Skyrim VRAM usage

[blitzen]

I've been having some issues with game freezing, CTDs, and infinite exterior loading. I thought I had resolved it, but apparently not. As Skyrim is running, it eventually fills up the available VRAM. It usually hovers around 2.3 - 2.4 GB of VRAM usage. At some point it will start to use more than 2.5 GB (that's how much VRAM the video card has), and the game will lockup or crash. I run with a single monitor at a resolution of 1600x900, so that should reduce how much VRAM is needed compared to running the game at 1920x1080. That seems a little high. I don't have any super hires textures. I was running with mountain textures at 4096, full SFO detail, and full Enhanced Blood Textures, but I reduced the mountains/rocks to 2048, used lower res trees for the floral ovehaul, and used 1024 blood textures (my game is very bloody, lots of fighting), but that didn't help much.

 

I was expecting maybe 2 GB max based on what I was seeing while running Skyrim in the past. This problem started to occur within the last 3 weeks, and I haven't installed any new texture replacers during that time period, only new content mods. I'm puzzled. If you don't know what your VRAM usage is, now is a good time to check. Download a tool such as GPU-Z and see what's going on with your video card, especially if your game is crashing, locking up, or stuck on loading screens.

 

How much VRAM usage are other people seeing?

Edited May 22, 2013 by blitzen

[BotOwned]

  On 5/22/2013 at 3:18 AM, blitzen said:

I've been having some issues with game freezing, CTDs, and infinite exterior loading. I thought I had resolved it, but apparently not. As Skyrim is running, it eventually fills up the available VRAM. It usually hovers around 2.3 - 2.4 GB of VRAM usage. At some point it will start to use more than 2.5 GB (that's how much VRAM the video card has), and the game will lockup or crash. I run with a single monitor at a resolution of 1600x900, so that should reduce how much VRAM is needed compared to running the game at 1920x1080. That seems a little high. I don't have any super hires textures. I was running with mountain textures at 4069, full SFO detail, and full Enhanced Blood Textures, but I reduced the mountains/rocks to 2048, used lower res trees for the floral ovehauls, and used 1024 blood textures (my game is very bloody, lots of fighting), but that didn't help much.

 

This seems a little high to me. I was expecting maybe 2 GB based on what I was seeing while running Skyrim in the past. This problem started to occur within the last 3 weeks, and I haven't installed any new texture replacers during that time period, only new content mods. I'm puzzled. If you don't know what your VRAM usage is, now is a good time to check. Download a tool such as GPU-Z and see what's going on with your video card, especially if your game is crashing, locking up, or stuck on loading screens.

 

How much VRAM usage are other people seeing?

 

Always at 1024MB for mine with a shared memory usage of between 700MB and 2.5GB which is to be expected since Cells and models are kept in a buffer for quicker loading times. < Nvm that was an observation not actual fact.

 

Anyway I rarely ever crash and I run with approx 200 mods, 15 of which are hi-res texture packs. I have rarely ever locked up and have a constant fps of 60 inside and 40 outside.

Edited May 22, 2013 by BotOwned

[Reynard131]

  On 5/22/2013 at 3:18 AM, blitzen said:

I've been having some issues with game freezing, CTDs, and infinite exterior loading. I thought I had resolved it, but apparently not. As Skyrim is running, it eventually fills up the available VRAM. It usually hovers around 2.3 - 2.4 GB of VRAM usage. At some point it will start to use more than 2.5 GB (that's how much VRAM the video card has), and the game will lockup or crash. I run with a single monitor at a resolution of 1600x900, so that should reduce how much VRAM is needed compared to running the game at 1920x1080. That seems a little high. I don't have any super hires textures. I was running with mountain textures at 4069, full SFO detail, and full Enhanced Blood Textures, but I reduced the mountains/rocks to 2048, used lower res trees for the floral ovehauls, and used 1024 blood textures (my game is very bloody, lots of fighting), but that didn't help much.

