Dell RTX 2070 problem

IdentityCrisis

Limp Gawd
Joined
Oct 29, 2012
Messages
258
So my neighbor has an Alienware Aurora that came with an OEM Dell RTX 2070. Randomly one day we stepped away from his pc while running CIV 5, when he came back, his framerates were bad and the system was running choppy. Upon reboot everything seemed fine, but his performance in games was awful, stuck at low FPS. I eventually swapped his 2070 for a 3060ti I had in a spare machine, and he gave me the RTX 2070 in return, because it seems to work as long as its not being used for 3d applications.

I installed it in a system with a 5960x and started to mess around with it.

Turns out the GPU isn't boosting to its rated speeds when I run timespy. Average GPU Speed during the runs is 680 mhz when it should be 1860 or so, and the temp was around 40*. My score was also about 4200

I updated the firmware from dell, along with the drivers, and the result is the same. I am not sure where to go from here, I see no errors in the Device Manager, and it seems stable at desktop. It will usually complete a 3dmark Timespy run if i leave it at stock settings. I think I tried tweaking voltages a bit, and it crashed during the run

What else can I look for/test to see if it is repairable?
 
It doesn't report high temps. it runs in the mid 40's or 50s during the time spy run. The clock speed doesn't stay boosted, I can't recall GPU usage, but i'm fairly certain it was low.
 
download and run ddu and then install the new drivers from nvidia not dell. the you could try afterburner and ticking all the boxes in the options for power and stuff.
 
The drivers were obtained from nvidia, using the clean install function, I'll try ddu, but this has carried over between 2-3 PCs now, so It is unlikely to be a driver issue.

I was going to tweak voltages further, than I did, but noticed just a slight change caused 3dmark to crash during it's run. Ill mess around with it some more and provide more information. It certainly seems like something failed on the card since the problem travels with it. When I installed the 3060ti in his PC, it ran perfectly with a driver update, so I am sure it is not his computer.
 
almost sounds like its got low power. I've seen weirdness like that with cards that work even if you forget to plug them in or even half plug them in.
 
almost sounds like its got low power. I've seen weirdness like that with cards that work even if you forget to plug them in or even half plug them in.
This particular card only has 1 power connector and has exhibited it across 3 systems and 3 power supplies. I wonder if an on board fuse is bad?
 
The obvious question would be is the computer still under warranty from Dell? Make them replace it.
 
The obvious question would be is the computer still under warranty from Dell? Make them replace it.
It is not under warranty, and my neighbor would have had to shell out $450 to reinstate the warranty and ship the whole PC back to Dell for service, which in my past experiences is hit or miss as best.

He opted to buy a 3060ti off of me for MSRP, and gave the card to me since I wouldn't have a GPU for my x99 system.

I spent time last night messing with it. I eventually maxed the voltage limit % to what Afterburner would let me, I think 114%? and then set the gpu core to 200mhz... and the card worked! My score jumped from 4300 to around 8800 for an RTX 2070. It was on the hot side, but it was stable and didn't crash once in the 8 or so runs I did last night.

I slowly backed off the voltage limit back to 100% and eventually removed the boost clock increase, and now the card is giving me consistent 8700ish 3dmark scores with the profile reset in after burner. I then took the card apart and repasted it which lowered my temps by 7-8 degrees. It seems to be fine, but I didn't really do anything to "Fix" it...

What could running a couple of Timespy runs and tweaking the cards profile have done?
 
If it's degradation, jolting it with more power might have temp fixed it, but it'll eventually happen again.

Unfortunately, Dell does itself no favors with their case designs/thermals when it comes to part life.
 
If it's degradation, jolting it with more power might have temp fixed it, but it'll eventually happen again.

Unfortunately, Dell does itself no favors with their case designs/thermals when it comes to part life.

That's what I figure as well, he left CIV V running and stepped away from his PC, which is on the floor near a gas fireplace. Not an ideal environment, and definitely not in one of those aurora cases, where the components are crammed in Tetris style.

The thermal paste was dry and crusty, but I don't see how that could have been the issue since initially I wasn't even getting load or much temp increase on the GPU itself
 
This particular card only has 1 power connector and has exhibited it across 3 systems and 3 power supplies. I wonder if an on board fuse is bad?
If it's a reference board, or one of those super cheap OEM blower-type ones, I'd be shocked if this card has any fuses on it.

