CPU Bottleneck again? WTF?!

[Opcode]

You know better dude..

Troll bait 

[Faa]

Troll bait 

Right.

 

It will depend if the GPU is being utilized to its fullest potential or not. The i3-4130 won't bottleneck a GTX 650 in any game that I know of. If you run the games you play with MSI Afterburner in the background. You can monitor GPU utilization in the charts on the MSI Afterburner main window. If its at a steady 99% while playing in the game, it means the GPU is being used at its fullest potential. Which also means if you overclock the GPU you will in fact see extra performance out of it in that particular game. Tho if your GPU isn't hitting the 99% area and is hovering around for example 75% utilization, that means the CPU is "bottlenecking" the GPU. So regardless to how much you overclock the GPU it will not improve frame rate.

http://linustechtips.com/main/topic/143993-helptips-in-overclocking-core-i3-4130-gtx650/#entry1923428

Lets compare it with your earlier post;

 

 

Check your CPU usage while gaming in these same games as well. If your CPU isn't punching 100% then neither the CPU or GPU are a bottleneck.

Just like you said even if the CPU is "punching" 100% it doesn't mean it's bottlenecking.

 

[i_build_nanosuits]

 

 

Just like you said even if the CPU is "punching" 100% it doesn't mean it's bottlenecking.

this is a coincidence though...you happened to catch a situation in which the cpu was JUST good enough for the game to run optimaly... battlefiled being extremely well optimised it happened to run even though the cpu was pinned...but most of the time when the cpu hits 100% like that either the gpu load will go down and the framerate will take a hit OR the game will stutter a little and take small pauses (usualy that's what happen and its very annoying and makes the game instantly unplayable...

 

 

BUT the other way around is not true...a cpu that runs only at 60% usage won't mean its not bottlenecking the gpu of course...

[Opcode]

Right.

 

 

http://linustechtips.com/main/topic/143993-helptips-in-overclocking-core-i3-4130-gtx650/#entry1923428

Lets compare it with your earlier post;

 

 

 

Just like you said even if the CPU is "punching" 100% it doesn't mean it's bottlenecking.

 

If your CPU peaks out (100% usage) under a full load you essentially run out of resources to feed the GPU. So even tho I personally don't care for the term "bottleneck" you do essentially run into one. Your screenshot above is a prime example as to this taking place, with the CPU topped out I am willing to bet you're seeing a GPU utilization ripple. Tho in this case where neither the CPU or GPU are hitting a full load ultimately results in neither being the culprit (like I stated in my previous post). It might be more along the lines of something like the PSU not feeding enough juice or something else.

[Faa]

this is a coincidence though...you happened to catch a situation in which the cpu was JUST good enough for the game to run optimaly... battlefiled being extremely well optimised it happened to run even though the cpu was pinned...but most of the time when the cpu hits 100% like that either the gpu load will go down and the framerate will take a hit OR the game will stutter a little and take small pauses (usualy that's what happen and its very annoying and makes the game instantly unplayable...

 

 

BUT the other way around is not true...a cpu that runs only at 60% usage won't mean its not bottlenecking the gpu of course...

Lol? The CPU was all the time at 100% load while the GPU never dipped below 80% load and averaged at 90%. SLI wasn't fully disabled, physics were offloaded to the 2nd card so add +x% to the 1st card. youtube.com/watch?v=o6IwHGTnG9I

 

 

If your CPU peaks out (100% usage) under a full load you essentially run out of resources to feed the GPU. So even tho I personally don't care for the term "bottleneck" you do essentially run into one. Your screenshot above is a prime example as to this taking place, with the CPU topped out I am willing to bet you're seeing a GPU utilization ripple. Tho in this case where neither the CPU or GPU are hitting a full load ultimately results in neither being the culprit (like I stated in my previous post). It might be more along the lines of something like the PSU not feeding enough juice or something else.

In a few posts above you said; the CPU will always bottleneck if its at 100% load. Which is bullshit. 1) The CPU load isn't any an indicator for CPU bottlenecks, only the GPU load is which you agreed on a long time ago. 2) Even if the cpu usage is 100% that doesn't mean that it will form a bottleneck. 3) CPU doesn't have to be at full usage to form a bottleneck.