 

This seems a little high to me. I was expecting maybe 2 GB based on what I was seeing while running Skyrim in the past. This problem started to occur within the last 3 weeks, and I haven't installed any new texture replacers during that time period, only new content mods. I'm puzzled. If you don't know what your VRAM usage is, now is a good time to check. Download a tool such as GPU-Z and see what's going on with your video card, especially if your game is crashing, locking up, or stuck on loading screens.

 

How much VRAM usage are other people seeing?

 

 

VRAM usage means nothing.... it is a normal function of the drivers...

 

your card is on its way out especially if you have not kept it clean, and or kept the fan spun up the VRAM is about 1/4" from an 80 deg C GPU.... the ambient temps are rising..... death to VRAM...

 

http://www.ixbt.com/video3/images/ref/gtx570-scan-front.jpg

Edited May 22, 2013 by Reynard131

[prod80]

@Reynard131;

That's a pretty harsh statement, lol.... can you elaborate as to why his card is "to an end" ? You buy a new card as soon as a game CTD? :wink: For the last 15 years Ive been running 80 C on my GFX cards, yet havent ever dropped a card away because it was dead, just because it was obsolete and couldnt run the games fast enough anymore. You think if the temps were a problem the GFX card manufactures would put the GDDR chips right next to the GPU these days... they do it because the chips can handle it, and the performance is better. It's perfectly fine these chips are heated up.

 

@OP;

Your game can CTD due to VRAM usage... but in that case it must be severely limited and you would see a lot of signs about your VRAM not coping with the load ( a lot of missing textures ). When the game engine is efficient it should always keep your VRAM filled up with textures it might need shortly ( last visited Cells ) to improve load times. This is normal. Crashes are mostly not related to VRAM but rather your system RAM running out, or simply the limitations of the game engine reached/broken. Running out of VRAM generally just causes severe FPS drops, missing textures, extremely long loading times.

 

You state that you use more than 2.5GB of VRAM and you downgraded to 2K/1K texture versions ( which is fine ) yet the problem did not resolve itself. Have you by any chance did any tweaks to your game's INI files? If so, then what did you change? Also content mods add, content. So yes that increases resource usage as well ( also textures ).

 

As to your question; I run a heavy modded game on 1GB VRAM... if you think that must look crap, see my gallery pictures ( just random pictures not "staged for screenshots and edited in Photoshop afterwards" pictures, just how it actually looks )

 

Or here are some;

http://static.skyrim.nexusmods.com/images/6440158-1368998157.jpg

http://static.skyrim.nexusmods.com/images/6440158-1368904015.jpg

http://static.skyrim.nexusmods.com/images/6440158-1368999821.jpg

 

[blitzen]

<p>@Reynard:  The card runs 60 - 65 degrees under load, the fan speed is adjusting itself nicely and does not seem to have any problems maintaining reasonable operating temperatures. Although there are certainly times where video card hardware problems begin to surface for some people, I don't believe that is what is happening for me at this point. It's something else. </p>
<p> </p>
<p>@prod80: Thanks for the input. Nice screenshots. Your game looks very good, especially when considering your hardware. I'm wondering what mods you are using. I don't think it's RAM, since I've checked that several times and it hasn't used more than 3 GB, and it's usually safely below that. It can use up to 4 GB, and the computer has 8 GB.</p>
<p> </p>
<p>The VRAM usage looked suspicious to me. I haven't been monitoring it for a while, but last time I did Skyrim was using around 1.2 GB of RAM. This was probably 6 months ago. I have been running GPU-Z consistently over the last several days due to the problems I've been having. There does seem to be a correlation between crashing/lockups and VRAM usage, but I can't be sure it's always full when something happens, and it does tend to spend a lot of time running at 2.4 GB, which is close to the limit. I recognize that an apparent correlation with a small data sample does not necessarily imply a cause and effect relationship. I'm hoping to get some input from the community on this. It's hard to draw any conclusions with a single data point.</p>
<p> </p>
<p>I use NMM to install mods. I can see each mod that I have installed since the problems started to occur (I can sort by install date), and I don't see anything that is likely to create this type of problem. It's possible that it has nothing to do with VRAM, but there have been a number of reports by people, since the release of Skyrim, about crashing due to VRAM filling up. I also run with Papyrus logging on, and I check the logs regularly. I make new characters often, so I'm not running with old saves filled with types and scripts that are baked in by old mods, which can destabilize a game. I just don't see a good reason for Skyrim to be having issues right now. Yet it is.</p>

Edited May 22, 2013 by blitzen

[prod80]

"I don't think it's RAM, since I've checked that several times and it hasn't used more than 3 MB [guess this should be GB], and it's usually safely below that. It can use up to 4 GB, and the computer has 8 GB."

