RTX Space Invaders Wanted

So we got bumped back in queue for a bit. We are not exactly priority with our huge one-card job. :) BUT! Good news so far.

Hey Kyle, they are moving along slowly due to some other priorities in the equipment queue.

I can confirm that we can consistently reproduce the corruption – so it is definitely not a system/user isolated issue. I am optimistically hoping that we can confirm it is not a thermal-sensitive issue by end of week. I will check in with you then.

At least we know it is not just Admiral_Ackbar6's problem.
 
thats hot - will smith,jpg

Ouch, a vid card recall in the near future?
 
Joust said:
My bet: he thinks he knows what's wrong. Needs to be able to repeat the cure.
Click to expand...

Drool! Hope he's got it figured out and holds the cure out to NVIDIA.....then snatches it away (or charges a hefty fee)! ;)
 
Maybe faulty memory chips that test fine but aren't really fine. Bad test procedures or maybe the memory itself is defective by design.
 
First of all, what's up with the anti nVidia and Intel vibe here on HardOCP? ... I wonder if people have forgotten that nVidia and Intel have the highest performing PC parts on the market? Between both of my accounts, going on 18+ years, I've never seen this before.

With that said, my MSI Gaming X Trio 2080 Ti is kicking ass and taking names.

I don't see space invaders, but I do see insane benchmarks. Hahaha. I mean, there might be space invaders or something on the screen but at 200fps @ 1440p ... shit is just moving too buttery smooth and fast for me to really notice. In fact, my FPS is so great that I fear I might go into a photosensitive epilepsy seizure anytime I load a game up. Hahah. If any of you out there share the same concern, get an AMD video card, you will never ever face that risk with one of those jobbies.

On a side note, for those of you wanting an nVidia BFGD, Samsung already has several models already on the market starting with a 55". They do 120hz native at 1440p and support Freesync and Freesync Ultimate. Another cool thing is, the only video card in the world that can handle this, the nVidia RTX 2080 Ti can very easily push High and Ultra settings at 120+ FPS @ 1440p.

Really hoping AMD can get it's act together in the high-end video card dept tho, for realz. I would love to have awesome performance that truly matches or rivals nVidia's best efforts. I hope that happens but I'm not counting on it.

BTW, as a responsible person who is passionate about my hobby, I've already started to save money for the $1200+ I will need to buy the next nVidia "Worlds Most Powerful Video Card" in 2020, whatever that model number will be called. And it's easy, $50 dollars a month off to the side. I do this by not swelling bawls energy drinks and stuffing my mouth full of Cool Ranch Doritos or buying anime action figures or a lot of other bs nonsense. I tell myself, I don't want to be that kid who is crying at how expensive these video cards are and being too broke to afford one when the time comes.

Yes, yes it's true ... Ferrai's and Lambo's do cost a pretty penny. But, you get what you pay for. Vroom! Vroom!
 
Last edited:
So for future reference, when people say that this is normal for this many things to fail we should tell to STFU?
 
SixFootDuo said:
First of all, what's up with the anti nVidia and Intel vibe here on HardOCP? ... I wonder if people have forgotten that nVidia and Intel have the highest performing PC parts on the market? Between both of my accounts, going on 18+ years, I've never seen this before.
Click to expand...

I agree with you to some extent, but if you've been following what's going on with the Intel hardware exploits and the nVidia boondoggles recently, it kinda makes sense.

That said, like you mentioned, they still have the highest-performing parts currently, so...yeah.
 
Ehren8879 said:
Kyle, have you seen this video from Tech YES City?



It's not quite a deep analysis of the problem, but it implicates Micron GDDR6 as a major source of failed RTXs
Click to expand...


That looks like a lot of guessing and not a lot of validation. He spent most of the video talking about how he was going to take the card apart so he could blame the RAM, with no evidence the RAM failed...
 
Did not forgot! Still waiting!

Maybe someone competent release information about this issue.
Because it seems over all the reviewers/influancers panel we dispose at the moment, no one got any follow into this story.
 
BoiseTech said:
I understand people don't "know" you. However if my Gigabyte 2080ti experiences an issue I won't hesitate to give you first dibs on it.
Click to expand...
He got his full purchase price and shipping paid directly without fees the moment he gave me a tracking number. :) Actually I think I gave him more than his cost IIRC. $1200.00
 
I remember getting a saphire geforce back in the day with massive artifacting. I took the heatsink off and discovered they had forgot to remove the plastic insert from the TIM material. After cleaning it off and adding a fresh coat of compound it worked like a champ, turned out to be a great clocker too.
 
