Amazon's New World is bricking RTX 3090 graphics cards

[silentbogo]

yea? Furmark comes to mind.

Furmark

Furmark only kills shitty or already heavily abused cards. I've only stopped using it for testing, cause it's old and uses wa-a-a-y too little VRAM. Superposition or Timespy work much better.
On the other hand, it's not the first time EVGA has released a poorly-designed card. GTX 10-series comes to mind, where most 1060s had abysmal VRM that liked to overheat, 1070s had an issue with VRAM power circuit design (VRM would usually burn the hole on the edge of the PCB) etc. etc.
Also quite fresh in my mind ASUS Cerberus GTX1070Ti, which had no VRAM cooling at all (just an ambient breeze). Lots of those died a long and painful death even without super-intensive loads and overclocking. Had one just a couple of months ago: giant f#$%ing triple-slot HSF that screams "Gaming Powa-a-a-a", while in reality it weighs less than some of my old ITX-friendly cards and doesn't do shit for actual cooling. Memory had to be downclocked to below stock in order to prevent artifacts from happening (and this was the "new-old stock", barely 6mo out of the box).

 

[R-T-B]

yea? Furmark comes to mind.

Furmark

TPU literally tests cards with furmark. It applies equally there: it should not be able to kill the card.

 

[MentalAcetylide]

I wonder if this happened to anyone that purchased an RTX A5000 when they couldn't get an RTX 3090. I can only imagine how pissed they would be if this happened to them with an A5000 or A6000...

 

[ThrashZone]

Furmark only kills shitty or already heavily abused cards. I've only stopped using it for testing, cause it's old and uses wa-a-a-y too little VRAM. Superposition or Timespy work much better.
On the other hand, it's not the first time EVGA has released a poorly-designed card. GTX 10-series comes to mind, where most 1060s had abysmal VRM that liked to overheat, 1070s had an issue with VRAM power circuit design (VRM would usually burn the hole on the edge of the PCB) etc. etc.
Also quite fresh in my mind ASUS Cerberus GTX1070Ti, which had no VRAM cooling at all (just an ambient breeze). Lots of those died a long and painful death even without super-intensive loads and overclocking. Had one just a couple of months ago: giant f#$%ing triple-slot HSF that screams "Gaming Powa-a-a-a", while in reality it weighs less than some of my old ITX-friendly cards and doesn't do shit for actual cooling. Memory had to be downclocked to below stock in order to prevent artifacts from happening (and this was the "new-old stock", barely 6mo out of the box).

Hi,
Most cards need custom fan curves, default is usually annoying anyway turning off and on.....
I wasn't on air very long before I switched to a water block on my 1080ti ftw3 but can say thermal pads and paste evga uses is not very good
Pads are more like greasy double stick tape and the paste they used on my ftw3 was already dry and cracking only after a couple months of benchmarking lol

 

[delshay]

Can someone point me to a thread where many cards are failing or is it just a small number of cards going kaput. EDIT: Never mind found a few threads, reading through them.

 

[silentbogo]

Can someone point me to a thread where many cards are failing or is it just a small number of cards going kaput.

AFAIK only two

Most cards need custom fan curves, default is usually annoying anyway turning off and on.....

Fan curves have very little to do with "bad" cards I mentioned. On EVGA you have a PCB track handling more current than it was designed to(right on the edge, where you barely even get ambient airflow), and as a consequence having higher thermal stress, hence it does not happen immediately, unless you mine ETH with maxed-out memory clocks. With Cerberus it's even worse - you don't need overclock, just a non-ACd room during hot summer and 30+ minutes of WoT. Memory becomes highly unstable even at stock clocks and fans blasting 100%.

Pads are more like greasy double stick tape and the paste they used on my ftw3 was already dry and cracking only after a couple months of benchmarking lol

Their thermal pads are OK from the performance perspective, but I know the thing you are talking about. Over time they start bleeding snot all over the PCB, which in turn becomes a supermagnet for dust bunnies and wooly mammoths. A real nightmare to clean up. Had a GTX1070 SC in my workshop not too long ago. I had to use carb cleaner and give it 3 runs of ultrasonic IPA bath to make it shine again. Replaced everything with Arctic pads, squeezed some MX4 onto the GPU and called it a day. Maybe not the best performing compounds, but they are consistent and long-lasting.

 

[The red spirit]

Nah I dont play games. Im mining with them currently and benchmark occasionally.

RX 580s were known to be great miners in 2018. They were most desirable cards for mining along with Vegas. You can tweak their vBIOS too. They are pretty fun to mess with.

 

[delshay]

I did some reading/watching videos & this is what I found out on one user card. .

One user says only when he removed the 8 pin PCI power cord the computer came on. So we already know there is a short across the 12v rail at the GFX card end. Please note the fault may appear different on other users cards & it's hard to tell how many user(s) are affected, but the numbers seem small to me. There are user's out there that did not have any problems when using the same GFX card, but stopped using the game after reading about the problem.

Anyway a short has appeared across the 12v 8 pin (GFX card end) after playing "New World".

 

[VeqIR]

AFAIK only two

Fan curves have very little to do with "bad" cards I mentioned. On EVGA you have a PCB track handling more current than it was designed to(right on the edge, where you barely even get ambient airflow), and as a consequence having higher thermal stress, hence it does not happen immediately, unless you mine ETH with maxed-out memory clocks. With Cerberus it's even worse - you don't need overclock, just a non-ACd room during hot summer and 30+ minutes of WoT. Memory becomes highly unstable even at stock clocks and fans blasting 100%.

