Why the high GPU Temps?

[LRBS 944]

54 minutes ago, fogboundturtle said:

stop it. Thermal is physical issue on how much  TDP you can dissipate in your case. This has nothing to do with software.  If you temp increase it means your case in unable to exhaust and dissipate the heat fast enough. So check your air flow, redo  thermal paste etc.... Stop blaming MSFS for your issue.

Wow!

It appears that I rocked your boat. I'm afraid that you can't accept those numbers where clearly is a temperature difference between the two. 

If it was an issue of airflow, you would expect to see high temperatures all across the platform, not only with one particular software. Granted, with the new update it appears to me that some optimization took place and I don't notice anymore so much difference.

 

[fogboundturtle 2,708]

3 minutes ago, LRBS said:

Wow!

It appears that I rocked your boat. I'm afraid that you can't accept those numbers where clearly is a temperature difference between the two. 

If it was an issue of airflow, you would expect to see high temperatures all across the platform, not only with one particular software. Granted, with the new update it appears to me that some optimization took place and I don't notice anymore so much difference.

 

because you speak nonsense. 

[LRBS 944]

3 hours ago, fogboundturtle said:

because you speak nonsense. 

Take it easy, will you?

Actually, about six months ago it was an update for MSFS where actually they did optimize the software. In case you forgot.Also a few months back it was an issue with a game that was overheating PCs especially GPUs to insane temperatures.

So, please don't say that has nothing to do with software or I speak nonsense. 

 

Edited February 7, 2022 by LRBS

[Adrian123 1,015]

12 minutes ago, LRBS said:

a few months back it was an issue with a game that was overheating PCs especially GPUs to insane temperatures.

I'm not disputing this but I don't recall seeing anything on that subject? Any rescouces?

[wims80 121]

GPUs are designed to run at nominal temperature 24/7 for the entire warranty period. Software should not be able to push the temperature above nominal in any situation. If you are seeing temperatures above nominal then there is something wrong with the cooling / airflow

Edited February 7, 2022 by wims80

[LRBS 944]

48 minutes ago, Adrian123 said:

I'm not disputing this but I don't recall seeing anything on that subject? Any rescouces?

I hear you loud and clear, no worries. It's just a discussion, sharing experiences etc.

 

https://www.gamingbible.co.uk/news/games-amazons-new-world-is-destroying-gaming-gpus-20211001

 

And @wims80, I agree with the statement that cooling could be an issue, but when a specific software is creating a temp issue and others are not, maybe we should entertain the idea that some coding or hardware might play a role.

 

A few months ago, Asobo did some game optimization on their end and solved many issues with CPU optimization and GPU leading to excellent results, you can find that on their forum.

 

Now, it's vey unprofessional  to see people telling you to stop or not to point discrepancies about something especially in our case where Asobo was aware about the issue and fixed it very nice.

 

[fogboundturtle 2,708]

Just now, LRBS said:

I hear you loud and clear, no worries. It's just a discussion, sharing experiences etc.

 

https://www.gamingbible.co.uk/news/games-amazons-new-world-is-destroying-gaming-gpus-20211001

 

And @wims80, I agree with the statement that cooling could be an issue, but when a specific software is creating a temp issue and others are not, maybe we should entertain the idea that some coding or hardware might play a role.

 

A few months ago, Asobo did some game optimization on their end and solved many issues with CPU optimization and GPU leading to excellent results, you can find that on their forum.

 

Now, it's vey unprofessional  to see people telling you to stop or not to point discrepancies about something especially in our case where Asobo was aware about the issue and fixed it very nice.

 

Even if specific instruction could in theory generates more heat, the issue is entirely based on the capacity of your airflow and cooler to dissipate that heat. Like it was mention before CPU/GPU are capable of running at full capacity under TJ max as long as you can remove the heat.  Whatever issue you are attribute this increase of heat, it's entirely on your external ambient temperature where your PC is located and the ability to exhaust that heat. It's a physical issue. It's not a software one.  Per example, if you run AVX 512 instruction on an Intel CPU, yes it will run hotter than if you no are not. This is an expected behavior as AVX-512 requires more transistor and more computably intensive to run. Your capacity to handle this is entirely all on YOU.  

