Space Engineers

Space Engineers

View Stats:
Quxudais Jan 5, 2016 @ 12:54am
Crazy high FPS in loading screen
Fired up a planetary save and happened to have my evga overlay on at the time, I noticed that while it was loading my overlay reported over 3,000 fps and my cpu was spiking to nearly 80c (gpu stayed roughly 65c). Anyone else noticed this? certainly cannot be normal.
< >
Showing 1-15 of 16 comments
Thalyn Jan 5, 2016 @ 12:59am 
It's displaying basically a static image. Do you really expect it to have a low framerate while doing so? Your GPU temperature was fine so it's not like it was causing any issues (your CPU temperature had little to do with the framerate).

If it bothers you, impose a limit. V-sync is the most universal, but I believe part of MSI's Afterburner software can impose a custom limit if you'd rather keep V-sync off.
Nitro Jan 5, 2016 @ 6:53am 
I prefer to keep V-Sync on, it allows for stability in my framerates and prevents any possible screen tearing.
Puppet #7 Jan 5, 2016 @ 8:26am 
Its a well known Nvidia issue called coil whine- V-sync as mentioned is the fix.
›Kolanaki Jan 5, 2016 @ 8:32am 
Originally posted by Puppet #7:
Its a well known Nvidia issue called coil whine- V-sync as mentioned is the fix.

This isn't what coil whine is...
Chip Patton Jan 5, 2016 @ 8:39am 
Originally posted by Quxudais:
Fired up a planetary save and happened to have my evga overlay on at the time, I noticed that while it was loading my overlay reported over 3,000 fps and my cpu was spiking to nearly 80c (gpu stayed roughly 65c). Anyone else noticed this? certainly cannot be normal.

Turn on adaptive sync in the Nvidia control panel. It should cap the load screen to 60fps and turn itself off if the fps is any lower than 60.

Originally posted by Thalyn:
It's displaying basically a static image. Do you really expect it to have a low framerate while doing so? Your GPU temperature was fine so it's not like it was causing any issues (your CPU temperature had little to do with the framerate).

If it bothers you, impose a limit. V-sync is the most universal, but I believe part of MSI's Afterburner software can impose a custom limit if you'd rather keep V-sync off.

It is not normal to run a GPU into working loads on a load screen... 2000+ fps can burn out a video card and that's not a good thing.
Quxudais Jan 5, 2016 @ 2:43pm 
Originally posted by Chip Patton:
Originally posted by Quxudais:
Fired up a planetary save and happened to have my evga overlay on at the time, I noticed that while it was loading my overlay reported over 3,000 fps and my cpu was spiking to nearly 80c (gpu stayed roughly 65c). Anyone else noticed this? certainly cannot be normal.

Turn on adaptive sync in the Nvidia control panel. It should cap the load screen to 60fps and turn itself off if the fps is any lower than 60.

Originally posted by Thalyn:
It's displaying basically a static image. Do you really expect it to have a low framerate while doing so? Your GPU temperature was fine so it's not like it was causing any issues (your CPU temperature had little to do with the framerate).

If it bothers you, impose a limit. V-sync is the most universal, but I believe part of MSI's Afterburner software can impose a custom limit if you'd rather keep V-sync off.

It is not normal to run a GPU into working loads on a load screen... 2000+ fps can burn out a video card and that's not a good thing.

This was why I posted. Many games have limits coded into their menus and screens to prevent this kind of thing from happening. Last time I saw numbers like that was the infamous Starcraft 2 debacle that melted a lot of peoples hardware.
Puppet #7 Jan 5, 2016 @ 5:06pm 
Originally posted by › The Cool Side of the Pillow:
This isn't what coil whine is...

Not that it matters but Coil whine happens when a card becomes stressed- spitting multiple thousand frames a second is likely considered being stressed- how is that not coil whine?

And either way V-sync does stop it in SE.
RayvenQ  [developer] Jan 5, 2016 @ 5:11pm 
Originally posted by Puppet #7:
Originally posted by › The Cool Side of the Pillow:
This isn't what coil whine is...

Not that it matters but Coil whine happens when a card becomes stressed- spitting multiple thousand frames a second is likely considered being stressed- how is that not coil whine?

And either way V-sync does stop it in SE.