Skyrim is a 32 bit application, and yes in theory it can use 4GB, however in reality it will CTD immediately when you go over 3.1GB, presumably 3.2GB. And dont forget that your system RAM is also used by you GFX card.

As for lockups...
...Good you already changed to Lite versions of the texture mods, you have a 1080 screen? 1920x1080 = ~2K ... so having a higher resolution than that will only benfit you when you wanna stand with your nose on it and still not see pixels - now how many times will that happen to any other objects besides faces? It doesn't. So use 2K textures when you can and lower res for smaller objects (note: obviously when an object can fit your screen >1 times [edit: in real size, not just because you can get closer] (eg. Dragons) you need higher textures. As in this example Dragons are ~5x your screen size, so with a 2K screen you need at least 10K textures to be able to view 'pixel-less' Dragons on 100% viewing distance)
...If you have things installed which have been tagged as being responsible for lockups multiple times, uninstall those ... a widely used mod that causes load screen to hang for some users is "A Quality World Map" .. so, if you have that, try and uninstall it.
...Skyrim.ini tweaks can make the game unstable, specifically uGridsToLoad tweaks cause issues in a lot of occassions. The default setting is 5, and actually it should stay at 5 if you want stability. The skyrim engine is not build around people tweaking it, and loading a zillion texture mods after it. There is a reason this setting is not included in the default ini file after installation. Increasing uGrids should only be used for some specific scenery screenshotting, not for playing.
...If your load screen hangs when there is a save point just outside, see if disabling auto-save completely in options resolves it. Use manual save instead (just dont forget to save then).
...Obviously if you have overclocked your CPU/GPU reduce it. I had to reduce overclocking as well by quite a bit... to give you an example... by CPU normally runs on 4.4Ghz (i5-2500K) but with Skyrim it will crash and I must reduce it to 4Ghz. My GPU can reach 932Mhz core speeds, but with Skyrim I get crashes/corruption with anything over 866Mhz. Especially your system RAM overclocks can cause issues, I'd suggest to use factory settings and not the maximum capable. Skyrim might be old, but when modded is very demanding and stressing on your system... more stress + overclock = less stability


Check these out, and I hope one of them fixes this issue.
cheers

 

 

PS. "I'm wondering what mods you are using"

 

What you see in those screenshots are mainly;

 

Realistic Lighting Overhaul

Climates of Tamriel

Skyrim HD Lite

Snow textures and shaders

Various smaller texture replacers for all objects, certain objects, smoke, etc, etc... too many :o

All NPCs are changed, all skin textures are changed, all armors, weapons, animals, etc, etc

Realistic Fire

Project Parallax

Project ENB which I have changed (quite a bit by now - not too much left of it)

Many people asking for this... I'm trying to get a list of NMM with all installed mods (regardless if they come with a plugin) but NMM doesnt seem to be able to extract this :(

[blitzen]

RAM usage is hovering at around 2.8 GB for TESV.exe, according to the windows Task Manager. It should be ok. I never mess with ugrids. I've been using A Quality World Map for ages. I'm relatively experienced with using mods for both Oblivion and Skyrim. The main point is not so much what I'm running with in general, but what the heck changed about 3 weeks ago that caused a noticeable destabilization. I just can't account for it.