In the case of new world, This was again a physical issue with capacitator on eVGA card and card that were exceeding the power limit. Again, another physical problem. New World exposed that issue but running uncapped framerate. It was a physical issue all along. It was never Amazon New World's fault. 

[LRBS 944]

2 hours ago, fogboundturtle said:

Even if specific instruction could in theory generates more heat, the issue is entirely based on the capacity of your airflow and cooler to dissipate that heat. Like it was mention before CPU/GPU are capable of running at full capacity under TJ max as long as you can remove the heat.  Whatever issue you are attribute this increase of heat, it's entirely on your external ambient temperature where your PC is located and the ability to exhaust that heat. It's a physical issue. It's not a software one.  Per example, if you run AVX 512 instruction on an Intel CPU, yes it will run hotter than if you no are not. This is an expected behavior as AVX-512 requires more transistor and more computably intensive to run. Your capacity to handle this is entirely all on YOU.  

In the case of new world, This was again a physical issue with capacitator on eVGA card and card that were exceeding the power limit. Again, another physical problem. New World exposed that issue but running uncapped framerate. It was a physical issue all along. It was never Amazon New World's fault. 

I agree with some of your statements but I'm afraid that there are some misunderstandings here.

I never said or implied that it was Amazon's fault.

I wonder what could it be that after Asobo optimized their code, the same PC with no changes, at the same psychical location, the same conditions, will not exhibit any fluctuations in temperatures as before? This is where I have a hard time agreeing with the notion (in this particular case) that was related to cooling, AVX instructions (that I didn't change) other than that it was just a coding issue. Especially where other games or applications on my PC (like photoshop video editing very heavy on CPU and somehow on GPU) did not have any problems.

There are known issues with certain games like WoW, Trove, Planet Gorgon, Shroud of the Avatar, Stonehearth, etc that are not properly coded and known to run hotter than normal.

If you can have any explanation for this particular situation (not in general) I am listening.

[martin-w 3,234]

On 2/7/2022 at 3:25 PM, hangar said:

My question about all this is, what can be done to limit the gpu while in the msfs main menu, for those of us who don't want to cap our frames during flight?

 

I don't think there's much that can be done unless the developers provide two frame rate limiters. One for the game and one for the menus. 

In regard to my issue, I've found out what the issue is. When I looked at GPU temperature while in the FS menus, I noticed that it wasn't so much GPU temp and CPU temp that was the issue, it was GPU memory temperature. DDR6 memory temperature was running at 94 degrees. While the GPU was ruining at pretty normal temps. GPU will still be around 50-60 degrees of course, due to the highly detailed hangar that rendering in the background.

94 degrees was in the menus and when running the sim. When running other software like COD Warzone and Farcry6, the cards m,ery temp was 90 degrees. Again, GPU temp normal.

DDR6 does run hot of course, that we know, but 94 degrees was right up to the max we should favour for long term use. Any higher and you are approaching the level where sustained use could be damaging in the long term. The memory manufacturer Micron recommend up to 95/110 degrees. Beyond 110 degrees is bad news.

My solution wasn't frame rate caps or anything like that, instead I used EVGA's Precision X1 to reduce the cards power target to 80%. When I did that, the temp of the DDR6 Maximum Junction Temperature dropped down to 88 degrees. Within an acceptable range for DDR6.

Fans are now much quieter in the menus and in tests I've seen virtually no drop in frame rate in the sim. Still at 70's to 80's on Ultra settings.

 

I think "buildzoid" is right in his videos where he claims that the 3090 was never designed to be pushed as far as some manufacturers are pushing them. 

 

I should add, many people are reverting to all manner of solutions to handle the high memory temp. Some are installing fans on the top plate of the cards, some are installing custom loops and EVGA hybrid cooling kits. 

 

 

 

 

Edited February 9, 2022 by martin-w

[martin-w 3,234]

On 2/7/2022 at 11:42 PM, fogboundturtle said:

In the case of new world, This was again a physical issue with capacitator on eVGA card and card that were exceeding the power limit. Again, another physical problem. New World exposed that issue but running uncapped framerate. It was a physical issue all along. It was never Amazon New World's fault. 

 

Not true, it was actually both!

