I still prefer Nvidia over AMD but the 2080 seems risky.
I feel for you man, I want another video card to upgrade the front room rig and I just can't bring myself to pickup another 20 series card. I don't care for the Radeon 7 and its too damn big for the Home Theater chassis I have. I may have to consider picking up a used 1080Ti now that I gave my old one to the cousin.
 
I feel for you man, I want another video card to upgrade the front room rig and I just can't bring myself to pickup another 20 series card. I don't care for the Radeon 7 and its too damn big for the Home Theater chassis I have. I may have to consider picking up a used 1080Ti now that I gave my old one to the cousin.

Radeon 7 seems like a great card but the reference one looks like crap. Too loud and hot. I hope Navi will bring something powerful.
 
If rumors are correct Navi 20 in 2020 is supposed to be high end, possessing Ray Tracing support / acceleration and may ... Be faster than the 2080Ti. Of course Nvidia will just release their 7nm part when that happens. The 20 series Nvidia parts are likely just a cash grab and I don't think they will be long lived.

Nvidia cannot allow AMD to regain market share. Like I've said before, if they lose any ground to AMD the next market shakeup will really hit their bottom line when Intel releases their cards.
 
Slighty off-topic, but since the Radeon VII is a 7nm GPU, that is the same GCN5thGen core architecture (Polaris), it should be a great indicator of the % gains to be had from going from GloFlo14nm to TSMC7nm.

Reviews comparing Vega64 to Radeon VII show around 20% to 40% speed improvements. Once nVidia does this with RTX, Raytracing will be a lot better performing.

Back OT: Really hope that by then, the issues that plagued these 20 series GPU's are resolved. Looking forward to going RTX on a 7nm GPU, with hopefully a lower cost.

These transistors are getting so tiny, it's amazing that the yields are as good as they are...
 
Slighty off-topic, but since the Radeon VII is a 7nm GPU, that is the same GCN5thGen core architecture (Polaris), it should be a great indicator of the % gains to be had from going from GloFlo14nm to TSMC7nm.

Reviews comparing Vega64 to Radeon VII show around 20% to 40% speed improvements. Once nVidia does this with RTX, Raytracing will be a lot better performing.

Back OT: Really hope that by then, the issues that plagued these 20 series GPU's are resolved. Looking forward to going RTX on a 7nm GPU, with hopefully a lower cost.

These transistors are getting so tiny, it's amazing that the yields are as good as they are...
If the past two generations of Founders edition boards from Nvidia are any indication to us, the next Nvidia video cards will either retain the RTX pricing or push even higher.

I look at it kind of like how this one contractor for low voltage is that I used before but no longer. The older he got, the more shady he became. Closing in on retirement he started pushing out of scope work, inflating projects from 500 bucks to three Grand by the time they ended. Now I use other contractors.

Nvidia sees profit loss on the horizon, so they're price gouging their consumer base now to make up for losses in the future, heavily competitive, market place where they won't be able to pull this shit anymore.

Once Intel enters the high end graphics market place the entire landscape will change and that, from a massive company like Intel, scares the shit out of them.
 
Who knows guys, we may yet see something. Not to get your hopes up, but that diagnosis from the engineering firm Kyle was using was taking a pretty damn long time because they were dissecting the GPU in detail. So, who knows, maybe they're still digging deep and/or compiling the information discovered.

It's a long shot, but hope is better than not having it. Not ready to let this one go just yet.
 
I haven't seen any recent news of new card purchase failures so that is a +.

Buuut I DID get a card with a 4 year manufacturer warranty on it... so there is that.
 
I haven't seen any recent news of new card purchase failures so that is a +.

Buuut I DID get a card with a 4 year manufacturer warranty on it... so there is that.
Yeah, I kicked my EVGA to 5 years because of the failure rate I initially encountered.
 
You're going to make me get all nostalgic thinking about BFG cards.

Yep, I went through a few BFG 280's before I finally got one that worked. They kept working with me till we found a good one.
 
I was hoping to hear about the results but I understand that if it's anything juicy it would be a bad look for Kyle to report on it after getting hired by a competitor to Nvidia so I'm resigned to not see the results. Maybe the results could be passed on to someone else to write about but if they lack industry connections it wouldn't be the same and anyone else with connections has likely signed the NDA at this point.

You're going to make me get all nostalgic thinking about BFG cards.