Coil whine is when your GPU makes a noise because of rendering way too many frames than it really needs to or should. But yes, enabling V-Sync, through a variety of means, eithe ringame, in your Graphics Contorl panel or via 3rd party software will stop the game trying to render a ridiculous amount of frames on a loading screen.
Thalyn Jan 5, 2016 @ 11:05pm 
Originally posted by Chip Patton:
It is not normal to run a GPU into working loads on a load screen... 2000+ fps can burn out a video card and that's not a good thing.
Framerate doesn't burn out GPUs. Inadequate cooling and malfunctioing failsafes do. Otherwise all modern graphics cards would burn out running older software, which they don't.

Any GPU worth its salt will have more than adequate cooling for the raster units, which are what are doing most of the work on a static screen. A GTX980Ti (for example) has 96 of these in total, compared with over 2,500 CUDA cores, so if it cannot keep itself in check when those CUDA cores are almost idle while the ROPs are flat-stick than it's not well designed.

I do, however, recall the StarCraft II debacle over NVidia's 252-series beta drivers, with framerates between 200 and 300fps during loading screens destroying cards. This wasn't because of how high the framerate was, but because the drivers had inadvertently disabled some of the protections to stop the card exceeding its limits. *Specifically they were being allowed to well-exceed their power draw limits, resulting in the PEG connectors de-soldering themselves from heat build-up.
Last edited by Thalyn; Jan 5, 2016 @ 11:14pm
Chip Patton Jan 5, 2016 @ 11:51pm 
Originally posted by Thalyn:
Originally posted by Chip Patton:
It is not normal to run a GPU into working loads on a load screen... 2000+ fps can burn out a video card and that's not a good thing.
Framerate doesn't burn out GPUs. Inadequate cooling and malfunctioing failsafes do. Otherwise all modern graphics cards would burn out running older software, which they don't.

Any GPU worth its salt will have more than adequate cooling for the raster units, which are what are doing most of the work on a static screen. A GTX980Ti (for example) has 96 of these in total, compared with over 2,500 CUDA cores, so if it cannot keep itself in check when those CUDA cores are almost idle while the ROPs are flat-stick than it's not well designed.

I do, however, recall the StarCraft II debacle over NVidia's 252-series beta drivers, with framerates between 200 and 300fps during loading screens destroying cards. This wasn't because of how high the framerate was, but because the drivers had inadvertently disabled some of the protections to stop the card exceeding its limits. *Specifically they were being allowed to well-exceed their power draw limits, resulting in the PEG connectors de-soldering themselves from heat build-up.

That's simply not true. A GPU has several elements that is designed for moderately high work-loads, and some elements for extrmely high workloads. However, what it is not designed for is an "across the boards" extreme workload, such as memory, gpu, cuda, frame buffers, comms, all blazing at once, can overload the factory cooling of a GPU and cause heat related failure, even an EVGA GTX card with turbo fans. When a loadscreen is blazing at 2,000 fps, this is a huge strain on every element of the card (in fact it is used in some burn ins), and to say it is inadquately cooled is like saying your little 4 banger is inadequately cooled because you cracked the block while tryiing to pull a trailor up a hill in Death Vally. The 2,000 fps screens are simply placing a load on the GPU that it was never meant to handle.
Last edited by Chip Patton; Jan 5, 2016 @ 11:52pm
Thalyn Jan 6, 2016 @ 12:13am 
Except, as I said, it's not "across the board" - the load on a screen like that is almost entirely on the ROPs. Just about everything else (TMU's, CUDA cores - even the RAM controllers) are sitting basically idle. The ROPs represent just a small fraction of the overall transistor allocation and power usage, and cooling them with a system designed for moderate-to-high sustained load on every subsystem should be trivial at best.

You're more likely to cause damage to your card at lower framerates, simply because that's more likely to indicate that more subsystems are active and generating heat, eg FurMark. Even that's not a perfect indicator, however, as limits can be imposed and the GPU isn't always the bottlenecking factor.

