MSI GTX 1080 Armor OC Power Limited

[TheJoker]

But try this out of curiosity to see what it does :

In Afterburner click on the voltage/frequency editor, select the the frequency that corresponds to 1.093 Volts, press L and apply. That should force the card to stay at the highest voltage state.

I formated my pc now, reseated the gpu, i will install the necessary drivers for mobo, nvidia driver and msi to test the voltage lock on heaven
Can you please tell me the real answer to my question that i got fooled on? I see that you know what are you talking about and i want to understand better why is it normal that my gpu hits power limit

 

[Vya Domus]

Every card eventually throttles under normal circumstance either because it hits the power limit or as the temperature gradually increases.

For Pascal once you are over 60C as the temperature goes up the card will slowly dial back the clocks in small 13 mhz steps. Same thing happens with the power limit. Reference 1080s can actually go as low as 1600mhz.

 

[TheJoker]

Every card eventually throttles under normal circumstance either because it hits the power limit or as the temperature gradually increases.

For Pascal once you are over 60C as the temperature goes up the card will slowly dial back the clocks in small 13 mhz steps. Same thing happens with the power limit. Reference 1080s can actually go as low as 1600mhz.

I understand that, but for sure i didn't understand why everyone was claiming that is so wrong to hit those limits. I made a couple of tests in benchmarks and i noriced a strange thing. I put the fan speed at max and ran a test. The core clocks were stable (~1934 stock) and voltages were between 1.093-1.080-1.062. I restarted my pc after a couple of minutes, and then the surprise: core clocks all over the place, because voltages wouldn't go any higher than 1.000-1.031, it even reached .973 one time. Then i restarted again, and it was all good again.

 

[FreedomEclipse]

Every card eventually throttles under normal circumstance either because it hits the power limit or as the temperature gradually increases.

For Pascal once you are over 60C as the temperature goes up the card will slowly dial back the clocks in small 13 mhz steps. Same thing happens with the power limit. Reference 1080s can actually go as low as 1600mhz.

Just read about this a few days ago myself though I wasn't sure if it was true or not. I'll need to run a test with my fans set to 80% on my crappy cooler to see if that really helps clocks

 

[phill]

Hello guys. I bought a couple of days ago a new gtx 1080 msi armor. Since then, whatever i play, or benchmark, the card is hitting the power limit and throttle back the cores. The original bios jas a 210watts limit.

I searched the data base here and i found another bios for my armor with 290watts being the limit. I flashed it and all good, the card doesn't hit the power limit and the clocks stay constant. But especially in 3d mark the app is closing itself while testing, without giving an error
In heaven, sometimes the screen gets black, and the app crashes, but without an error from the nvidia driver

I then flashed back my original bios and to my surprise i found that the card now can't keep the voltages. It runs 1900mhz on 1.000 or sometimes it drops to even 0.980 voltages. It never goes up higher that 1.041 now. Is there a possibility that i just broke my card?

Have you tried running it stock/are you running it stock with it boosting on it's own with standard power/temp limits etc to see if it's still causing the same problems?

Just read about this a few days ago myself though I wasn't sure if it was true or not. I'll need to run a test with my fans set to 80% on my crappy cooler to see if that really helps clocks

It will

I never use the standard/stock fan curve, I always manually set the fan speed at a set amount as it will help with performance on the card.

Lent a mate from work my EVGA 1070 SC, stock drivers and speeds he was hitting 80C (as max is 83C it'll just run below) said to him to stick it to about 50% or so (depending on noise and what he's happy with) and came in the next day saying that it never surpassed 50C and the boosts where higher
Auto is fine if it's not a massively demanding title, but for anything else that makes it work, slap a decent fan speed on it, it'll boost away

I think my 1070 used to boost to over 1950Mhz with that fan speed set, same goes for my 1080 Ti as well.. Keeps the temps down which keeps the GPU core speed higher

 

[TheJoker]

Have you tried running it stock/are you running it stock with it boosting on it's own with standard power/temp limits etc to see if it's still causing the same problems?