Video above proves enough, ditch the GPU I used to a 760 which would have easily ran at 99% load all the time since its a significantly weaker card. Essentially no CPU bottleneck regardless of the CPU load. You'll get the best performance if your GPU is completely max'ed out which was happening so the CPU load is meaningless. 

Since the OP here has a xeon, goodluck finding a game that would take that xeon to 100% usage. Lets all just buy a 15 core xeon which will never hit 100% load anyways so it will never ever be a bottleneck according to you.

[Opcode]

 

Lol? The CPU was all the time at 100% load while the GPU never dipped below 80% load and averaged at 90%. SLI wasn't fully disabled, physics were offloaded to the 2nd card so add +x% to the 1st card. youtube.com/watch?v=o6IwHGTnG9I

 

 

In a few posts above you said; the CPU will always bottleneck if its at 100% load. Which is bullshit. 1) The CPU load isn't any an indicator for CPU bottlenecks, only the GPU load is which you agreed on a long time ago. 2) Even if the cpu usage is 100% that doesn't mean that it will form a bottleneck. 3) CPU doesn't have to be at full usage to form a bottleneck.

Video above proves enough, ditch the GPU I used to a 760 which would have easily ran at 99% load all the time since its a significantly weaker card. Essentially no CPU bottleneck regardless of the CPU load. You'll get the best performance if your GPU is completely max'ed out which was happening so the CPU load is meaningless. 

Since the OP here has a xeon, goodluck finding a game that would take that xeon to 100% usage. Lets all just buy a 15 core xeon which will never hit 100% load anyways so it will never ever be a bottleneck according to you.

 

The CPU will always become a bottleneck when at 100% load, unless you're fortunate enough to hit the sweet spot of both the CPU and GPU being evenly paired. It doesn't matter what your GPU is doing if there is a hold up at the API layer (CPU drives API, API drives GPU). Run the game again with 780 Ti and a Celeron 430 and let me know how that works out for you (we both know the end result). Also as I have said in the past, there's more factors involved with "bottlenecks" than just hardware utilization. Your own video backs up everything that I have been saying. Look at that GPU utilization ripple. In Battlefield 3 my A10-6800k pushes my HD 5870 to 99% without any ripple at all (solid 99% utilization line) while the CPU peaks out around 70-80%.

 

You're basically agreeing to everything I have said, CPU usage means next to nothing when the GPU is pushed to 99% without ripple.

 

You'll still hit a performance wall even with a 30 core Xeon, most games barely utilize four cores as it is. And overall game performance is solely dependent on how fast core #0 is. This is why the G3258 can match the 4770k in a few games when overclocked.

[Demonking]

 

Lol? The CPU was all the time at 100% load while the GPU never dipped below 80% load and averaged at 90%. SLI wasn't fully disabled, physics were offloaded to the 2nd card so add +x% to the 1st card. youtube.com/watch?v=o6IwHGTnG9I

 

 

In a few posts above you said; the CPU will always bottleneck if its at 100% load. Which is bullshit. 1) The CPU load isn't any an indicator for CPU bottlenecks, only the GPU load is which you agreed on a long time ago. 2) Even if the cpu usage is 100% that doesn't mean that it will form a bottleneck. 3) CPU doesn't have to be at full usage to form a bottleneck.

Video above proves enough, ditch the GPU I used to a 760 which would have easily ran at 99% load all the time since its a significantly weaker card. Essentially no CPU bottleneck regardless of the CPU load. You'll get the best performance if your GPU is completely max'ed out which was happening so the CPU load is meaningless. 

Since the OP here has a xeon, goodluck finding a game that would take that xeon to 100% usage. Lets all just buy a 15 core xeon which will never hit 100% load anyways so it will never ever be a bottleneck according to you.

 

Ok now i'm confused my I5 4670K hits close to 100% in BF4 with my GTX 760 hitting over 90% is that a issue or not ?