Much like your 4-pot in Death Valley, you're not going to be moving very quickly with that trailor going up a hill. But you can also go light on the gas on dead-flat ground. The difference here which renders the analogy moot, however, is you'd still destroy the 4-pot going flat-stick with zero superfluous weight, where a GPU can handle that just fine (as evidenced by the OP's GPU reaching a mere 65ºc).
Chip Patton Jan 6, 2016 @ 12:16am 
Originally posted by Thalyn:
Except, as I said, it's not "across the board" - the load on a screen like that is almost entirely on the ROPs. Just about everything else (TMU's, CUDA cores - even the RAM controllers) are sitting basically idle. The ROPs represent just a small fraction of the overall transistor allocation and power usage, and cooling them with a system designed for moderate-to-high sustained load on every subsystem should be trivial at best.

You're more likely to cause damage to your card at lower framerates, simply because that's more likely to indicate that more subsystems are active and generating heat, eg FurMark. Even that's not a perfect indicator, however, as limits can be imposed and the GPU isn't always the bottlenecking factor.

Much like your 4-pot in Death Valley, you're not going to be moving very quickly with that trailor going up a hill. But you can also go light on the gas on dead-flat ground. The difference here which renders the analogy moot, however, is you'd still destroy the 4-pot going flat-stick with zero superfluous weight, where a GPU can handle that just fine (as evidenced by the OP's GPU reaching a mere 65ºc).

Not really, using Nvidia Inspector, everything is ramped up at the load screen, granted for me its at about 82%(ish), but that's still pretty damn high for a load screen, and toeing the line to what these 980's can take.
Thalyn Jan 6, 2016 @ 12:44am 
I've run my 980, overclocked (ASUS Strix factory cooling), continually at 95%+ (according to Afterburner) for in excess of an hour with no negative side effects. And that was definitely doing more than just spinning a small loading indicator on an otherwise motionless image - power usage was over 100% (limited to 125%), and the temperature had hit 83ºc (my self-imposed limit).

The issue comes with how that percentage is calculated, which no-one outside of NVidia really knows for sure. Does 82% mean 82% of all the card's resources are in use, or that the highest usage on any individual subsystem is 82%? Both are perfectly valid metrics, but I'd be reluctant to believe that a near-motionless screen could possibly be using 82% of all available resources (which would be matched by an 82%+ power usage measurement).

*Just testing now, loading without a limiter had me at anything from 2,000fps to 2,800fps (I'm transcoding as well at the moment, reducing maximum potential). GPU usage hovered in the high-50s, low-60s percentage, but GPU power usage was under half that (it admittedly did reach 50%-ish but the baseline is 20% for reasons I need to investigate).
Last edited by Thalyn; Jan 6, 2016 @ 12:56am
Silnar Jan 6, 2016 @ 2:02am 
It's a bit rediculous, SE is the only game in which I have to turn vsync on and it's only because of this menu issue.

In no other game menu does my card ever ramp up the fan to make it sound like a helecopter taking off. Even on Ultra settings in games like Planetside 2 its never like the SE menu.
Last edited by Silnar; Jan 6, 2016 @ 2:03am
Chip Patton Jan 6, 2016 @ 2:34am 
Originally posted by Thalyn:
*Just testing now, loading without a limiter had me at anything from 2,000fps to 2,800fps (I'm transcoding as well at the moment, reducing maximum potential). GPU usage hovered in the high-50s, low-60s percentage, but GPU power usage was under half that (it admittedly did reach 50%-ish but the baseline is 20% for reasons I need to investigate).

I think you're starting to see...

Simply put, it happens because the frame render is just so easy to do -- it takes 1,000th of the time to render a frame on the main menu as it does in the game itself. When the game doesn't impose a limit, the GPU will render as hard and fast as possible. This can be quite simply put by stating the analogy -- You have to complete both tasks individually in under 5 seconds: Move a 50lbs log 5 inches. Now in under 5 seconds move 118 decks of unwrapped, unboxed playing cards, each card individually 5 inches. Which do you think you'll overheat first in doing? 1 big log, or a few thousand playing cards?. You may think you're having no negative side-effects, but there are things that are overheating that the heatsync doesn't make contact with and very well may not have a thermometer, and with normal games, it doesn't need it.
Last edited by Chip Patton; Jan 6, 2016 @ 2:35am
< >
Showing 1-15 of 16 comments
Per page: 1530 50

Date Posted: Jan 5, 2016 @ 12:54am
Posts: 16