Skyrim CPU OC Scaling?

CPU utilization differs from area to area, I think the AI of the NPCs and enemies accounts for some of it, how many physics objects are being simulated and things like that, so it does differ, I'm sure there's places with less than approx 50% utilization.

It's also a hard thing to measure accurately, it was more than 50% on some cores and less on others, 50% is an approximate across all 4 cores and doesn't take into account the lack of hyperthreading use etc. It seems to behave like it has one master thread which runs at a higher usage than the secondary threads. It seems likely that whatever thread is in charge of the render/update cycle in the engine would be a bottleneck and whenever that thread maxes out the core it's running on the whole engine would be CPU limited, irrelevant of what is happening on the other threads.

It's very possible due to lack of threading optimisation (which is incredibly common in games) that the CPU will never be able to run 100% CPU usage or even close to it. However in these circumstance where the CPU becomes the bottleneck you would expect the GPU usage to start dropping and becoming erratic, and that's not happening in my tests, GPU usage is locked at 99% which is classic sign of it being the bottleneck.

As I explained prior, simply demonstrating task manager to be a bad way of measureing CPU performance is only half the battle, you'd have to explain why the GPU was also fully loaded, in order to reasonably support the conclusion that in this specific circumstance the game was CPU limited and not GPU limited.
 
Are you saying you have Skyrim running across 4 cores? I thought it was limited to only ever using two cores? What am I not understanding?
 
Skyrim is using all 4 physical cores but is not touching the hyperthreaded cores.

Some people made the assertion that it was dual core only and that turned out to be false, I had read that in a few places before actually testing for myself and stated it on this forum and corrected myself once I'd actually checked.
 
Skyrim is using all 4 physical cores but is not touching the hyperthreaded cores.

Some people made the assertion that it was dual core only and that turned out to be false, I had read that in a few places before actually testing for myself and stated it on this forum and corrected myself once I'd actually checked.

When I posted those benchmarks from walking around Riverwood I also tested with two cores. Here is what I saw.

4 cores

2011-11-13 18:56:52 - TESV
Frames: 1831 - Time: 36489ms - Avg: 50.180 - Min: 41 - Max: 69

2011-11-13 18:57:57 - TESV
Frames: 1234 - Time: 25334ms - Avg: 48.709 - Min: 38 - Max: 62

2011-11-13 18:58:57 - TESV
Frames: 1723 - Time: 34741ms - Avg: 49.596 - Min: 41 - Max: 64

2011-11-13 19:00:08 - TESV
Frames: 1852 - Time: 39905ms - Avg: 46.410 - Min: 39 - Max: 59

2 cores

2011-11-13 19:05:22 - TESV
Frames: 1567 - Time: 33431ms - Avg: 46.873 - Min: 39 - Max: 64

2011-11-13 19:06:22 - TESV
Frames: 1384 - Time: 32183ms - Avg: 43.004 - Min: 35 - Max: 53

2011-11-13 19:07:26 - TESV
Frames: 1442 - Time: 33867ms - Avg: 42.578 - Min: 36 - Max: 60

2011-11-13 19:08:28 - TESV
Frames: 1480 - Time: 32776ms - Avg: 45.155 - Min: 35 - Max: 57

The article they test in 1680x1050, your sig says you're using a 1920x1080 monitor, having a higher resolution causes a higher load on the GPU with a similar load on the CPU and so tends the game towards being more GPU limited.

The more GPU limited you become the less important changes to CPU speed become when you finally become completely GPU limited changing the CPU speed makes next to no difference.

Any proof to back that statement up?

I know that I have seen people running more mods in games like Oblivion that I ever could with a crappier computer than I use but at a lower resolution. The resolution has an effect on the cpu in this game.

Again this is from a poster who's opinion that I trust.

i have it but don't oc much :LOL:
but i did get about 10fps more going from my i7 930 to a 3930k both at stock with the same drivers at 1600p
i will test oc on it later on

Just saw this over at the same Rage3d thread.

Given that the game isn't stressing a single GTX 570 and is being held back by an i7-875K (should be a little faster than a 920) in my case, SLI wouldn't do much for you anyway.

I overclocked my CPU to 3.6GHz last night and that got me 5-10fps in shadow heavy areas (towns mostly).

The game isn't dropping below 45fps anywhere for me now, but unfortunately it's still a bit noticeable when I walk into a town or over a big expanse and all the sudden the framerate drops from 60 to 45. I can deal with it though, not enough to make we want a new processor at this point.

It's likely never going to happen, but I really wish they would put out a patch to do shadows on the GPU; shadows on CPU is for consoles that have low end GPUs.
 
Last edited:
That's about what I'd expect to be honest, while it uses 4 cores it's certainly not maxing them even once I dump the graphics settings down to minimum, mine was about 85% usage on the CPU and was at low GPU usage so no longer GPU bottlenecked.

There's room for optimization on the CPU for sure by having all of the threads (presumably 4?) running on 2 cores rather than 4 you optimize the usage of each core, not enough to maintain the same frame rate as 4 cores but enough to get pretty close. Fairly typical for games these days, they're only usually threaded to a basic degree like physics in 1 thread, AI in another thread, it's never usually more fine grain than that.
 
i tested my system an AMD 5000+ 4Gb memory GTX 550Ti
average FPS 36 tested with Fraps. OCing to 3.1Ghz just becomes unstable. it is time for an upgrade on this rig though. the motherboard has become unstable as the sound and networks ports on back are starting to work and not work and any over clock over stock has become very unstable.

i am running at a quality setting of High with 8x AA and 8x AF everything plays ok unless there is a lot going on.

this is also the most intense game i enjoy on my TV lol.
 
I would like to report that after bumping my 2500k from 3.3ghz to 4.5Ghz there was a 8fps improvement on the min fps.
 
Back
Top