GPU < 50% load when CPU at 100%

[Stunt-Monkey]

So...
Im running a 4790K @4.6 and a GTX 980Ti at stock, its coming up to winter here in Perth Australia so decided to stress test my system for temps for possible overclocks (almost impossible in Australian summers lol, anyway i digress :P

I used Furmark for gpu stress and Intel Extreme Tuning for cpu, but when i run IET or any cpu stress tests it makes my gpu load less than 50% even when it was currently running at 100%

Any idea why this is, my guess is that the program used to stress gpu is just not being given enough cpu processes to keep running the tests 100% efficiently?

Hmmmm

[Prowlcorp]

Sounds like that could be it

[_I_]

could be an unstable overclock

check with cpuz while the cpu is loaded
if its not staying at 4.6ghz its throttling for some reason
most likely cpu or vrms overheating

[Bad 💀 Motha]

Also, regardless of CPU loads, a higher-end GPU like 980 Ti should really not be hitting loads above 50-60% @ around 1080p. Unless perhaps the game or benchmark is not well optimized and just ends up making the GPU hit a high % regardless of settings and res, etc.

Such a GPU as this would have plenty of breathing room to run say 1440p or even 2160p; so for 1080, 50% loads on GPU alone are fine and normal.

Now for CPU, stress tests and things of this nature will use 100% CPU, that is the whole point of them. To fully stress and bench your CPU. Now if you seeing that kind of CPU load in games during normal gameplay (as CPUs can hit 100% load during times of loading screens/maps) then I would say the CPU is set to something that is not stable. Or the CPU is throttling due to heat, or the VRMs are unstable/hot, forcing a CPU-throttle during peak loads.

[Stunt-Monkey]

So both run stable at 100% load when running seperate tests, its only when both run together that the the gpu with throuttle to around 50-60% when both are running together, cpu stays at 100% when both tests are running. Ill test again with stock cpu settings. Is it possible that the motherboard maybe bottlenecking the gpu when cpu is at 100%. Motherboard is the Sabertooth Mark S

[Bad 💀 Motha]

I only suggest that both be tested "stress" wise at the same time so a user can judge the max temps of both. As when u reach these in stress related tests, they should reach much higher then you'd ever see within any games.

However looking the loads; often times you will feel as though something is wrong if you try stressing both at the same time (for example, running a Prime95 CPU test & FurMark GPU test together side by side). One or the other will usually prevent one from reaching max load.

I would not bother looking at loads on these kinds of tests, especially when ran together. The focus on such tests should be: do they stay cooled while under such stress for lengthy periods of 30-40 mins.

During games however, you should get lower loads, even with a game maxed out visual settings-wise @ 1080p or whatever it can typical handle.

Try running a benchmark run-through using say:
- CineBench
- FutureMark ~ 3DMark & PCMark
- Unigine ~ Heaven & Valley

When you use the ones from say Futuremark or Unigine, pick from one of the Presets prior to running a test, this will lock in certain settings based on preset. This helps so that your end result scores can better compare to that of others; rather than running a range/mix of settings.

[Stunt-Monkey]

I pretty much have used all of those and getting good stable clocks and temps. i wanted to try both together to see what my max would be be. Obviously running seperate tests would give me a decent reading on temps but both tests would show max levels, thats when i came across the issue.

Running The division, it uses pretty much 100% gpu and around 60% cpu, temps from sensors like PCH, PCIE, VCORE would reach there 60s/70s after quite a few hours gaming and case fans at 100%, i can honeslty say that upgrading from a 970 to a 980Ti there quite a big difference in temps, cought me off guard and had to re do all my fan settings.

At least I can heat my room in winter for free now haha

[Bad 💀 Motha]

Originally posted by Stunt-Monkey:

I pretty much have used all of those and getting good stable clocks and temps. i wanted to try both together to see what my max would be be. Obviously running seperate tests would give me a decent reading on temps but both tests would show max levels, thats when i came across the issue.

Running The division, it uses pretty much 100% gpu and around 60% cpu, temps from sensors like PCH, PCIE, VCORE would reach there 60s/70s after quite a few hours gaming and case fans at 100%, i can honeslty say that upgrading from a 970 to a 980Ti there quite a big difference in temps, cought me off guard and had to re do all my fan settings.

At least I can heat my room in winter for free now haha

Nothing wrong with anything there; all normal.

If you turn VSync off and use the High or Ultra preset in Division at 1080p; what are you benchmark results looking like? The game has its own bench in options.

[Stunt-Monkey]

On my personal Ultra settings, some are higher some are lower, (Ultra doesnt actually give you all highest settings?

MY SETTINGS
Avg 67.6
Typ 68.5
Avg CPU 50
Avg GPU 90

ULTRA
Avg 67.2
Typ 77.2
Avg CPU 56
Avg GPU 85

Should note that ive avctaully noticed since playing inbetween these posts that both cpu and gpu will run 100% together hmmm

[Stunt-Monkey]

sorry just realised vsync was on for ultra settings

[Bad 💀 Motha]

Yea everything according to all that info; completely normal.

Have you tried running this same benchmark at say 1440p or 2160p?

Just curious, would like to compare #s

Also Division is an odd game, but in a certain sense that I like. What I mean by this is, you can leave VSync=On in-game; now make sure you have an FPS counter enabled, like from Steam settings for example. When Division loads up, see if the counter stays @ 60; if not, press ALT+ENTER to make the game actually full-screen; I like that ALT+ENTER works cause I can just use that to toggle VSync on/off at will with this game.

[Stunt-Monkey]

ive never actually used vsync in any game ive played, i run a samsung 120hz monitor and i can never see the difference, well i can when im looking for it but when playing games i never notice it. Havent tried upscalling but will give that a shot and see how it performs

[Stunt-Monkey]

Ultra, no vsync, 2160p gets me

Avg 77.4
Typ 78.3
Avg CPU 49
Avg GPU 86

[Bad 💀 Motha]

Originally posted by Stunt-Monkey:

Ultra, no vsync, 2160p gets me

Avg 77.4
Typ 78.3
Avg CPU 49
Avg GPU 86

Do u have the actual FPS Counter on-screen? You will see the font get super small @ 2160p

See the game can run in windowed @ desktop res if u do not press ALT+ENTER after the game reaches main menu; those #s all look like 1080p bench #s

As the CPU load should be way higher for 1440p or 2160p
Not to mention; no single GPU will net that FPS @ 2160p

[Stunt-Monkey]

hmmm good point, it wasnt going small it was just going blurry, interesting, ill look into that

[N0REGARD4LIFE]

Originally posted by Bad-Motha:

Also, regardless of CPU loads, a higher-end GPU like 980 Ti should really not be hitting loads above 50-60% @ around 1080p. Unless perhaps the game or benchmark is not well optimized and just ends up making the GPU hit a high % regardless of settings and res, etc.

Such a GPU as this would have plenty of breathing room to run say 1440p or even 2160p; so for 1080, 50% loads on GPU alone are fine and normal.

Now for CPU, stress tests and things of this nature will use 100% CPU, that is the whole point of them. To fully stress and bench your CPU. Now if you seeing that kind of CPU load in games during normal gameplay (as CPUs can hit 100% load during times of loading screens/maps) then I would say the CPU is set to something that is not stable. Or the CPU is throttling due to heat, or the VRMs are unstable/hot, forcing a CPU-throttle during peak loads.

My i7 4770/980ti Classified combo sees me hitting 93-99% gpubusage in the Division at 1080p with everything but shadows maxed. The Division seems pretty well optimized to me.