Ocellaris said:
That looks like a lot of guessing and not a lot of validation. He spent most of the video talking about how he was going to take the card apart so he could blame the RAM, with no evidence the RAM failed...
Click to expand...

He did test the RAM at a slower clock speed and it was fine, but at stock it failed. So in that sense, it did seem related to RAM. Taking it apart seemed pointless other than to confirm that it was Micron RAM, as he suspected.
 
SixFootDuo said:
First of all, what's up with the anti nVidia and Intel vibe here on HardOCP? ... I wonder if people have forgotten that nVidia and Intel have the highest performing PC parts on the market? Between both of my accounts, going on 18+ years, I've never seen this before.

With that said, my MSI Gaming X Trio 2080 Ti is kicking ass and taking names.

I don't see space invaders, but I do see insane benchmarks. Hahaha. I mean, there might be space invaders or something on the screen but at 200fps @ 1440p ... shit is just moving too buttery smooth and fast for me to really notice. In fact, my FPS is so great that I fear I might go into a photosensitive epilepsy seizure anytime I load a game up. Hahah. If any of you out there share the same concern, get an AMD video card, you will never ever face that risk with one of those jobbies.

On a side note, for those of you wanting an nVidia BFGD, Samsung already has several models already on the market starting with a 55". They do 120hz native at 1440p and support Freesync and Freesync Ultimate. Another cool thing is, the only video card in the world that can handle this, the nVidia RTX 2080 Ti can very easily push High and Ultra settings at 120+ FPS @ 1440p.

Really hoping AMD can get it's act together in the high-end video card dept tho, for realz. I would love to have awesome performance that truly matches or rivals nVidia's best efforts. I hope that happens but I'm not counting on it.

BTW, as a responsible person who is passionate about my hobby, I've already started to save money for the $1200+ I will need to buy the next nVidia "Worlds Most Powerful Video Card" in 2020, whatever that model number will be called. And it's easy, $50 dollars a month off to the side. I do this by not swelling bawls energy drinks and stuffing my mouth full of Cool Ranch Doritos or buying anime action figures or a lot of other bs nonsense. I tell myself, I don't want to be that kid who is crying at how expensive these video cards are and being too broke to afford one when the time comes.

Yes, yes it's true ... Ferrai's and Lambo's do cost a pretty penny. But, you get what you pay for. Vroom! Vroom!
Click to expand...
Good to know someone else happy their purchase. Lucky you...

As far as the Intel / Nvidia hatred is concerned... I don't hate either , I run both. Though I'm on my 3rd 2080Ti that may very well be starting to manifest issues more frequently with a system lockup just this past evening.

Some people are here to stick it to Nvidia to be certain, however, I like to believe that the rest of us are more interested in getting to the bottom of why these video cards are failing. I just want my damn card to run flawlessly... Hopefully this testing sheds some light on what the hell is going on.
 
I'll be watching this very carefully, too. Am really curious what's causing this for so many cards while others seem 100% fine.
 
can't wait to see the results of these tests and what the real problem with the 2080Ti cards is.
 
Last edited by a moderator:
keep-calm-and-stay-on-topic-2.png
 
So it's been some time since all the 2080Ti issues came to light. Primarily heard about issues with the Founders Edition cards, and that the Micron memory was a suspect, and a cause was "test escapes", or as I understood it, failures in an automated card testing or component testing process that was just passing everything, so bad parts made it through and on to customers.

Have we hard anything more from nVidia about it? What has been the failures rates after the issue was discovered and corrected? i.e. on cards manufactured after the testing process was fixed? We heard on the forums a few months ago someone who had 3 I think it was, bad cards in a row. Whatever happened with that?
 
MavericK said:
He did test the RAM at a slower clock speed and it was fine, but at stock it failed. So in that sense, it did seem related to RAM. Taking it apart seemed pointless other than to confirm that it was Micron RAM, as he suspected.
Click to expand...

Sure, something RAM related failed. That could be a memory controller in the GPU die, power circuitry to the RAM, or the RAM itself... Heck it could even be a trace issue on the PCB wafer. And he could have looked up the RAM type in GPUZ and not taken the card apart like a noob to stretch out his video.
 
Ocellaris said:
Sure, something RAM related failed. That could be a memory controller in the GPU die, power circuitry to the RAM, or the RAM itself... Heck it could even be a trace issue on the PCB wafer. And he could have looked up the RAM type in GPUZ and not taken the card apart like a noob to stretch out his video.
Click to expand...
Bingo! Just because you can downclock the VRAM and "fix it" hardly means that the actual VRAM is the issue. Hopefully we can find something out with PROOF.
 
knowom said:
I think it's TSMC chemical related at this point given recent events. Sure it's probably in part memory related as well. More specifically Micron's rather than Samsung's because the IC's are probably more closely boarder line in terms of headroom and safeguarding from stuff like a power surge/spike for example wrecking these poor RTX GPU's. In this case teh chemicals causing premature unforeseeable degradation is likely what is going on. Perhaps though it's not exactly the memory IC's though, but rather the PCIe bus interconnects or memory controller degrading which would likely cause similarly quirky results being intertwined. In the end I think it's degradation probably from heat due to bad quality control by Nvidia and/or it's memory supplier or fabrication partner.
Click to expand...

