CPU upgrade....i5 or i7?

I'm in similar situation to you Maci but I need the constant 90fps for oculus (can't seem to drive as well with 45 fps), so with the same i5 2500k @4.something and a 980 I can maintain 90fps at quite low settings on most tracks, but does occasional dip in parts of track or when I'm taking an off road excursion

been thinking about new i5 or i7 K chip setup for a while, couldn't really justify the 30 odd percent gain for the money, would be interesting to hear from anyone with a rift had a similar setup to mine and has since upgraded to i7 setup

 

Just upgraded to an i7 7700k from an i7 2700k oc to 4.3 & a titan x pascal from a 980 ti.

 

A

And den?
Oculus or screens? Before/after frame rate? CPU load?

 

CPU? ... Here's a vote for i7 due to multi-threading.

Equally note the following where you can explicitly tell AC to use all or most threads for the cpu demands of large tracks & the OSRW expanded KS grid sizes (associated use of physics threads) (eg. 40 car SPA, or 30+ Nords endurance etc.)

\SteamLibrary\SteamApps\common\assettocorsa\system\cfg\assetto_corsa.ini

[PHYSICS_THREADING]
THREADS=-1 ; -1 = automatic , 0 = disabled , 1...N number of threads dedicated to physics

[THREADING]
USE_TIMER_PROCESS=-1 ; -1=Automatic, no timer process for CPU with more than 2 cores, timer process for dual cores. 0=Always off. 1=Always on.
SET_THREAD_AFFINITY_MASK=0 ; 0=Windows scheduler will decide where to run the threads. 1=Main thread assigned to Core0, physics thread assigned to Core1

 

Is there any advantage to tweaking this?

I'm running a 3930K (6 cores 12 threads) with slight overclock to 3.8Ghz on all cores

 

I did some benchmark tests to compare HT on and off using my i7 6700k @4.6ghz with a gtx 1080, 16GB DDR4

Here are the tests

HT on at 4k Max settings (except no motion blur or FXAA)

Code:

AC VERSION: 1.11.3 (x64)
POINTS: 11584
FPS: AVG=79 MIN=56 MAX=89 VARIANCE=0 CPU=27%

LOADING TIME: 13s
GPU: NVIDIA GeForce GTX 1080 (3840x2160)
OS-Version: 6.2.9200 () 0x100-0x1
CPU CORES: 8
FULLSCREEN: ON
AA:4X AF:16X SHDW:4096 BLUR:0
WORLD DETAIL: 5 SMOKE:5
PP: QLT:5 HDR:1 FXAA:0 GLR:5 DOF:5 RAYS:1 HEAT:1

HT on at 1080p Max settings (except no motion blur or FXAA)

Code:

AC VERSION: 1.11.3 (x64)
POINTS: 22401
FPS: AVG=152 MIN=66 MAX=180 VARIANCE=1 CPU=53%

LOADING TIME: 13s
GPU: NVIDIA GeForce GTX 1080 (1920x1080)
OS-Version: 6.2.9200 () 0x100-0x1
CPU CORES: 8
FULLSCREEN: ON
AA:4X AF:16X SHDW:4096 BLUR:0
WORLD DETAIL: 5 SMOKE:5
PP: QLT:5 HDR:1 FXAA:0 GLR:5 DOF:5 RAYS:1 HEAT:1

HT on at 1080p min setting (To eliminate GPU bottleneck and test CPU better)

Code:

AC VERSION: 1.11.3 (x64)
POINTS: 59048
FPS: AVG=403 MIN=196 MAX=562 VARIANCE=21 CPU=86%

LOADING TIME: 12s
GPU: NVIDIA GeForce GTX 1080 (1920x1080)
OS-Version: 6.2.9200 () 0x100-0x1
CPU CORES: 8
FULLSCREEN: ON
AA:1X AF:0X SHDW:512 BLUR:0
WORLD DETAIL: 0 SMOKE:0
PP: QLT:5 HDR:0 FXAA:0 GLR:5 DOF:5 RAYS:1 HEAT:1

Here are the tests with everything the same but just Hyperthreading turned off.