[Faa]

The CPU will always become a bottleneck when at 100% load

Always? This was proven a post above wrong.

 

 

You're basically agreeing to everything I have said, CPU usage means next to nothing when the GPU is pushed to 99% without ripple.

You said; "If your CPU isn't punching 100% then neither the CPU or GPU are a bottleneck." Meaning that your CPU can only be a bottleneck at 100% load. Which I didn't agree with. Few months ago you clearly stated that the CPU load is not an indicator of CPU bottlenecks.

 

You'll still hit a performance wall even with a 30 core Xeon, most games barely utilize four cores as it is. And overall game performance is solely dependent on how fast core #0 is. This is why the G3258 can match the 4770k in a few games when overclocked.

Not sure why you are mentioning this. It was a counterargument for your claim that CPU's can only bottleneck at 100% load.

 

 

Look at that GPU utilization ripple. In Battlefield 3 my A10-6800k pushes my HD 5870 to 99% without any ripple at all while the CPU peaks out around 70-80%.

I'm not saying that your CPU will never bottleneck when it's at 100% load. I'm saying that it's not an indicator of CPU bottlenecks and that even if your cpu usage is at 100% load it doesn't mean your CPU bottlenecks. 

5 secs out of the 300secs we see the GPU dropping below 99%. Give youtube some time to encode it

https://www.youtube.com/watch?v=AzE9IQrWsI4

 

 

Ok now i'm confused my I5 4670K hits close to 100% in BF4 with my GTX 760 hitting over 90% is that a issue or not ?

Without Vsync, your gpu is supposed to hit near 99% load that just means your GPU is performing at its full potential and it also means that the CPU didn't bottleneck at all. The CPU load tells literally nothing, only the GPU load however does.

[i_build_nanosuits]

Ok now i'm confused my I5 4670K hits close to 100% in BF4 with my GTX 760 hitting over 90% is that a issue or not ?

as long as your gpu is being fed all it can handle it's fine...but 90% cpu usage should indeed start to worry you...once the cpu hit 100% it means it can no longer execute all the instructions required by the game and will become the bottleneck soon or later...regardless of what Faa has to say about it thats just how it works...you never want a cpu to be pinned at 100% while gaming, thats it!

[Sintezza]

intel fans please dont feel butthurted right now. 

Bottlenecking term, just gets miss-used too much.

[Faa]

as long as your gpu is being fed all it can handle it's fine...but 90% cpu usage should indeed start to worry you...once the cpu hit 100% it means it can no longer execute all the instructions required by the game and will become the bottleneck soon or later...regardless of what Faa has to say about it thats just how it works...you never want a cpu to be pinned at 100% while gaming, thats it!

I5's pin all the time above 90% load in BF4 while keeping the GPU at full load consistently.

[i_build_nanosuits]

I5's pin all the time above 90% load in BF4 while keeping the GPU at full load consistently.

sure! never said otherwise...but how much more will it handle? not much...battlefield 5?! you know what i mean i'm sure...

[Faa]

sure! never said otherwise...but how much more will it handle? not much...battlefield 5?! you know what i mean i'm sure...

And how much more would a i7 handle? Hyperthreading can cause a huge performance loss sitting below a i5 even. If you really think, Hyperthreading which gives a 10-15% improvement at best will ever be a saviour then the 8350 will be chosen for the worlds best performing and efficient cpu ever made.

What about Battlefield 5? Battlefield 4 was hardly better multithreaded than BC2. That game hardly took my CPU above 40% load at 1.2GHz, BF4 has been nothing more besides a junk multithreaded game. 99% of the time we are getting bottlenecks at 30% load which equals in this case for me 4 cores so no need for a i7. youtube.com/watch?v=2VstWexLBDA

Whenever games will fully saturate all threads a 4770K/4790K has to offer, the 5820K will stomp over it by up to 50% clock for clock or a 5960x by 100% so let me ask you the following question. "Sure never said otherwise...but how much more will the 4770K/4790K handle? not much..Battlefield 6?! you know what I mean i'm sure..."