If this is the case and who knows, that is very interesting. Hopefully the engineers that Kyle is using are able to pick that up. I have no idea if that would even be possible.
 
MavericK said:
He did test the RAM at a slower clock speed and it was fine, but at stock it failed. So in that sense, it did seem related to RAM. Taking it apart seemed pointless other than to confirm that it was Micron RAM, as he suspected.
Click to expand...

You can't draw a conclusion by downclocking VRAM. His "testing" means nothing without more information and context from a qualified hardware engineer. Kyle did the 100% right thing by having such people examine a failed card. That Brian nitwit from the "Tech Yes City" YouTube channel just spoke out of his ass because he likes the sound of his voice. Read between the lines of what he said around the 8:15 mark. I am paraphrasing his whole diatribe about free/review samples, but he essentially said that he is not going to pursue the issue with fixing his card with Gigabyte because he likes getting free stuff.

He also said it is not their (Gigabyte's) fault. I laughed out loud when he said that. If the product is faulty and it does turn out to be RAM, Gigabyte is responsible and they have to fix their product. Saying it isn't their fault is like me saying it isn't Chrysler's fault that they put crappy Takata airbags in my Dodge Charger Rallye. They could have used Autoliv, Delphi or TRW for airbags, but they elected to use cheap asian crap and now my car is unsafe if I get in a wreck and I have to wait 9-12 weeks for the parts to arrive for the recall.

It's not Gigabyte's fault. Translation: Gigabyte, don't be mad at me and keep sending me free stuff. What a hack fraud.
 
mps said:
You can't draw a conclusion by downclocking VRAM. His "testing" means nothing without more information and context from a qualified hardware engineer. Kyle did the 100% right thing by having such people examine a failed card. That Brian nitwit from the "Tech Yes City" YouTube channel just spoke out of his ass because he likes the sound of his voice. Read between the lines of what he said around the 8:15 mark. I am paraphrasing his whole diatribe about free/review samples, but he essentially said that he is not going to pursue the issue with fixing his card with Gigabyte because he likes getting free stuff.

He also said it is not their (Gigabyte's) fault. I laughed out loud when he said that. If the product is faulty and it does turn out to be RAM, Gigabyte is responsible and they have to fix their product. Saying it isn't their fault is like me saying it isn't Chrysler's fault that they put crappy Takata airbags in my Dodge Charger Rallye. They could have used Autoliv, Delphi or TRW for airbags, but they elected to use cheap asian crap and now my car is unsafe if I get in a wreck and I have to wait 9-12 weeks for the parts to arrive for the recall.

It's not Gigabyte's fault. Translation: Gigabyte, don't be mad at me and keep sending me free stuff. What a hack fraud.
Click to expand...

This ^ +1

I, too, am careful about the "influencers" I pay attention to. I mean wtf with (e.g.) the eye diagram at 2:20? Did he even know what he's looking at?

Me, I'm waiting on what Kyle puts together after he hears back from the lab.
 
Last edited:
mps said:
You can't draw a conclusion by downclocking VRAM. His "testing" means nothing without more information and context from a qualified hardware engineer. Kyle did the 100% right thing by having such people examine a failed card. That Brian nitwit from the "Tech Yes City" YouTube channel just spoke out of his ass because he likes the sound of his voice. Read between the lines of what he said around the 8:15 mark. I am paraphrasing his whole diatribe about free/review samples, but he essentially said that he is not going to pursue the issue with fixing his card with Gigabyte because he likes getting free stuff.

He also said it is not their (Gigabyte's) fault. I laughed out loud when he said that. If the product is faulty and it does turn out to be RAM, Gigabyte is responsible and they have to fix their product. Saying it isn't their fault is like me saying it isn't Chrysler's fault that they put crappy Takata airbags in my Dodge Charger Rallye. They could have used Autoliv, Delphi or TRW for airbags, but they elected to use cheap asian crap and now my car is unsafe if I get in a wreck and I have to wait 9-12 weeks for the parts to arrive for the recall.