HT off at 4k Max settings (except no motion blur or FXAA)

Code:

AC VERSION: 1.11.3 (x64)
POINTS: 11336
FPS: AVG=77 MIN=34 MAX=237 VARIANCE=0 CPU=28%
LOADING TIME: 18s
GPU: NVIDIA GeForce GTX 1080 (3840x2160)
OS-Version: 6.2.9200 () 0x100-0x1
CPU CORES: 4
FULLSCREEN: ON
AA:4X AF:16X SHDW:4096 BLUR:0
WORLD DETAIL: 5 SMOKE:5
PP: QLT:5 HDR:1 FXAA:0 GLR:5 DOF:5 RAYS:1 HEAT:1

HT off at 1080p Max settings (except no motion blur or FXAA)

Code:

AC VERSION: 1.11.3 (x64)
POINTS: 21968
FPS: AVG=149 MIN=65 MAX=187 VARIANCE=1 CPU=53%
LOADING TIME: 14s
GPU: NVIDIA GeForce GTX 1080 (1920x1080)
OS-Version: 6.2.9200 () 0x100-0x1
CPU CORES: 4
FULLSCREEN: ON
AA:4X AF:16X SHDW:4096 BLUR:0
WORLD DETAIL: 5 SMOKE:5
PP: QLT:5 HDR:1 FXAA:0 GLR:5 DOF:5 RAYS:1 HEAT:1

HT off at 1080p min setting (To eliminate GPU bottleneck and test CPU better)

Code:

AC VERSION: 1.11.3 (x64)
POINTS: 59845
FPS: AVG=408 MIN=197 MAX=576 VARIANCE=23 CPU=86%
LOADING TIME: 12s
GPU: NVIDIA GeForce GTX 1080 (1920x1080)
OS-Version: 6.2.9200 () 0x100-0x1
CPU CORES: 4
FULLSCREEN: ON
AA:1X AF:0X SHDW:512 BLUR:0
WORLD DETAIL: 0 SMOKE:0
PP: QLT:5 HDR:0 FXAA:0 GLR:5 DOF:5 RAYS:1 HEAT:1

Going by the results at maxed settings there was a slight gain with HT enabled but surprisingly with the min settings at 1080 there was a tiny gain with HT off, however at those high fps the difference is so minor that the results would be too close to call in that test.

 

ATM I use a i7 3770k and a GTX1050 low profile. Previously there was a GTX750 low profile built in. With the old GC the CPU got a load about 40% and graphics settings from mid to high. With the 1050 and the same graphics settings the CPU load has been 72% and the FPS raised from average 56 to average 141. This tells me that my i7 3770k isn´t at it max performance yet. Maybe if I will build in a GTX 1080 or better the CPU will be the bottleneck one day. But this would be more power than I ever need for Assetto Corsa (I only play racing games, and only one in a period of years). Even for VR.
So I bought a CPU in late 2013, which got enough power to fit my needs for 5 years for sure. The main reason for this CPU was the low power consumption, which produces less heat. Less power and heat implies a longer life-cycle for all components as well. I follow this principle for all my PC components and I never regret it.

So my advice would be: Buy a good CPU with lowest possible power consumption.

 

Thanks a bunch for that!!

I've just upgraded my Gfx to a 1080, and was unsure if it would be powerful enough to run AC at 4K.
Averaging 79 FPS in the benchmark on max settings would just cut it for me, I may reduce smoke or shadow reflection update frequency slightly to bring min FPS up to around 70.
I think I can risk getting a 4K TV.
One question: What clocks were you running your 1080 at in these benchmarks?

 

The core is running at around 2000mhz with memory at 5500mhz x2, the min fps in the benchmark really is a bit off as this seems to always be while the bench is loading. Once it is actually on screen the fps was fairly close to average in all the tests with no dips and never do I see it close to the min fps. I think in the 4k benches the lowest I saw the FPS drop to once it loaded was about 76.

