Why does someone have to own something to have an opinion? Why does someone have to own something to wonder what the root cause of an issue is?

Why do people feel the need to say “my works fine”? Great. Thumbs up and congratulations. People loved saying that as users were having issues with their 3090FTW3’s, as if it somehow negates the fact something is happening. My brain can’t comprehend that thought process.

When I bought my car, there was a consumer report that showed the tires used on an SUV were meant for a much lighter car, so the rubber would be destroyed at an unreasonably low usage. I contacted ford and they said “it’s only a few instances, you have nothing to worry about” until I was paying $1300 for new tires at 30,000 miles, when they were rated for 70,000. But hey, other peoples were fine, so why worry?

When I spend the money on a new GPU, I prefer to know for a fact it isnt going to have a strange issue that more than a couple of people are reporting. The bigger thing to remember is 1: not everyone is checking their connector. 2: not everyone is reporting on Reddit that they have had an issue.

It is entirely reasonable for people to be curious before making a purchase. It should be expected, rather than being shrugged off or condemned.

As for JonnyGuru, I’ve only formed an opinion based on his public conduct. When he wasn’t making claims like he has recently, I didn’t have any issues with him, and actually enjoyed his website. Just not a fan right now.

the_Scarlet_one
Why does someone have to own something to have an opinion? Why does someone have to own something to wonder what the root cause of an issue is?

Why do people feel the need to say “my works fine”? Great. Thumbs up and congratulations. People loved saying that as users were having issues with their 3090FTW3’s, as if it somehow negates the fact something is happening. My brain can’t comprehend that thought process.

When I bought my car, there was a consumer report that showed the tires used on an SUV were meant for a much lighter car, so the rubber would be destroyed at an unreasonably low usage. I contacted ford and they said “it’s only a few instances, you have nothing to worry about” until I was paying $1300 for new tires at 30,000 miles, when they were rated for 70,000. But hey, other peoples were fine, so why worry?

When I spend the money on a new GPU, I prefer to know for a fact it isnt going to have a strange issue that more than a couple of people are reporting. The bigger thing to remember is 1: not everyone is checking their connector. 2: not everyone is reporting on Reddit that they have had an issue.

It is entirely reasonable for people to be curious before making a purchase. It should be expected, rather than being shrugged off or condemned.

As for JonnyGuru, I’ve only formed an opinion based on his public conduct. When he wasn’t making claims like he has recently, I didn’t have any issues with him, and actually enjoyed his website. Just not a fan right now.

Your brain can't comprehend that there are about 100k RTX 4090s out in the wild and only about 30 of them are experiencing an issue? 
 
I'm stating that my card is fine not as a way to counter anything happing, I'm stating is a matter of fact, nothing more nothing less. You seem to be caught up the mass paranoia that roughly 30 cards have experienced when by and large the vast majority of owners are having no issue what so ever. You are ignoring the fact that they are largely issue free. These things happen, things fail all the time it's part of a normal product's expected variance. There isn't a product mass released in the history of the planet that doesn't have a failure rate. It's commonly accepted that a failure rate of about 1% is acceptable. Do you know what the failure rate is for 30 bad cards out of 100,000? It's 0.03%. This whole thing is just drama and you and most others are eating it up. 

The claimed 30 may be a small percentage, but it is abnormal and abnormally rapid. We have seen a handful in the past, but it was sporadic and usually way late in the product's life.
It may be a small number, but it is a significant indicator of something abnormal.

I've heard over and over that 100k 4090s have been sold and are out there being used.
I thought that info was being misinterpreted and that it was only the chips that Nvidia sent to AIBs.
 
I was right. Here's a quote from Guru3D:  https://www.guru3d.com/news-story/nvidia-produced-over-100000-rtx-4090-units-thus-far,4.html
 
Note that 100,000 units do not equal the total number of GPUs made or sold. Instead, it is the number of chips that were sent to partners so they could make RTX 4090 models.
 
