Asus Strix 980 not fully stable? Far Cry 4 - Display driver stopped responding

[96Firebird]

I think you've RMA'd several cards before this too, eh?

I think we're at 10+ now, if I'm not mistaken.

Listen, litwicki... Don't take this the wrong way, but I'm not sure PC gaming is for you. I don't like to say that, but if a crash here or there, or a drop in FPS in a certain area makes you believe you have hardware problems and might need to send in yet another RMA request, it's time to give it up.

PC games are buggy, PC drivers are buggy, PC operating systems are buggy. That is how it is, whether we like it or not.

 

[xorbe]

I think we're at 10+ now, if I'm not mistaken.

Listen, litwicki... Don't take this the wrong way, but I'm not sure PC gaming is for you. I don't like to say that, but if a crash here or there, or a drop in FPS in a certain area makes you believe you have hardware problems and might need to send in yet another RMA request, it's time to give it up.

PC games are buggy, PC drivers are buggy, PC operating systems are buggy. That is how it is, whether we like it or not.

First, I agree that OP has gone through a zillion cards across multiple forums. Sadly, he never provides enough info to sort out the issue, or reports back if any suggestions were tried. All we know months and years later, is that this game crashes, that game doesn't ... "is it the driver or card or game" ... same question thread after thread. I really wish we could help sort out his system if he's legit and not trolling.

But second, crashing is not the norm for a stable system! My 2500k + Titan ran for over 2 years with no notable issues. Everything went to crap when I popped in the new Titan X: tdr, ctd, reboots. Which all go away when I put the original Titan back in. Some of these stock cards are truly faulty. I also have a faulty stock 560 -- the batch that needed voltage bumps. Yes there is the occasional glitch in paradise -- most systems don't have ECC ram, etc. But when multiple games are crashing on an average of every 90 minutes ... there is a real problem.

 

[litwicki22]

Ok i think i will stay with card. Thx anyway for interest in topic and answers. I am very greatful

 

[flexy]

I have seen a thread somewhere where a guy flashed his Asus DCUII with a gigabyte Bios to get it to run stable at higher clocks. (this was after unlocking etc) Asus's uses a different voltage controller which has lower voltages than most 780's, so the Gigabyte bios treated the card like it was a normal card voltage wise so yes too much or too little voltage can be a problem.

This (possible) issue with the voltage tables in the BIOS is actually more complex than it might appear at first.

I give you an example of an EVGA SC ACX2.0, but I've seen this in pretty much ALL BIOS for GTX 970, EVGA, Gigabyte, Asus, MSI...I looked at a lot.

* First, the voltage table, means the increase in voltages per clock in the table is of course not linear. It's NOT that the same value is added with each higher clock rate.

* The maximum voltage specified in the BIOS for the max boost clock can be WAY, WAY higher than what the card (voltage controller) can actually provide.

My EVGA can only supply an absolute maximum 1.212V, yet the voltage table specifies 1.870V at clock 74. ALL Bios I looked at have as the last voltage at the end of the table some way high value they cannot even specify.

* Now, all cards are different, not only with their voltage controllers (some are programmable and might INDEED allow increasing V, some are volt locked to 1.2)....of course all cards don't boost to the same clock entry either. (ASIC quality might play a role etc. Some boost to 1380 stable, some to 1447 etc...)

BUT THE BIOS, at least the voltage table entries, are pretty much the same for all those cards even with these differences in hardware. Which means that this table is not/cannot be optimized for each and every specific card. Which also means that most cards (in my opinion) would benefit from a custom-modifed BIOS where the voltage table (aka: The voltages for each particular clock) are adjusted accordingly.
As a result, and especially since the table is not linear, it's *possible* that some cards don't get the right voltage at certain clocks. This is my THEORY.

 

[xorbe]

* First, the voltage table, means the increase in voltages per clock in the table is of course not linear. It's NOT that the same value is added with each higher clock rate.

As expected, this is a good thing.