It will

I never use the standard/stock fan curve, I always manually set the fan speed at a set amount as it will help with performance on the card.

Lent a mate from work my EVGA 1070 SC, stock drivers and speeds he was hitting 80C (as max is 83C it'll just run below) said to him to stick it to about 50% or so (depending on noise and what he's happy with) and came in the next day saying that it never surpassed 50C and the boosts where higher
Auto is fine if it's not a massively demanding title, but for anything else that makes it work, slap a decent fan speed on it, it'll boost away

I think my 1070 used to boost to over 1950Mhz with that fan speed set, same goes for my 1080 Ti as well.. Keeps the temps down which keeps the GPU core speed higher

The unfortunate thing is that my temps never get past 65-67, and still, the card is downvolting mid-3dmark benchmark to 1.000-1.021v, and the clocks are dropping. In unigine valley, it stays smooth and at high voltages 1.081-1.093-1.075 and the core fluctuates just by 25mhz up and down. Now if i put my valley in full hd and high settings the pattern from 3d mark repeats itself, the card is downclocking hard. If i put valley on ultra, x8aa and 2k(native res) then it's all smooth, voltages and clocks

 

[eidairaman1]

Where are your system specs at? Please provide the brand and model of power supply you are using

 

[phill]

The unfortunate thing is that my temps never get past 65-67, and still, the card is downvolting mid-3dmark benchmark to 1.000-1.021v, and the clocks are dropping. In unigine valley, it stays smooth and at high voltages 1.081-1.093-1.075 and the core fluctuates just by 25mhz up and down. Now if i put my valley in full hd and high settings the pattern from 3d mark repeats itself, the card is downclocking hard. If i put valley on ultra, x8aa and 2k(native res) then it's all smooth, voltages and clocks

Have you got Afterburner installed so you can track the GPU core volts and GPU core clock when you run a test?

Where are your system specs at? Please provide the brand and model of power supply you are using

This too.....

 

[TheJoker]

My system specs:
i5-8600k 4.8ghz stable, tested with prime and aida
Corsair 16gb ram ddr4 rgh 2666mhz xmp (memtested, no error)
Mobo aorus z370 gaming 3 (latest bios F7)
PSU Sirtec Astro GD 850w 80+gold, bought last year
Monitor LG 2k 165hz G-Sync 31,5"
GPU MSI gtx 1080 Armor Oc
Well ventilated case, positive air pressure(4intake-3 exhaust fans)
Cpu cooled with a cooler master v8 (no more than 70°C and that on one core)

Msi installed. Just played Vampyr, all on ultra, 2k. Core clock 2012, voltage 1.081 during gameplay(63°C max). As soon as i enter 3dmark to test the gpu i got 19.200 points in firestrike default, but the clocks and voltages are again all over the place, voltages even dipping to 1.000v

 

[eidairaman1]

My system specs:
i5-8600k 4.8ghz stable, tested with prime and aida
Corsair 16gb ram ddr4 rgh 2666mhz xmp (memtested, no error)
Mobo aorus z370 gaming 3 (latest bios F7)
PSU Sirtec Astro GD 850w 80+gold, bought last year
Monitor LG 2k 165hz G-Sync 31,5"
GPU MSI gtx 1080 Armor Oc
Well ventilated case, positive air pressure(4intake-3 exhaust fans)
Cpu cooled with a cooler master v8 (no more than 70°C and that on one core)

Msi installed. Just played Vampyr, all on ultra, 2k. Core clock 2012, voltage 1.081 during gameplay(63°C max). As soon as i enter 3dmark to test the gpu i got 19.200 points in firestrike default, but the clocks and voltages are again all over the place, voltages even dipping to 1.000v

Ok its a problem with the benchmark and gpu driver at this point report it to the driver team and benchmark team.