One possibility here is that, for reasons unclear, communication on the I2C bus that the GPU uses to communicate with the power regulation and RGB disco lights got interrupted. The symptoms you describe are very similar to what happens when the I2C bus stops responding to the GPU - it reverts to maxing out at the highest known-safe clock speed.

Can you post some photos of the board? Is there any sign of corrosion or physical damage, or tons of dust caked on it anywhere?
 
If it's a reference board, or one of those super cheap OEM blower-type ones, I'd be shocked if this card has any fuses on it.

One possibility here is that, for reasons unclear, communication on the I2C bus that the GPU uses to communicate with the power regulation and RGB disco lights got interrupted. The symptoms you describe are very similar to what happens when the I2C bus stops responding to the GPU - it reverts to maxing out at the highest known-safe clock speed.

Can you post some photos of the board? Is there any sign of corrosion or physical damage, or tons of dust caked on it anywhere?
This card has no RGB inside it, its a fairly plan Dual Fan Design with a single 8 Pin power connector.

I opened it up last night to repaste after it seemed to be revived. and the board was pretty clean, no major dust, fins were all clean, and my untrained eye didnt see anything obvious in the way of burnt or damaged components. But this was after someone else worked on it prior to it getting into my hands, So I don't know the original state.

I have attached a picture of the card.
 

Attachments

  • 20200116_110313bb.jpg
    20200116_110313bb.jpg
    287.7 KB · Views: 1
Whoever the AIB is for Dell on some of their cards, like that one, that's actually a decent cooler assuming you fix the pad/paste contact issues.
 
Whoever the AIB is for Dell on some of their cards, like that one, that's actually a decent cooler assuming you fix the pad/paste contact issues.
Thats good to know, I mean after repaste and testing under timespy the card dropped a good 7-8c under load and at stock voltage and frequency settings.

Ill have to look up the issues you're talking about, I've never dealt with fixing those issues and probably should for the health of my 3080 Gaming Z
 
This card has no RGB inside it, its a fairly plan Dual Fan Design with a single 8 Pin power connector.

I opened it up last night to repaste after it seemed to be revived. and the board was pretty clean, no major dust, fins were all clean, and my untrained eye didnt see anything obvious in the way of burnt or damaged components. But this was after someone else worked on it prior to it getting into my hands, So I don't know the original state.

I have attached a picture of the card.
It may not have the LEDs, but it probably still has some vestigial circuitry from them, but the important thing is that the I2C bus that gets used for comms with the disco light controller (which is frequently on the cooler, not the board) is also used for comms between the GPU and power monitoring. This was why bent pins in the reference 2080 and 2080 Ti fan connectors could cause the card to fail to boost. I doubt that's exactly what's happening here, but it could be that communication on the I2C bus is failing for some other reason, hence the improperly controlled boost algorithm.
 
It may not have the LEDs, but it probably still has some vestigial circuitry from them, but the important thing is that the I2C bus that gets used for comms with the disco light controller (which is frequently on the cooler, not the board) is also used for comms between the GPU and power monitoring. This was why bent pins in the reference 2080 and 2080 Ti fan connectors could cause the card to fail to boost. I doubt that's exactly what's happening here, but it could be that communication on the I2C bus is failing for some other reason, hence the improperly controlled boost algorithm.
Could be, ill probably load some games on that PC and try it out, see if it holds up to extended sessions of load or not.

Thanks for the replies.
 
So small update, I continued to mess with the card, tried it again the next day, and the card wouldn't boost, couldn't even get it to work with a forced boost level or anything. seems like the card is junk for 3d work, it was worth a shot. could be just as RazorWind suggested, there doesn't appear to be any physical damage and I don't know how to test or even repair an issue I may find.
 
So small update, I continued to mess with the card, tried it again the next day, and the card wouldn't boost, couldn't even get it to work with a forced boost level or anything. seems like the card is junk for 3d work, it was worth a shot. could be just as RazorWind suggested, there doesn't appear to be any physical damage and I don't know how to test or even repair an issue I may find.
Does it crash when you try to run 3D apps, or is just kind of slow?
 
Back
Top