* The maximum voltage specified in the BIOS for the max boost clock can be WAY, WAY higher than what the card (voltage controller) can actually provide.

My EVGA can only supply an absolute maximum 1.212V, yet the voltage table specifies 1.870V at clock 74. ALL Bios I looked at have as the last voltage at the end of the table some way high value they cannot even specify.

I'm sure this is some technical detail that doesn't matter.

* Now, all cards are different, not only with their voltage controllers (some are programmable and might INDEED allow increasing V, some are volt locked to 1.2)....of course all cards don't boost to the same clock entry either. (ASIC quality might play a role etc. Some boost to 1380 stable, some to 1447 etc...)

BUT THE BIOS, at least the voltage table entries, are pretty much the same for all those cards even with these differences in hardware. Which means that this table is not/cannot be optimized for each and every specific card. Which also means that most cards (in my opinion) would benefit from a custom-modifed BIOS where the voltage table (aka: The voltages for each particular clock) are adjusted accordingly.
As a result, and especially since the table is not linear, it's *possible* that some cards don't get the right voltage at certain clocks. This is my THEORY.

I've always assumed that the "asic value" controlled the application of voltage to the clocks, probably as in lower asic needs higher voltage. Maybe the asic value isn't even a single number, but some sort of bit mask.

 

[Magic Carpet]

@flexy

Seems unlikely that some chips / p-states are coming "undervolted" out of the box. They always try to give a little bit more voltage, just to be on the safe side. Interesting theory, though.

 

[xorbe]

Got a refund auth for my first card today from Digital River / nVidia Direct, and the second Titan X arrived from EVGA. Guess what, it crashes in FC4 too in the exact same way. I throw my wager down for "driver bug with Maxwell cards" because my Kepler OG Titan simply does not crash. (Hopefully it's not a hardware design bug!)

 

[amenx]

Got a refund auth for my first card today from Digital River / nVidia Direct, and the second Titan X arrived from EVGA. Guess what, it crashes in FC4 too in the exact same way. I throw my wager down for "driver bug with Maxwell cards" because my Kepler OG Titan simply does not crash. (Hopefully it's not a hardware design bug!)

Dont rule out system synergy. In rare cases some Maxwell cards do not play well with certain older motherboards, ie, incl some P67, P68, Z68 boards. But usually fixed with bios updates. Have seen this happen on 970 and 980 cards across a few forums, so maybe Titan-x as well? Its why I always think it invaluable to have a 2nd rig for cross-testing purposes.

 

[xorbe]

edit: There's big thread on the GeForce forums where other people have come to the same conclusion. In FC4, Kepler works but multiple Maxwell cards crash. One end-user tried three 980 cards. Consensus is that there is a Maxwell driver bug.

https://forums.geforce.com/default/topic/799274/geforce-drivers/nvlddmkm-keep-crashing-in-far-cry-4/

Dont rule out system synergy. In rare cases some Maxwell cards do not play well with certain older motherboards, ie, incl some P67, P68, Z68 boards. But usually fixed with bios updates. Have seen this happen on 970 and 980 cards across a few forums, so maybe Titan-x as well? Its why I always think it invaluable to have a 2nd rig for cross-testing purposes.

My system is a brand new Asus Maximus VII Hero (Z97) + 4790k + Win8.1, I can't imagine that Titan X would be incompatible. I saw that some others on OCN are also complaining about mysterious Maxwell crashes in GTA V, and others say they can play forever no issues. Makes no sense that an original Titan can run 12 hours no problem, but two different Titan X cannot make it to 3 hours.

In the motherboard bios setup, I am trying Gen2 (instead of Gen3) to see if that changes anything ... well, the Buzzer only lasted 20 minutes circling in the sky before CTD.

The only other difference that I can think of is that Titan X pushes higher fps than Titan, maybe revealing some game or gsync driver bug ... so the FC4 Buzzer is now circling with 60 Hz vsync (4 pennies on the keyboard with a light book on top to hold the key down ...)

I also switched from a 60Hz DVI screen to a 144Hz DP screen ... ugh, so many things. But the original Titan has no issues with the DP-connected 144Hz screen.