Enjoy your games

 

[phill]

My system specs:
i5-8600k 4.8ghz stable, tested with prime and aida
Corsair 16gb ram ddr4 rgh 2666mhz xmp (memtested, no error)
Mobo aorus z370 gaming 3 (latest bios F7)
PSU Sirtec Astro GD 850w 80+gold, bought last year
Monitor LG 2k 165hz G-Sync 31,5"
GPU MSI gtx 1080 Armor Oc
Well ventilated case, positive air pressure(4intake-3 exhaust fans)
Cpu cooled with a cooler master v8 (no more than 70°C and that on one core)

Msi installed. Just played Vampyr, all on ultra, 2k. Core clock 2012, voltage 1.081 during gameplay(63°C max). As soon as i enter 3dmark to test the gpu i got 19.200 points in firestrike default, but the clocks and voltages are again all over the place, voltages even dipping to 1.000v

Voltages will go up and down when in game depending on game and complexity etc, same as the core clock, same as the core temp.. Personally I'd have gone for a different PSU but otherwise it looks fine. When you're saying that the voltages go all over the place, is this inbetween the test screens, so where the next bench test runs?

 

[TheJoker]

Voltages will go up and down when in game depending on game and complexity etc, same as the core clock, same as the core temp.. Personally I'd have gone for a different PSU but otherwise it looks fine. When you're saying that the voltages go all over the place, is this inbetween the test screens, so where the next bench test runs?

In 30 mins i will be back with some screens of msi and hwinfo64 after the 3dmark test

 

[phill]

Cool Hopefully we can put your worries to rest

 

[TheJoker]

https://www.3dmark.com/3dm/29577306
You see, my voltage dropped to 0.993 at only 65C and it is very unstable. My question is how i can manage 19.500score with those fluctuations
Now i will run valley unigine in extreme preset, 2k and i will post the results. It will be a big difference

Spoiler

Now the clocks never dipped below 2012 and voltage never dipped below 1.062, and that for a second. And i ran the test on extreme 2k. If i run this test in 1080p high, the same story with 3dmark repeats. Although i don t think it can be a cpu bottleneck

Spoiler: Valley Unigine

 

[Vayra86]

Every card eventually throttles under normal circumstance either because it hits the power limit or as the temperature gradually increases.

For Pascal once you are over 60C as the temperature goes up the card will slowly dial back the clocks in small 13 mhz steps. Same thing happens with the power limit. Reference 1080s can actually go as low as 1600mhz.

We are going in circles I think. This was explained earlier but his card doesnt hit over 70C so the voltage capping and/or major downclocking should not happen. It only happens above 83C.

Conclusion: card is not working properly and a voltage curve edit to 'get stable' is a big fat RMA sticker in my book

 

[TheJoker]

We are going in circles I think. This was explained earlier but his card doesnt hit over 70C so the voltage capping and/or major downclocking should not happen. It only happens above 83C.

Conclusion: card is not working properly and a voltage curve edit to 'get stable' is a big fat RMA sticker in my book

Yea, you are right. But if you are in a shitty country like i am, i need to check every possibility before i send it to the store that i bough to RMA because if they find it working(and they do that, because a lot of them don't know what are they doing, and if the pc powers on, they say all is working) i need to pay half the price of the card to see it back

 

[Vayra86]

Gotcha, I tend to forget this from time to time... here in NL we have awesome customer rights, usually things are easy.

Some tips about RMA:

- Be very clear about specifying the problem, and leave all emotion out of your email. Be factual, and use this line of thought: Business is business, and the product provided does not work as advertised. That is all. Replace with a fully functional card with full warranty. Thank you for your trouble.

And on top of that, be explicit about you not accepting an RMA refusal, and put an advance warning in there that you will escalate this case to wherever it needs to go to get resolved. For that you may have several options. If you have a legal assistance insurance, be sure to include that in your warning. If you don't, find a customer rights commission in your country (it should exist) and wave that in front of them.

- Make pictures of the card, especially PCB, and make sure they are highly detailed and well lit+sharp. We have seen companies destroy/damage cards so they don't need to RMA them. Make sure you include that information (not the pictures themselves) in your message as well.