See buildzoid video above, where he points out that some software pushes the cards past their power limit. Often in terms of spikes. 

Actually the EVGA New World issue was a defective solder joint for one of the mosfets and New World was pushing the cards beyond their power limit which was revealing the defect in the card. Furthermore, it was never just EVGA cards. Gigabyte cards also had issue with New World, as did some other manufacturers.

As buildzoid points out in his video, those who claim that software can't brick a card are in error. And those who have run Furmark in the past will remember the box that used to pop up warning the user that the software could damage cards and you were to use at your own risk.

There is series of Jayz2cents videos were he conducts tests with New World and points out the software issue that were causing 3090's and some 3080's to spike way above the power limit.

 

Essentially it was Amazons dodgy programming that was pushing cards beyond their power limits. Obviously when that happens any cards with defects will be exposed. 

 

 

Edited February 9, 2022 by martin-w

[martin-w 3,234]

On 2/8/2022 at 2:04 AM, LRBS said:

Even if specific instruction could in theory generates more heat, the issue is entirely based on the capacity of your airflow and cooler to dissipate that heat.

 

Its not theory that software can damage cards. Its well know fact. Something we have known for a very long time.

And its not heat that's the issue, it's overloading components that aren't designed to accommodate power spikes.

Furthermore, even if it was heat... your cooling isn't going to help you because the issue here is transient spikes well above the power limit. Beyond the capability of cooling to react. 

The solution is to lower the power target so that the spikes have some headroom and don't exceed the power limit. And for developers like Amazon to code better.

[martin-w 3,234]

 

On 2/7/2022 at 10:51 PM, wims80 said:

GPUs are designed to run at nominal temperature 24/7 for the entire warranty period. Software should not be able to push the temperature above nominal in any situation. If you are seeing temperatures above nominal then there is something wrong with the cooling / airflow

 

Well obviously software is capable of doing that. Doesn't matter how good your cooling is, it's still possible to place a software demand on the component that exceeds that cooling capacity. There are software stress tests out there that place all manner of varying loads on our systems, and yes, they can be coded to exceed the cooling capacity of even a custom loop.

But what then, a chiller perhaps? And what if we code software that exceeds the cooling capacity of a chiller? A chiller plus Peltier perhaps?

And what if the software is coded to exceed the cooling capacity further? Liquid nitrogen perhaps?

Obviously we have to be sensible here and consider the software an issue if it generates a load that exceeds a "reasonable cooling capacity". 

Edited February 9, 2022 by martin-w

[martin-w 3,234]

On 2/7/2022 at 5:55 PM, fogboundturtle said:

Stop blaming MSFS for your issue.

 

We can blame a developer for a temp issue if the software generates a heat load that is beyond the average customers likely cooling capacity. 

 

Edited February 9, 2022 by martin-w

[fogboundturtle 2,708]

4 hours ago, martin-w said:

 

We can blame a developer for a temp issue if the software generates a heat load that is beyond the average customers likely cooling capacity. 

 

That is your problem to solved not the developer. Your cooling capacity is a physical issue with your setup. Dev has nothing to do with this.

[martin-w 3,234]

1 hour ago, fogboundturtle said:

That is your problem to solved not the developer. Your cooling capacity is a physical issue with your setup. Dev has nothing to do with this.

 

Wrong! As we've already established, software is perfectly capable of forcing hardware beyond its power limits or introducing dangerous power spikes. So clearly the "Dev" can have a lot to do with it.

Wrong again! If a product is sold, knowingly, that pushed temp  beyond the capability of most people's systems to cool properly, then obviously, the developer is selling a product that is not fit for purpose for most of its intended customers. 

So with your logic, if a software was released that caused all of the systems running it to overheat, you would say its not the developer fault, get better cooling. Weird logic my friend. Its fairly obvious that a product has to be "fit for purpose" and if it overheats the majority of people PC's, then its not fit for purpose. And being "fit for purpose" is as legal requirement. 

 

As Is said... "We can blame a developer for a temp issue if the software generates a heat load that is beyond it's average customers likely cooling capacity."

you would also have to wonder about the "ethics" of a hypothetical company that deliberately released a product that most of its intended customer base couldn't run.

 

Edited February 9, 2022 by martin-w