Their thermal pads are OK from the performance perspective, but I know the thing you are talking about. Over time they start bleeding snot all over the PCB, which in turn becomes a supermagnet for dust bunnies and wooly mammoths. A real nightmare to clean up. Had a GTX1070 SC in my workshop not too long ago. I had to use carb cleaner and give it 3 runs of ultrasonic IPA bath to make it shine again. Replaced everything with Arctic pads, squeezed some MX4 onto the GPU and called it a day. Maybe not the best performing compounds, but they are consistent and long-lasting.

Actually MX4 is not great for a GPU due to low heatsink mounting pressure and effectively “pump out” over time. Use denser compounds, like Thermalright TFX or SYY-157.

 

[GerKNG]

Actually MX4 is not great for a GPU due to low heatsink mounting pressure and effectively “pump out” over time. Use denser compounds, like Thermalright TFX or SYY-157.

i agree MX4 turns into silicon oil soup on a GPU.
but MX5 is fantastic.
it is now almost 100 Days on my 6900XT and performs like on day one (and it is a couple °C better than my Noctua NT H2)

 

[metalfiber]

Let's spread the joy.

 

[MentalAcetylide]

Let's spread the joy.

View attachment 209348

Something is up, and we have a possible common denominator, but there's not enough evidence to definitively point fingers. I hope all of those reporting failures and shut downs are also submitting full system specs & any pertinent OC settings that were being used.

 

[Shrek]

A card should be able to survive what is thrown at it; I would not blame the game.

 

[Grog6]

I wpuld bet they're maxing the power supply; if you saturate the inductors, they act like a piece of wire.
You can even go into 'shoot thru, where the 12V shows up on the output.
That's usually when you hear the pop.

 

[Shrek]

A design should not allow an inductor to saturate but rather stop before getting to that point.

 

[Mussels]

I did some reading/watching videos & this is what I found out on one user card. .

One user says only when he removed the 8 pin PCI power cord the computer came on. So we already know there is a short across the 12v rail at the GFX card end. Please note the fault may appear different on other users cards & it's hard to tell how many user(s) are affected, but the numbers seem small to me. There are user's out there that did not have any problems when using the same GFX card, but stopped using the game after reading about the problem.

Anyway a short has appeared across the 12v 8 pin (GFX card end) after playing "New World".

did they show what cable they used?
My 3090 undervolted, burned and melted an extender - if this is people burning out cables they may not even realise the GPU's are fine and just the aftermarket cables died

 

[delshay]

did they show what cable they used?
My 3090 undervolted, burned and melted an extender - if this is people burning out cables they may not even realise the GPU's are fine and just the aftermarket cables died

Contact the user of this video directly & ask, but it looks like cables are ok as he swapped out the faulty card..

New World Bricked My RTX 3090 - YouTube

 

[The red spirit]

I wpuld bet they're maxing the power supply; if you saturate the inductors, they act like a piece of wire.
You can even go into 'shoot thru, where the 12V shows up on the output.
That's usually when you hear the pop.

With RX 590s? It's the worst Polaris card, but it should be rather safe from power fluctuations.

 

[erocker]

This is a 3070 Ti with limiting frames to 60, in the menu. That's still drawing a lot of power.

 

[Hachi_Roku256563]

With RX 590s? It's the worst Polaris card, but it should be rather safe from power fluctuations.

its really similar to a overclocked 580 it runs hot and heavy
i think its cards that just get to hot and work to hard being pushed parst limits

 

[The red spirit]

its really similar to a overclocked 580 it runs hot and heavy
i think its cards that just get to hot and work to hard being pushed parst limits

Heat so fast doesn't kill. And while it consumed quite a lot of power, that was still not as bad as Vegas. Those aren't reported yet.

 

[R-T-B]

Maybe not the best performing compounds, but they are consistent and long-lasting.

I don't know about the Thermal Pads, but the paste is bulk order Shin Etsu, which is at least ahead of most subpar OEM pastes.

 

[DeathtoGnomes]

should not

key words here.

 

[R-T-B]

key words here.

Yeah. I think you know my point: Don't buy cards that can't run furmark, or otherwise explode. If you do, RMA them with extreme prejudice.

 

[Falkentyne]

Furmark only kills shitty or already heavily abused cards. I've only stopped using it for testing, cause it's old and uses wa-a-a-y too little VRAM. Superposition or Timespy work much better.
On the other hand, it's not the first time EVGA has released a poorly-designed card. GTX 10-series comes to mind, where most 1060s had abysmal VRM that liked to overheat, 1070s had an issue with VRAM power circuit design (VRM would usually burn the hole on the edge of the PCB) etc. etc.
Also quite fresh in my mind ASUS Cerberus GTX1070Ti, which had no VRAM cooling at all (just an ambient breeze). Lots of those died a long and painful death even without super-intensive loads and overclocking. Had one just a couple of months ago: giant f#$%ing triple-slot HSF that screams "Gaming Powa-a-a-a", while in reality it weighs less than some of my old ITX-friendly cards and doesn't do shit for actual cooling. Memory had to be downclocked to below stock in order to prevent artifacts from happening (and this was the "new-old stock", barely 6mo out of the box).

Furmark GREATLY downclocks the card massively.
At 400W Power limit, a 3090 or 3080 card that would normally run at 1920 mhz would run at 1300 mhz in furmark.

You can destroy the card if it doesn't.
Proof? Try renaming furmark.exe to Quake3.exe or UnrealTournament.exe
Don't blame me for what happens.