[Opcode]

Always? This was proven a post above wrong.

Always, if you don't understand how everything works at a software layer you'll never understand how it works in conjunction with hardware. Hardware is only as fast as the software that's running on it and vise versa. Once the game process itself exhausts all of your CPU's resources, threads get pushed back in queue waiting to be executed. These are thousands of batches that are delayed that the GPU could be executing. This is why you see frame and utilization dips in your dual core tests.

 

You said; "If your CPU isn't punching 100% then neither the CPU or GPU are a bottleneck." Meaning that your CPU can only be a bottleneck at 100% load. Which I didn't agree with. Few months ago you clearly stated that the CPU load is not an indicator of CPU bottlenecks.

If your CPU isn't being pushed to its limits, there shouldn't be a bottleneck. Below 100% utilization means there are plenty of cycles to spare. If you're playing a game and your CPU utilization doesn't peak out, it's easily ruled out as not being the cause of a bottleneck. Tho if your CPU peaks out and your GPU doesn't during gaming, that's typically an indicator that the CPU is to slow to drive the API to feed the GPU.

 

Not sure why you are mentioning this. It was a counterargument for your claim that CPU's can only bottleneck at 100% load.

CPU's do bottleneck at 100% load if the game itself is exhausting all of them resources. Maybe I will be a little more clear next time to avoid the confusion.

 

I'm not saying that your CPU will never bottleneck when it's at 100% load. I'm saying that it's not an indicator of CPU bottlenecks and that even if your cpu usage is at 100% load it doesn't mean your CPU bottlenecks. 

I never stated that a just because a CPU reads out 100% utilization that it will create a bottleneck. What was being implied is if the game itself is eating up every resource the CPU has to offer, it will bottleneck.

 

5 secs out of the 300secs we see the GPU dropping below 99%. Give youtube some time to encode it

https://www.youtube.com/watch?v=AzE9IQrWsI4

Unigine Valley is primarily a GPU bound benchmark (even you should know that). What you see in the video doesn't explain a CPU bottleneck (dealing at a software level now). I am interested to see what happens if you ran Tomb Raider and a CPU benchmark (designed to benchmark each core independently) at the same time. I imagine you would see skewed results across the board due to each core that the game utilizes. I guess you could try running six instances of HyperPi and force them to all use an independent core.

[Faa]

Unigine Valley is primarily a GPU bound benchmark (even you should know that). What you see in the video doesn't explain a CPU bottleneck (dealing at a software level now). I am interested to see what happens if you ran Tomb Raider and a CPU benchmark (designed to benchmark each core independently) at the same time. I imagine you would see skewed results across the board due to each core that the game utilizes. I guess you could try running six instances of HyperPi and force them to all use an independent core.

Was done with prime95 in the background if you havent noticed it.

[Opcode]

Was done with prime95 in the background if you havent noticed it.

Prime95 doesn't reign thread priority. I am curious what would happen if you forced Prime95 to the highest level of thread priority and then ran the game again.

[Majestic]

If you're playing a game and your CPU is edging near 100% on any of the cores (please stop talking "100%" on a multicore CPU, because load is not 100% independent), eventhough it's technically not being a "bottleneck", it will create less drawcalls than it would were it not so close to being saturated. If you have more overhead, the CPU has oppertunity to create an extra drawcall on a GPU idletime. I know this because there is no point at which CPU stops mattering. Even in GPU bottlenecked situations you can see faster IPC CPU's create more fps. Sure, you'll reach a point of diminishing returns where 50% extra overhead only renders 2% extra fps, but it's not this black & white situation that is being depicted here. My i5-760 at 4.2ghz was able to mostly saturate my 670 to 99%. But i still got an increase in fps when i switched to the 4670K.

 

The benefits basicly operate under first order system graph.

 

And surely if only 1 core is saturated, it becomes a limiting factor all. Especially since that is usually the mainthread, which is dependent, and will cause idle/wait in the application/api. This is why Intel does so much better at min. fps, due to the way it handles the mainthread (by IPC). It has thus better scaling across the rest of the cores during stressful moments.