Tomorrow I get a new SS-860W PSU, and if that doesn't help, I'll re-assemble the old Asus P8P67 + 4GHz 2500K + Win7 and see what the story is there. This is insanity.

 

[litwicki22]

Hey guys. Little update. I run memtest and shows 34 errors after 5 minutes. It can be problem then with memory with display crashing? ( TDR )? Display crashing ( TDR ) can be caused by memory then?

Also today i am buying new memory set.

 

[flexy]

@flexy

Seems unlikely that some chips / p-states are coming "undervolted" out of the box. They always try to give a little bit more voltage, just to be on the safe side. Interesting theory, though.

I am actually more and more tending towards that UNDER-VOLTAGE is actually the major problem of many cards. (There is also a "black screen" crash issue with many GTX 970, say at max stock boost).

This, in the same way as when you OC a CPU would normally mean I increase voltage until the CPU/GPU is stable - PROBLEM: When they use a "cheap" VRM that is volt-locked and can only provide 1.212V maximum, you CAN not increase the voltage. If it turns out the card crashes out at 1380 at max voltage, you can't do a thing but underclocking until it gets stable.

That some cards are simple undervolted, and then also through the entire voltage table from the max boost clock downwads ...IMHO is likely because the observed problems, crashes etc. make me suspect this. I bet any money that two notches up in V would solve those problems...but of course you can't with cards that use volt-locked VRMs like many do.

 

[xorbe]

Hey guys. Little update. I run memtest and shows 34 errors after 5 minutes. It can be problem then with memory with display crashing? ( TDR )? Display crashing ( TDR ) can be caused by memory then?

Also today i am buying new memory set.

Hell yeah it can be. Did you check your memory timings are properly set? THANK YOU for finally posting some hardware testing results!

 

[Magic Carpet]

@flexy

Things you own, end up owning you, it sucks. I suppose, with all the "the less power the better" trend as of late, these things are bound to happen. Not really surprised. From a components point of view though, Asus 980 Strix looks to be quality, but maybe even Asus began slipping in the QC department. Isn't it possible to unlock voltage on Maxwell parts yet?

Unfortunately, I don't own any Maxwell parts so can't really comment. As to Kepler, it's gotten definitely trickier to tame these things, when compared to Fermi or AMD GCN. But eventually, one could solve this with custom bios modding. It's a bit too much work though, when you expect a $400+ product to work out of the box.

In rare cases some Maxwell cards do not play well with certain older motherboards, ie, incl some P67, P68, Z68 boards. But usually fixed with bios updates. Have seen this happen on 970 and 980 cards across a few forums, so maybe Titan-x as well? Its why I always think it invaluable to have a 2nd rig for cross-testing purposes.

This is so true with Kepler as well. You just don't know, until it happens to you.

 

[xorbe]

Hey guys. Little update. I run memtest and shows 34 errors after 5 minutes. It can be problem then with memory with display crashing? ( TDR )? Display crashing ( TDR ) can be caused by memory then?

Also today i am buying new memory set.

Also, reseat your CPU and check for any bent pins in the socket.

 

[flexy]

@flexy
Things you own, end up owning you, it sucks. I suppose, with all the "the less power the better" trend as of late, these things are bound to happen. Not really surprised. From a components point of view though, Asus 980 Strix looks to be quality, but maybe even Asus began slipping in the QC department. Isn't it possible to unlock voltage on Maxwell parts yet?

It depends on the VRMs being used on the PCB, as far as I know. The ASUS Strix, while a good card with good cooler also is not exactly an "enthusiast" card, this is obvious alone from the fact they use only one 150W PCIE connector. (I could be wrong tho, but it's certainly not a card made for overclocking records because of this alone)
I think it's as simple as that many makers try to cut costs manufacturing cards while at the same time of course try to compete with each other by whose cards has more "overclock" out of the box because those 50 or so mhz more (lol) seem to matter, especially in reviews where they mention which card "comes with the highest stock OC".
And then of course issues with QA.