 

I looked at the VRAM and thought it wouldn't hurt to get info from others. You may be correct in that the game engine is smart enough to make use of whatever is available, so having it fill up is normal. GPU-Z reported a GPU load of 80% at the last crash, when VRAM usage was maxed, so the card's not even fully stressed. The last crash occurred when looking at the inventory. It's more likely to crash when interacting with UI elements, but it can crash at apparently random times. I'm not seeing a load on either CPU or GPU that would indicate a hardware limitation. I do run with about 250 .esp/.esm files installed, but I've been doing that for months so, again, it's not so much what I'm doing but what I'm doing differently from a month ago when everything was working well. I haven't overclocked anything on the computer at any point. The computer was buiIt by a gaming system company and I instructed them to not overclock a single thing, and I upgraded power supply and cooling to ensure maximum stability.

 

I have made quite a few .ini changes over time, but nothing dramatic, and I don't believe I changed anything in there around the time the game destabilized. I have had autosaving enabled the entire time I've been playing, during which I've had periods of horrendous instability, and periods of very solid, reliable performance. I don't consider that a factor, and am skeptical about the advice to turn that off, in spite of the fact that many people give that advice. I know there have been times where the game was much more likely to crash quite soon after saving, so there may be issues with saving in general, but I haven't seen any difference between auto saving and explicit saving in years of using Oblivion, FO3, FNV, and Skyrim, and have always left it on for each of those 4 games.

[prod80]

Well, you know... there's nothing wrong in giving it a try... :smile:

It's not like a game-change that you cannot revert back very easily and/or could potentially break your save (disable QWM and/or Autosaves)...

As what has changed.. god I wouldn't know... I assume you already run all the little things a modded game should be running (eg. SKSE.ini edit to clear invalid registrations and the likes)... It's very hard to pinpoint such a thing. With most people it's pretty obvious, been doing quite a bit of troubleshooting here in the past weeks. Maybe you can post your load order, perhaps anyone can point out a mod that has been causing issues. It's always worth a shot. It can also be you've changed your Papyrus settings away from defaults while using SKSE... this also has potential to break stuff eventually according to the SKSE developer.

 

Exterior crashes are mostly related to the game expecting something somewhere, but not getting it or conflicting (removed object/mod incompatibility - this kind of crash should be more or less repeatable) or your system, more likely the game engine, being overloaded and simply saying "no more!" (which is a more random event)

 

Load screens, I've seen a lot of issues related to what I already said about QWM and Autosaves, and of course a simple overload of textures which your system just cant handle at the time (either size related, or the quantity to load). I have noticed on my VRAM limited system that at times my load screen could take more than a minute for actually no appearant reason... I thought it hanged... wanted to ctrl-alt-del and pop- there it was. Next time, same load screen and < 5 seconds it went through.... specifically that seems to happen in Markarth and Solitude for me, I guess there are some big*ss objects to load over there causing this, which sometimes must be reloaded into an already cramped 1GB card, and sometimes are still buffered and are loaded a lot faster.

[Reynard131]

  On 5/22/2013 at 9:52 AM, prod80 said:

@Reynard131;

That's a pretty harsh statement, lol.... can you elaborate as to why his card is "to an end" ? You buy a new card as soon as a game CTD? :wink: For the last 15 years Ive been running 80 C on my GFX cards, yet havent ever dropped a card away because it was dead, just because it was obsolete and couldnt run the games fast enough anymore. You think if the temps were a problem the GFX card manufactures would put the GDDR chips right next to the GPU these days... they do it because the chips can handle it, and the performance is better. It's perfectly fine these chips are heated up.

 

Or here are some;

http://static.skyrim.nexusmods.com/images/6440158-1368998157.jpg

http://static.skyrim.nexusmods.com/images/6440158-1368904015.jpg

http://static.skyrim.nexusmods.com/images/6440158-1368999821.jpg

 

 

 

 

The VRAM is a FIFO buffer for the GPU the FIFO buffer throughput is part of driver functionality the driver determines the Hardware VRAM Size and utilizes it based upon given algorithms specific to the GPU, and Hardware VRAM.... Data integrity in this process is very important as it truly determines what, and how it is displayed.

