STEP Extended + Alternate Start fails "Harpalus bandit test" on reasonable hardware

[shaunlewis]

All, 

 

Firstly, thanks for the work on these guides, and please forgive me if I have posted in the wrong forum.

 

I have recently completed a marathon few days of Skyrim installation; my wife says, I am starting to think she is right, that I enjoy modding this game more than actually playing it! 

 

I had originally installed SR:LE, but while trying to add my own selection of other, non graphical, mods on top, it was clear my hardware wasn't going to cope too well. Therefore, through the magic of MO, I reverted to a STEP install without ENB, considering that might be a less challenging base on which to build on. I have installed a stock, as-per-the-guide STEP Extended install, with the sole addition of the Alternate Start - Live another Life mod. Before installing the LAL mod, I benchmarked my install, with the following results:

 

 

It appears that, graphically, STEP Extended isn't taxing my hardware too much, which is as I expect - my GPU is a couple of tiers higher than that recommended for my resolution according to Toms Hardware tier list, and it appears only my CPU (at 3.6GHz rather than "closer to 4.0GHz" in the background section) falls anywhere short of the "Beyond Step" recommendations.

 

I have however, before I commence with adding additional mods tested the stability of the install using harpalus' test, posted on Dreadflopps modular patches page (which I intend to use). All seems fine with the Solitude dragon spawning test and the bandit camp clearing test (I used KnifepointRidge01).

 

There seems to be some difficulty with the NPC spawning test however. I tgm and spawn 100 bandits in solitude. Skyrim quickly freezes after 3-4 seconds. I should note that, trying the same test with the minimum modded 'vanilla' I consider playable - that is; SKSE, ENBoost, USLEEP, SkyUI, Cleaned ESMs and Optimized Textures - results in nothing worse than an 20FPS drop while I hack every single one of them to death with an iron sword. 

 

Clearly, something in STEP is effecting my ability to spawn and fight large numbers of NPCs. My questions are:

 

• Should I consider my inability to pass this 'test' as something indicative of a problem with my install?
• Is this a hint that I will indeed need to upgrade my CPU to at least a 4GHz-er before I can continue building on it?
• Or can it be this test just hokum, and I can ignore this? I admit I havn't played a full play-through of Vanilla Skyrim and all the DLCs so I don't know how likely I am to run into such large spawns. 

I would be grateful for your advice on the matter.

Edited February 21, 2016 by shaunlewis

[Greg]

I think spawning 100 bandits is an unrealistic test that you should never encounter in the game -- unless you somehow manage to anger every NPC in Skyrim to the point that they come swarming for you head on a pike and assuming there are actually 100 NPCs in all of Skyrim. In reality, I don't think you can ever encounter much more than a dozen or so NPCs in combat. Even the Civil War quests are more like the Hatfields vs McCoys than an actual Civil War. With that many NPCs in active combat at once, it's more likely you are gumming up the Papyrus engine.

 

For a more realistic test, you might walk into the middle of Halted Stream Camp and the lunar forge area (can't remember the name of this one off the top of my head). If you want to spawn NPCs, limit it 20 or less.

[Nebulous112]

Try Memory Blocks Log. Install it, and do the test again. You may need to increase the heap size in skse.ini, but the log will tell you.

[Darth_mathias]

My bad throught this was SRLE Extended

Edited February 21, 2016 by Darth_mathias

[shaunlewis]

Thanks for the speedy reply.

 

Installing and repeating with Memory Blocks Log gives some respite - I dropped down to 5fps and it took about 90 seconds, rather than a few seconds, before it froze. I am confident that is because the issue is not completely reproducible rather than any improvement simply by installing the log plugin.

 

The log itself looks like this: 

 

logging of blocks enabled

logging max values only

Timer disabled

Block1 Block2

512MB 256MB

85 8

85 8

85 9

....

312 166

312 167

312 168

312 169

312 170

312 171

312 172

312 173

312 174

312 175

312 176

312 177

312 178

 

I believe that means that the SKSE Memory patch is working as intended, but correct me if I am wrong. I did try adding --forcesteamloader to the SKSE executable, but this results in no basic change. 

 

My SKSE.ini is:

[General]

ClearInvalidRegistrations=1

 

[Memory]

DefaultHeapInitialAllocMB=768

ScrapHeapSizeMB=256

 

[Display]

iTintTextureResolution=2048

 

and the memory section of enblocal.ini:

 

[MEMORY]

ExpandSystemMemoryX64=false

ReduceSystemMemoryUsage=true

DisableDriverMemoryManager=false

DisablePreloadToVRAM=false

EnableUnsafeMemoryHacks=false

ReservedMemorySizeMb=128

VideoMemorySizeMb=3714

EnableCompression=true

AutodetectVideoMemorySize=false

 

 

The 3714 comes from vramsizetest reported value of 4064 - 350 (for windows 8.1).

 

Am I mis-reading the log and I do indeed need to make changes to my ini files? 

 

Thanks for your assistance.

 

[shaunlewis]

Also, thanks Greg. I can confirm that spawning 20 bandits inside Halted Stream Camp does not cause any issues other than a 15FPS loss.

 

I suspected that the conditions of the test were unlikely to occur during normal gameplay, but I wondered if it was similar to the TCL and fly around at high speed test - not something ever likely to happen in game play, but something you want to be able to do without crashing, if I understand correctly; and of course, my vanilla-ish profile handled it just fine.

 

I guess that, unless someone can tell me I have made a mistake with my configuration above, the best thing to do is continue slowly at test by spawning 20 bandits, unless I start to hit freezes in actual gameplay.

Thank you for your help. 

[Nebulous112]

Your Memory Blocks log confirms the settings are fine, and your skse.ini and the Memory section of enblocal.ini is fine.

 

The forcesteamloader argument is not necessary for you.

 

It just comes down to what Greg mentioned...you'll never fight 100 enemies at the same time. What Greg is saying is 100% correct.

 

Edit: Didn't see your last comment. I think you are probably just fine!

[Greg]

I personally think the speed test is completely bogus as well because a crash to desktop is more likely to be caused by spamming the papyrus and rendering engine. If you think about it, you're literally causing cells to load at a very rapid rate so the rendering engine may not be able to keep up and this isn't taking into consideration the number of activators and other events (like spawning NPCs and creatures) that require scripts to run as you cross cell zones.