I still have my BFG 8800gt in the original box, nostalgia is the only reason I haven't gotten rid of it at this point.
 
So I guess this one goes unsolved...

upload_2019-4-1_17-35-32.jpeg
 
Paul or Brent could post the results..

However, I think it's a safe bet to say that it is the GPU chips themselves, possibly due to those tainted chemicals TSMC reported.

It's not the memory, cards with both micron and samsung have had issues.

It's not the PCB, as custom cards have had the issue. And if the custom cards use their own brands/supplies of the VRM's and other components, it is likely none of those parts either.

The common component is the GPU, and it affects all of the 20xx GPU's as far as we can tell.

The fact that 80% of the GPU's are fine, means the chip design can likely also be ruled out.

My Theory: TSMC had issues that affected yields very significantly.
Testing on the GPU's memory controllers was insufficient, or unable to detect the issues. Or it was just as nVidia said, 'test escape' which I understand to mean that the automated testing equipment malfunctioned and bad parts made it through. If that was true I would expect by now for all of those bad cards to be out of the sales channels. I don't know any current failure rates compared to the first couple months. For all I know the issues are no longer affecting newly made cards.

One more thing that points to the GPU's themselves is that the company doing the testing has had a tough time nailing down any other component as the cause.
 
Last edited:
Paul or Brent could post the results..

However, I think it's a safe bet to say that it is the GPU chips themselves, possibly due to those tainted chemicals TSMC reported.

It's not the memory, cards with both micron and samsung have had issues.

It's not the PCB, as custom cards have had the issue. And if the custom cards use their own brands/supplies of the VRM's and other components, it is likely none of those parts either.

The common component is the GPU, and it affects all of the 20xx GPU's as far as we can tell.

The fact that 80% of the GPU's are fine, means the chip design can likely also be ruled out.

My Theory: TSMC had issues that affected yields very significantly. Testing on the GPU's memory controllers was insufficient

Kyle would have to give them the results before they give anything to us and that would still cause the same problems.
 
Kyle would have to give them the results before they give anything to us and that would still cause the same problems.


No he wouldn't. He could just say to the engineering firm he hired "I can't be a part of this anymore, and these individuals have agreed to take up this task in my place and pay for everything that hasn't already been paid for. So get in touch with them to share your findings. They'll do whatever they feel is the right thing to do with whatever information you give them."

Granted, people will still be free to assume that Kyle is puppeteer-ing from the backstage to shill for Intel, but the types that will do that would be the types that would have done that regardless.
 
No he wouldn't. He could just say to the engineering firm he hired "I can't be a part of this anymore, and these individuals have agreed to take up this task in my place and pay for everything that hasn't already been paid for. So get in touch with them to share your findings. They'll do whatever they feel is the right thing to do with whatever information you give them."

Granted, people will still be free to assume that Kyle is puppeteer-ing from the backstage to shill for Intel, but the types that will do that would be the types that would have done that regardless.
There's a really good chance his life was under a microscope hiring into Intel. Background checks these days will dig through social media and even these forums and pull relevant data. Even better chance that Intel flagged him for what he was working on and told him to drop it. Distance himself from it and not speak another word about it. Background checks are all sorts of invasive these days. You're best bet for getting hired anywhere is to place all your online accounts into a deep freeze / shut down state and then start applying for jobs. I know this because my boss even commented on how the background checking firm couldn't find any online social presence for me when I was hiring in. ;) That's because I deep sixed damn near anything where I had said anything online. Fired those accounts back up well after I got the job. Now, in Kyle's case, he couldn't instantly shutter the [H] not 20 years of work... It's part of his resume. So, the Intel decision is likely why we won't see anything here.

When all this shit cools down, when Nvidia pulls this crap again... I wouldn't be surprised to see an article come out of some other journalist with inside information about the 20 series and the whatever bullshit Nv is pulling series, from an anonymous source. Who knows, I still have a couple fragments of hope the story about this series could trickle out at some point. It's a bummer, but nothing we can do about it.
 
Last edited:
I like how he taps a screw driver to the bare GPU dies and the CPU socket pins hahaha

I cringed. Even knowing the boards were "faked" (very well built, too, btw), it still caused me to flinch.
 
How did he make those... must be the same guy that made the BitchinFast3dfx2000 or whatever it was called... He likes to say "so.. yeah.."

Setting them on fire, and then peeing on it was a nice touch.

Lol.
 
Back
Top