Then of course I could be entirely wrong and GTX970, the majority of them are flawless and I only have the impression many have problems because I hang out on manufacturer's forums where naturally everyone goes and posts where they have problems...

 

[litwicki22]

Guys. I resolved issue with TDR crash in FC4. I bought new memory set. Now its working no crashes ok with 350.12 WHQL drivers.
My previous memory set was giving random errors in memtest so i am thinking thats why i had tdr crashes.

Check your memory in memtest i suggest.

I bought that one:
HyperX 2x4GB 1600MHz DDR3 Non-ECC CL9 DIMM Plug n Play (KHX1600C9D3P1K2/8G)

 

[litwicki22]

Little update.
Sorry i blame drivers because Far Cry 4 was crashing TDR display . It was my faulty memory. I tested for 3 days Far Cry 4 for 10 hours without single TDR crash even on 350.12 WHQL drivers. ))) So new ram helped a lot. So my previous ram was just faulty. And 350.12 WHQL are just fine.So drivers are ok
Now my pc is stable. New rams:
HyperX 2x4GB 1600MHz DDR3 Non-ECC CL9 DIMM Plug n Play (KHX1600C9D3P1K2/8G)
Using 350.12 WHQL

 

[xorbe]

Even with enhanced godrays and fur simulated? Or did you disable those?

 

[litwicki22]

All options enabled ( Nvidia God Rays , Fur, Txaa 2x ) and details on ultra.
Today is 4 day when i test Far Cry 4 for 10 hours without single TDR crash even on 350.12 WHQL drivers. So new ram resolve TDR issue.
I was thinking that my graphic card is unstable. But that was ram.

 

[xorbe]

Well, I already swapped ram and psu and gfx card, I'll try memtest overnight for the heck of it. I'll try to narrow it down to fur or godrays. I've just been disabling both at once. Or maybe it interacts with the AA ... I'm using a different one than you.

 

[litwicki22]

Little update. TDR crash with FC64.dll is back even with new rams. 5 days i was testing without single crash. Today TDR's back even with new ram. So all of my hope that was ram issue is gone. Now with new ram the same TDR crashing.
Card is Asus Strix 980 and my previous rams was 2x4gb corsair vengeance 1333mhz , new ram are Kingston Hyperx 2x4gb 1600mhz 1.5v.

And till today i get my first TDR crash with new rams App stopped, FC64.dll, and minidump file in watchdog folder reporting that nvldkmm crashed. Is my pc not stable or is this a game issue ?
I am using 350.12 WHQL.

Psu is COrsair 750 RM and cpu on stock 4790k. So that was not ram. On new rams TDR crashing back.But for 5 days i was dont get single crash. Strange, huh, but why TDR started today ?!Anyway, issue is still on my pc on Far Cry 4. EHhh <sad>

 

[litwicki22]

I am thinking about RMA my card. Somebody said:"Because it's one game that shows the card isn't stable" . So its not a driver issue but its my card yes? I am using Asus Strix 980 stock.

 

[Dave3000]

I think the latest drivers might be a bit crashy, quite a few people have been complaining of similar problems, completely locked up in various games at random times. It's even started doing this on my 780.

It's most likely the overclocks on either your GPU or CPU that are causing your crashes in games as you said it also is doing this on your 780. Try running every piece of hardware in your system at stock settings and see if your games still crash on you.

 

[Dave3000]

It's most likely the overclocks on either your GPU or CPU that are causing your crashes in games as you said it also is doing this on your 780. Try running every piece of hardware in your system at stock settings and see if your games still crash on you.

I must have interpretted wrong what you said. I took it as the 780 was your spare GPU but based on your posting your system specs show a 780 GPU listed. I guess I need to get some sleep now as I was awake all night and maybe can't think straight. Still I suggest running both your GPU and CPU at stock settings and see if it solves your system crashing in games.

 

[3DVagabond]

FC4 has issues. I wouldn't use it as the sole determination that a card is defective.