 

If you have something empirical, that shows VRAM usage.. Would be interested in seeing, but most are anecdotal generality's...

 

Having more VRAM does not effect frame rates. except at very high resolution IE: 2560 x 1600 and above and that effect is like 1-2 frames..

 

Hiccups and glitches are mostly due to errors of one type or another as operating temperatures are driven up, the electron leakage is increased by a number of factors this contributes to errors and there are typically not an ECC environment everything is unbuffered in most desktops

 

Semiconductors are a commodity and the interoperability specifications are laid down by JEDEC http://www.jedec.org/ Joint Electron Devices Engineering Council they specify the timing/binning form factors and operating conditions along with all kinds of other standards.

 

GDDR5 has a temperature sensor that is built in that that per the JEDEC spec adjust the interface timings. If the sensor does not function correctly this creates errors.....

 

 

 

  Quote

Hynix GDDR5 Tech Manual.

 

5.22. TEMPERATURE SENSOR

GDDR5 SGRAMs incorporate a temperature sensor with digital temperature readout function. This function

allows the controller to monitor the GDDR5 SGRAM die’s junction temperature and use this information

to make sure the device is operated within the specified temperature range or to adjust interface

timings relative to temperature changes over time.

The temperature sensor is enabled by bit A6 in Mode Register 7 (MR7). In this case the temperature readout

is valid after tTSEN. Hynix applies 10us to tTSEN.

The temperature readout uses the DRAM Info mode feature. The digital value is driven asynchronously

on the DQ bus following the MRS command to Mode Register 3 (MR3) that sets bit A7 to 1 and bit A6 to 0.

The temperature readout will be continuously driven until an MRS command sets both bits to 0.

The GDDR5 SGRAM’s junction temperature is linearly encoded as shown in Table 35. Hynix has the readout

to a subset of six digital codes out of Table 35, corresponding to six temperature thresholds.

Table 34 Temperature Sensor Readout Pattern

Temperature [°C]

Binary Temperature Readout

MF=0:

MF=1:

DQ[5:0]

DQ[31:26]

＜ 45 000000

55 000001

65 000011

75 000111

85 001111

95 011111

＞ 95

 

See Page 123.

 

www.hynix.com%2Fdatasheet%2Fpdf%2Fgraphics%2FH5GQ1H24AFR(Rev1.0).pdf&ei=aPScUbfvE8bmyQHJjYDwDA&usg=AFQjCNGERRGoFmuk7O73OTtMch9OmlRKDw&sig2=RdFHE-GYA8R9n7WpR9OGYA

[prod80]

Yes, so.. his card is still well below any temps which should be considered alarming. Failing hardware is mostly a lot more visible then just an occasional CTD, certainly if it's failing memory as that often shows as graphic corruption before crashing completely (of course not always) and would likely crash the whole system when it fails requiring a complete reboot. Yes heat causes damage, electrical current causes damage... connections go old fast... lotta things go bad over time. But don't state immediately;

 

"your card is on its way out"

 

without any diagnostic as to where the infinite load screens or CTDs come from. You need a lot more diagnostic to determine a card is about to die then 1 forum post as the OP. And besides that, it's about VRAM usage as he's obviously looking for a solution by himself and just asks input from here. I do not doubt he knows enough about HW/SW to determine where the issue comes from himself and solve it when found. He already says he has been monitoring his hardware for the past weeks to find a relation, and noticed his CTDs seem to happen when his VRAM fills up.

 

and

 

"Having more VRAM does not effect frame rates"

 

Correct when everything is loaded into VRAM that the scene needs to render, but when you need a lot of textures loaded after each other you will have "stutters" where they are loaded - because hey; less VRAM means less can be loaded simultaneously, the game engine cannot take advantage of the extra space it has, and if you need a LOT more textures loaded in high res and you have a low VRAM card (like me @ 1GB) you receive a heavy penalty because it cannot place/swap everything it needs into memory at sufficient speed.

 

You can debate it all you want but fact is if I remove my high res texture replacers for lands, mountains and water, I go from 27-30 FPS to >40 FPS in outside world.