It's not Gigabyte's fault. Translation: Gigabyte, don't be mad at me and keep sending me free stuff. What a hack fraud.
Click to expand...
While the guy in the video may be taking the wrong approach to his verdict, he isn't wrong. It ISN'T Gigabyte's fault the cards failed, because they only sourced the components from one of two manufacturers --again that's assuming RAM is the issue-- so it's odd that he wouldn't want to pursue an RMA... because it IS Gigabyte's responsibility to RMA products with their name on them. It'd be better publicity to have him show off the card failed, and Gigabyte RMAed it quickly and painlessly.
 
Teenyman45 said:
The real kicker is that the last few generations of Nvidia cards, from the 970/ 980's onward, have been at least somewhat memory bandwidth limited. Or least more obviously limited when overclocking.
Click to expand...

No they haven’t memory OC’s have made barely any difference to gaming which also increases bandwidth as a side effect.
 
I think TSMC is likely the cause of the problem inadvertently though Nvidia is certainly at least partially liable for their own quality control and testing being a bit subpar and not catching it sooner. You might say the same about memory IC suppliers too if they are teetering that close to the edge of unreliability and failure. From where I stand I wouldn't feel comfortable buying a RTX series card right now and even less so overclocking one.
Maddness said:
If this is the case and who knows, that is very interesting. Hopefully the engineers that Kyle is using are able to pick that up. I have no idea if that would even be possible.
Click to expand...
Well it does give credence towards the test escapes claims by Nvidia. It could be related to the memory IC's and bad judgement by Nvidia or it's memory supply, but the TSMC bio chemical wafer attack seems most probable culprit to me. I would certainly hope Nvidia's quality control testing would pick up on more common voltage/temperature/frequency related problems myself than unforeseeable degradation due to impurities, corrosion, or acidity or who the F knows as a result of a chemical used and whatever dodgy dark alley way supply source they bought them from to save a nickel.
 
aokman said:
No they haven’t memory OC’s have made barely any difference to gaming which also increases bandwidth as a side effect.
Click to expand...

Using Brent's and Kyle's review of the MSI GeForce RTX 2080 GAMING X TRIO as a reference, the standard average operating gaming frequency is 1980MHz and the Memory is 1,750MHZ. These are the baselines BEFORE overclocking. While the GPU frequency can only be increased to 2055MHz, a 4% overclock, memory goes to 2,000MHz, a 14% overclock. When gaming while overclocked, if memory was not a limiting feature then GPU clock increases would lead to purely linear increases in framerate, without ever touching memory speed/bandwidth.

However, the card with both GPU and memory overclocked actually runs about 7-10% faster rather than only 4% faster. At stock in-game speed, memory bandwidth was the limiting factor, likely on the order of 3-6% and while that's nowhere near how badly Vega 64 was hamstrung for lacking memory bandwidth, it's still lost performance and having to dial back memory speed (with the resulting further cut to bandwidth) will impair an RTX card's performance.
 
I still don't understand how something like this makes it passed all the checks. Unless its cheaper just to not test that much and roll the dice?
 
tunatime said:
I still don't understand how something like this makes it passed all the checks. Unless its cheaper just to not test that much and roll the dice?
Click to expand...
Well in the case of TSMC chemical wafer issue it's easier to understand how it kind of went a bit more unnoticed. The degradation might not be immediately noticed, but over time over a period of heating and cooling cycles. Outside of that without knowing their quality control standards it's sort of anyone's guess. They may have tested a initial batch of wafers that checked out fine and a later batch was effected. Both batches though may be part of the initial stock pile of GPU's they shipped since otherwise they get complaints about "paper" launches.
 
SixFootDuo said:
First of all, what's up with the anti nVidia and Intel vibe here on HardOCP? ... I wonder if people have forgotten that nVidia and Intel have the highest performing PC parts on the market? Between both of my accounts, going on 18+ years, I've never seen this before.
Click to expand...

Seriously?

I mean especially the intel thing. You shouldn't have to even question it. Intel has been dragging their feet for a while with no competition and high prices. They made a number of decisions that have come back to bite them in the ass. On top of it, AMD has, for the moment, gotten their act together regarding CPUs. This is high drama game of thrones type shit for hardware enthusiasts. It will be covered. It will look bad for intel because.... IT IS BAD FOR INTEL. If you are a hardware enthusiast and you haven't broken out your bowl of popcorn and are watching the show, something is wrong with you.

As for NVIDIA, well they started it acting like assholes. On top of that, you have a whole lot of money going into not much performance increase. There's a lot there to speculate about, and most of it is not going to look good for nvidia. It would also not look good for AMD, but you wouldn't notice that part as that discussion has been ongoing for a few years now at least.
 
You must log in or register to reply here.
Back
Top