- Include the MSI rep forum statement and ask MSI how they think this is correct customer service on a faulty product, and about how apps crash with the new BIOS. Let em sweat a little, too - tell them you can expose this on social media if the circumstance requires it. Tell them to either explain or apologize for this. Including this gives weight to your complaint, because as a customer you feel you have been mistreated and you have evidence to support that. Its a bit of 'carrot VS stick' treatment: they can try to fix this (carrot) and make it a win-win situation, or they can make it worse for all those involved (stick).

All of those elements together in a well worded, factual message will move any company into action. The right action. Good luck!

 

[TheJoker]

I will follow your advice and write an email to both the store and msi, and i will compose it by your guidelines. I told msi already and they told me that if the store is behaving, they will intervene and help facilitate the rma.
Can you please also take a look at the results i posted on those benchmarks? It seems to me that something is weird in 3dmark compared to valley?

 

[Vayra86]

I will follow your advice and write an email to both the store and msi, and i will compose it by your guidelines. I told msi already and they told me that if the store is behaving, they will intervene and help facilitate the rma.
Can you please also take a look at the results i posted on those benchmarks? It seems to me that something is weird in 3dmark compared to valley?

Go through the store first, its the normal way of doing it. Don't walk the other track outside of the store. If you go through the store, make sure you tell them that you will not accept a returned card, and leave it at that. They aren't to blame, they just need to do the work.

IF the store somehow refuses, go through MSI and include the above.

EDIT: as for your Valley run: yes, that is how it should be looking, but you can also see the voltage will not spike over 1.081v. Valley is less demanding and it will achieve the best GPU clocks and lower temperatures than 3DMark. This is also why its not a great stability testing tool (but a fantastic tool to test overclocking and max clocks), and 3DMark is a lot better for that purpose. The bottom line remains: all applications should function normally and they do not.

Oh one last thing: make sure the serial number is clearly visible on a photo.

 

[FreedomEclipse]

- Include the MSI rep forum statement and ask MSI how they think this is correct customer service on a faulty product, and about how apps crash with the new BIOS. Let em sweat a little, too - tell them you can expose this on social media if the circumstance requires it. Tell them to either explain or apologize for this. Including this gives weight to your complaint, because as a customer you feel you have been mistreated and you have evidence to support that. Its a bit of 'carrot VS stick' treatment: they can try to fix this (carrot) and make it a win-win situation, or they can make it worse for all those involved (stick).

All of those elements together in a well worded, factual message will move any company into action. The right action. Good luck!

This sounds very threatening

 

[Final_Fighter]

did you try going into the nvidia control panel and turning power management mode to "prefer maximum performance"?

 

[TheJoker]

did you try going into the nvidia control panel and turning power management mode to "prefer maximum performance"?

View attachment 109000

Yeap, i did that and i also set high performance power plan in windows settings. This monday i will rma the card, but still i would like to get more opinions. Have any of you had any gtx 10x card that downvolted by itself as far as 0.983-1.000v at 65°C?

 

[Vayra86]

This sounds very threatening

Yes, re-reading it, it does. At the same time, as a customer, you have every right to be pissed when a company sends you into the bushes with false advice.

 

[Vya Domus]

As soon as i enter 3dmark to test the gpu i got 19.200 points in firestrike default, but the clocks and voltages are again all over the place, voltages even dipping to 1.000v

Firestrike has a CPU test and a combined test in which GPU clocks drop significantly. My question is : is there an actual game where performance does in fact plummet ?

 

[TheJoker]

Firestrike has a CPU test and a combined test in which GPU clocks drop significantly. My question is : is there an actual game where performance does in fact plummet ?

Well, MSI just confirmed that they will send me a new one free and i can keep this one. I re-flashed the bios that they indicated(the one with 290w as the power limit). Well, with it it's a total difference. Firstly, the firestrike score raised to 20.000pts. The voltages are staying at 1.062-1.050 without dropping and the clocks are between 1940-1932 without oc. The power limit is completely gone. What the heck is this? Because the first time i flashed this bios, i got random crashes everytime. I will get back to testing this bios hard to see if it crashes. It seems that this card has a mind of it s own