As far as I can tell we don't know how many built 4090s have been sold. At least, I can't find the data.

ty_ger07
The claimed 30 may be a small percentage, but it is abnormal and abnormally rapid. We have seen a handful in the past, but it was sporadic and usually way late in the product's life.
It may be a small number, but it is a significant indicator of something abnormal.

While I agree this is abnormal, there really isn't a time table on how these things can happen especially when the issue can't be recreated despite many efforts in a controlled setting. They have tried to compromise these cables, they have tried to get them to melt to demonstrate it happening and maybe better understand it or how to prevent it. They cannot, so all we are left with is roughly 30 reports and here-say or anecdotal evidence from the end users. They are definitely not going to own up to doing anything wrong and potentially risk a warranty replacement. However as I've said, if they've intentionally damaged these cables, broken solder joints and overloaded the cables and still cannot get them to melt. The cable most likely isn't the issue. I'm starting to agree this is mostly user error and not plugging in the cable all the way. The way the cable is designed has all but been ruled out as a cause for failure with what these reviewers have subject them to. What else is left really? What reasonable assumption would you make?  
 

frankd3
I've heard over and over that 100k 4090s have been sold and are out there being used.
I thought that info was being misinterpreted and that it was only the chips that Nvidia sent to AIBs.
 
I was right. Here's a quote from Guru3D:  https://www.guru3d.com/news-story/nvidia-produced-over-100000-rtx-4090-units-thus-far,4.html
 
Note that 100,000 units do not equal the total number of GPUs made or sold. Instead, it is the number of chips that were sent to partners so they could make RTX 4090 models.
 
As far as I can tell we don't know how many built 4090s have been sold. At least, I can't find the data.

Sure, that's fair but we can assume it's a very large number of cards out in the wild. Even if it were only half of the estimated 100k GPUs that were sold through to users that would only make the 30 cases a 0.06% failure rate. Hell, even if it were only a quarter of the 100k units sold through to user that would make it a 0.12% failure rate. 

How many times, in the past, has NVidia told AIB’s to send their own cards back to NVidia for investigation when something comes up with a product?

You think users wouldn’t admit to having the cable loose, yet we have people come onto the very forums complaining that their Hydrocopper cards were hit 100-105c and shutting down and they couldn’t figure out why, when they had never even realized that the card required external cooling. Yeah, people admit things, usually inadvertently. It’s a little strange that every user reporting the issue swears it wasn’t them, when someone typically would admit their fault even if it was just a slip up, because not everyone is a terrible person.

Crap, I’m not sure why you are so severely offended by me posting. You seem absolutely hell bent to prove me wrong about something or shut me down for some reason. You do you. Have fun, pour your heart all over the keyboard. Get it all out. Just know that it doesn’t change my curiosity even .01%. It also won’t change my opinion that NVidia could be leveraging the power at the PCI connection either. So far your card is an outlier when looking at 4090 GPU-Z sensor tabs that i can find through google, and you are pulling 2 to 3 times more wattage than most users that have posted screen shots. Most sensor tab screenshots I can find seem to hover between 7w and 17w.

CraptacularOne
 
Sure, that's fair but we can assume it's a very large number of cards out in the wild. Even if it were only half of the estimated 100k GPUs that were sold through to users that would only make the 30 cases a 0.06% failure rate. Hell, even if it were only a quarter of the 100k units sold through to user that would make it a 0.12% failure rate. 

Hey now, assuming the "30" that were posted about are all the failures is a bit of a stretch, isn't it? That's just the people who bothered to post on reddit about it. Apparently Gamer's Nexus's email opened for this issue had hundreds of submissions, and even then that wouldn't be all of them. Quite a few folks with this issue, maybe even the majority, would just return the card for a refund or RMA and we'd not hear about it. There's even a good chance there are quite a few folks out there with melted adapters that aren't even aware of it.
 
Too many unknowns to be certain.
 
And it's very interesting that these 4090s have such a low PCIE slot power draw - coming from someone who had a 3090 FTW3 Ultra that drew over 93W on the PCIE slot.

the_Scarlet_one
How many times, in the past, has NVidia told AIB’s to send their own cards back to NVidia for investigation when something comes up with a product?