To get some extra FPS turning down some reflection settings will be a considerable boost to get the min fps a bit higher but I just ran tests that way to stress the GPU in the max settings to see how it is effected by the i7 with its HT and then the min settings for pure cpu performance.

It doesnt look like there would be a big difference between an i7 or i5 running at the same clocks as HT doesnt seem to make much difference. If doing more stuff in background such as teamspeak and possibly 3rd party apps using CPU then the more threads could be a benefit same as some extra background programs. But with minimal background apps running it doesnt seem to be a significant difference having hyperthreading on a quad core cpu.

 

Thank you for your tests...fantastic help for me. You said right about backround apps...in AC there are many apps (in particular third part ones) that are harmful for framerate...maybe an i7 can manage better these situations

 

While the K chips are theoretically for overclocking with their unlocked multiplier they also come with a decently higher stock clock speed as well so wouldn't discount them, generally only a few $ more

 

Same here. Set at 2100/5570 and in game clocks around 2050-2080. I've silently hoped you would be running at stock speeds, but it is very promising still, specially if the perceived min FPS is higher during the actual run, as you state.
By itself 60 FPS or a frametime of 17 ms are mighty fine, but in situations where the FPS actually drop from e.g 100 down to e.g 60 chances are high that the occasional frame in between takes up to e.g 30-40ms to render, and this stutters are inconvenient.
To avoid such, I try to keep min frames at around 70 at least, and driving begins to feel natural for me at around 80-90 average FPS

Now I only need to decide if I should choose a 60hz TV or opt for a native 100/120hz model to finally see all those frames that are being rendered...decisions, decisions
Wish there were 21:9 TVs on the market...

 

Why not jump on the VR train and get a rift? The immersion is epic

 

Maybe at a later time, I'm not yet comfortable with VR kind of completely disconnecting me from reality, not only with regards to my wheel, shifter, buttons, and the keyboard, but also I would not be able to notice anymore whats going on around me, wife, kids, the lot.

 

Hehe, for me this is the main positive aspect of VR .

mobile greets, Wolfgang

 

lol, its not that I dont want to, its more I don't dare to...

 

Since I'm recording via NVENC from the same PC the actual Core CPU may differ a bit. The game is locked at 60fps with vsync to have a better idea of the actual CPU usage on each thread.


If you leave the FPS unlocked, obviously the CPU usage will sky rocket as long as the GPU can still draw more frames. Max setting with SD(480P) resolution on a slow CPU will not maintain 60fps even with the best GPU on the market.

For a full grid, I suppose having more cores is better for AC. The CPU is overclocked at 4.3GHz.

 

Haha, yeah, can understand that. Some weekends I think ill do a couple of laps in the evening and by the time I decide to call it it's 3am... time disappears in VR. Kind of wish the in car clocks in the cars that have them worked

 

While I'm open to having it pointed out, that it could be placebo, - i want to think and feel that extra threads and hyper threading really comes into its own with oversized grids. See the the OSRW KS-modded tracks (eg. 40 car Spa, 32+ Nords endurance/cup etc). The notion being that individual cars each have their own associated physics calculations and required physics threads. Increases in thread quantity should ease CPU stress when grid size goes far in excess of modest numbers (re: the quantity found in the benchmark). However with small grid sizes, the benefit of the extra threads may diminish in contrast. Oh, they'll be useful but only to a minor degree.

I'll even admit to salivating at a 'project' pc rig concept involving use of slightly older server grade multi CPU (chip) mobo's/cpu's to run genuine setups with 16core/32-hyper threads to drive huge modded KS-track grids. For instance running a SPA race with 65 AI would be fun. Particularly if a 4th gen i7 will happily support 40 cars with 8 threads.

 

And den i post here about it. not had chance to test it sorry. so will try have a play tonight. unfortunately i did not bench mark my previous build. If there is anything specific you want me to test happy to do that. Im using oculus but be aware i still use my old monitor an old 1280 x 1050 for other stuff. not sure what monitor to upgrade to but obviously using VR where i can.