You think users wouldn’t admit to having the cable loose, yet we have people come onto the very forums complaining that their Hydrocopper cards were hit 100-105c and shutting down and they couldn’t figure out why, when they had never even realized that the card required external cooling. Yeah, people admit things, usually inadvertently. It’s a little strange that every user reporting the issue swears it wasn’t them, when someone typically would admit their fault even if it was just a slip up, because not everyone is a terrible person.

Crap, I’m not sure why you are so severely offended by me posting. You seem absolutely hell bent to prove me wrong about something or shut me down for some reason. You do you. Have fun, pour your heart all over the keyboard. Get it all out. Just know that it doesn’t change my curiosity even .01%. It also won’t change my opinion that NVidia could be leveraging the power at the PCI connection either. So far your card is an outlier when looking at 4090 GPU-Z sensor tabs that i can find through google, and you are pulling 2 to 3 times more wattage than most users that have posted screen shots. Most sensor tab screenshots I can find seem to hover between 7w and 17w.

I'm not at all offended by your posts, everyone is entitled to their own opinion. I'm looking at this from a numbers point of view and in that respect things are certainly not anywhere near as bleak as you or others are making it out to be. As for this being the first time Nvidia requested cards to be sent back no, Nvidia did in also request early samples of 30 series cards to be returned to them for inspection when the whole POScap hysteria was running rampant in the beginning, same thing when some GPUs were dying while playing New World and that was EVGA and Nvidia wanting them back for an autopsy. So Nvidia wanting a closer look isn't really anything new. 
 
As for the cards PCIe power draw we just don't have a big enough sample size to make any determination, but I think this is largely irrelevant too. In comparison the theoretical max is a drop in the bucket and wouldn't really alleviate any real strain when we are talking about a connector that can deliver 600w. 

arestavo

CraptacularOne
 
Sure, that's fair but we can assume it's a very large number of cards out in the wild. Even if it were only half of the estimated 100k GPUs that were sold through to users that would only make the 30 cases a 0.06% failure rate. Hell, even if it were only a quarter of the 100k units sold through to user that would make it a 0.12% failure rate. 

Hey now, assuming the "30" that were posted about are all the failures is a bit of a stretch, isn't it? That's just the people who bothered to post on reddit about it. Apparently Gamer's Nexus's email opened for this issue had hundreds of submissions, and even then that wouldn't be all of them. Quite a few folks with this issue, maybe even the majority, would just return the card for a refund or RMA and we'd not hear about it. There's even a good chance there are quite a few folks out there with melted adapters that aren't even aware of it.
 
Too many unknowns to be certain.
 
And it's very interesting that these 4090s have such a low PCIE slot power draw - coming from someone who had a 3090 FTW3 Ultra that drew over 93W on the PCIE slot.

We don't know we can only go by what has been claimed or posted about. We have no way of knowing otherwise and must go with what we know and really that's all that we know as of now. Also if your RTX 3090 was drawing 93w from thew PCIe slot you need to RMA it, something is not right with it's power delivery and load balancing. It should never draw that much from a slot that is only rated for a theoretical max of 75w. Of all the cards that I own (and believe me I own quite a few) I haven't seen any of them pull more than 66-70w at max and only for a few seconds. 

I don't think the low slot power readings are all that unusual.
With my 3090Ti, as I kept pushing the overclock higher (clock, memory and voltage) to get a better Timespy score I noticed that the board power went up and the slot power went down with each adjustment.
 
Then I ended up with my second best Timespy record:

and the slot power was at it's all time lowest 8.1W and board power at it's highest 506.3W (using the XOC bios)
However, normal non-oc gaming the slot power is always in the 50's.
 
I didn't know what to make of that so I never said anything.
But maybe the 4090 overclocks have something to do with the low slot power?

Interesting
 
4090 Asus TUF. all stock, GPUZ shows its drawing 1.050 max I guess I need another bios for 1.1?
 
PCIE Slot Power
 
At IDLE 6.4
Just setting fans 90% it goes to 10
I have yet to see it go over 16 at all. This was with 450 watts total draw in GPUZ and +175 core. Porte royal.