• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

When do you notice the speed difference?

bamavooHF

bamavooHF

n00b
Joined
Mar 30, 2017
Messages
30
if I am running an i5 357k @ 4.5GHz what CPU would I need to see a difference? I know I would have to change platforms.
 
When your current platform feels slow in current games and apps, that is the time to upgrade.
 
CPU speed just doesn't seem to be progressing very fast. Is a 4 thx on an x99 platform really faster than the 1155?
 
actually everything depend entirely on what are you doing with your machine, if you are gaming, you would see high benefits even by jumping to a 4c/8t 3770K at the same speed, as most newer games not only benefit from more threads for performance but also on the "feelings" with gaming, the smoothness, the stability on frames, frametimes are much better with newer games and 4/c8t chips, I did a jump last year from a 3770K@4.5ghz to a 6700K first at same 4.5ghz then 4.8ghz and the performance jump was highly noticeable specially on some heavy single threaded games as Fallout 4 as example where minimums skyrocketed at least 2x.. and as a 120hz user minimums are a high concern always, so it was a totally worth jump to me, everything just feels much better, heavy CPU games are noticeable smoother even on GPU limited scenarios..

so, sometimes isn't just about numbers but feelings.. you don't know normally how needed an upgraded is until you upgrade and notice everything is just better, and that's a general case from all the guys here who have upgraded from Sandy/Ivy platforms to Skylake/Kaby Lake staying on the same 4c/8t platform.. but now, worth at this point jump to another 4c/8t chip? absolutely nope, actually 6c/12t is the sweet spot to aim for gaming if one is looking for a platform longevity and with the prices AMD are offering with ryzen it's just plain stupid to look at anything else than a 6c/12t chip, I don't recommend to anyone at this point in the market to go with a 4c/8t CPU even for exclusive gaming, im actually waiting on what exactly is intel going to offer (besides the late rumors) to counter AMD on price and performance as the 1700X/1800X are every day more and more tempting..
 
bamavooHF said:
if I am running an i5 357k @ 4.5GHz what CPU would I need to see a difference? I know I would have to change platforms.
Click to expand...
depends what you do.

Nearly everything is single thread so getting a 5GHz KBY is a big improvement in snappiness across your whole system but it only matters if you are tuned into snappiness

my 4.4GHz 1650v3 is way slower to me than my 4.8GHz SKL

I use single thread more because i use day to day programs. chrome, OS, paint, explorer, PDF, OCR, and so on. They are all single thread limited.

I am also aware of snappiness and can tell the difference because i have my win7 remove all forced animations and transitions.
 
bamavooHF said:
if I am running an i5 357k @ 4.5GHz what CPU would I need to see a difference? I know I would have to change platforms.
Click to expand...

it depends on how much you CPU is the limiting factor in what you do. if its not you will not see any difference.
Do a CPU bottleneck measurements before considering to upgrade a cpu
 
So Intel is still the way to go? I actually have an AMD 8300 system and this 3570k. They are pretty equal in speed on all things we do with them. From gaming to surfing the web. They both have SSD main drives. The AMD has a 7950 and the k has an R9 390, which does provide more frames in games.
 
SvenBent said:
it depends on how much you CPU is the limiting factor in what you do. if its not you will not see any difference.
Do a CPU bottleneck measurements before considering to upgrade a cpu
Click to expand...
have a second screen and watch task manager you can easily tell if your single thread bottlenecked. control p in chrome is single thread limited...fun fact. So are all web pages.
 
SomeGuy133 said:
have a second screen and watch task manager you can easily tell if your single thread bottlenecked. control p in chrome is single thread limited...fun fact. So are all web pages.
Click to expand...

This is not optimal way to do it because you will not be able to see core speed bottleneck. just overral CPU bottleneck

for corespeed bottlenekc use process explorer . select you game/program in the process list and select to see the threads.
there you cna se every threads CPU usage.
if a threade hits close to 100%/number of logical cores. you are hitting a corespeed limitation
 
SvenBent said:
This is not optimal way to do it because you will not be able to see core speed bottleneck. just overral CPU bottleneck

for corespeed bottlenekc use process explorer . select you game/program in the process list and select to see the threads.
there you cna se every threads CPU usage.
if a threade hits close to 100%/number of logical cores. you are hitting a corespeed limitation
Click to expand...
you can clearly see it in taskmanager and ThrottleStop. If a process sits at 13% usage and takes 1 second to complete it is single bottlenecked. If its at a fixed 20% consistent it is 100% single thread bottlenecked.

TS is a better method because you can see the C0% (IIRC the correct one to look at off my head)

C0% is the one that says a core is at 100% max so if a core is 100% laoded that means your single thread bottlenecked.
 
SomeGuy133 said:
you can clearly see it in taskmanager and ThrottleStop. If a process sits at 13% usage and takes 1 second to complete it is single bottlenecked. If its at a fixed 20% consistent it is 100% single thread bottlenecked.

TS is a better method because you can see the C0% (IIRC the correct one to look at off my head)

C0% is the one that says a core is at 100% max so if a core is 100% laoded that means your single thread bottlenecked.
Click to expand...

If a process sits a 13% how do you know it is not 2 threads running at around 6% each, but limited by GPU or I/O?
You can also see core utilization in taskmanager but that does not inform you of Corespeed limitations either due to how quanta/CPU slices are being handed out to threads.

The same issue you have with C0% is that due to threads being spread across core over time you don't know if you are actually hitting a max.
Looking at core utilization does NOT inform you of hitting a corespeed bottleneck, unless you for sure have locked the thread to a giving core
On the contrary measurement C0X you are moving even farther away from what you want to measure because you are no longer measuring the software utilization of the CPU/cores, but instead how the CPU powermanagement is reaction to CPU utilization with its inherited delays.
So it is actually even more inaccurate than CPU/core utilization that measures it in cycles used by a thread/process


Let me illustrate why your method is horrible unoptimal for the goal of seeing Corespeed limitation

Core_STuff.png

As you can see over the time measuring interval of 1sec (1sek) you don't have any core that is being used more than 25% of a core or more than 3.125% of the CPU utilization
The entire CPU usage over this time interval is 12.5% but we of this time don't know if that is one thread or more without looking at thread utilization.
But looking at thread utilization we can see one thread is using 12.5% of the total CPU which is exactly 100% of a core.
 
Last edited:
SomeGuy133 said:
you can clearly see it in taskmanager and ThrottleStop. If a process sits at 13% usage and takes 1 second to complete it is single bottlenecked. If its at a fixed 20% consistent it is 100% single thread bottlenecked.

TS is a better method because you can see the C0% (IIRC the correct one to look at off my head)

C0% is the one that says a core is at 100% max so if a core is 100% laoded that means your single thread bottlenecked.
Click to expand...

if you disagreee then can you please inform me how to analyse this and tell me if im currently running with a CPU bottleneck or not?
and how many CPU heavy threads are running (if any)
Untitled.png
 
SvenBent said:
if you disagreee then can you please inform me how to analyse this and tell me if im currently running with a CPU bottleneck or not?
and how many CPU heavy threads are running (if any)
Untitled.png
Click to expand...

You are not CPU limited there.. the max utilization of a single CPU thread is 58.8% on Thread #4 who share resources with Thread #5 which sit at 0.4% load (if they are named from #0 to #7 as it should normally) so not even resource stolen by HT on your CPU.. second highest load is from thread #6 which share resources with thread #7, so both are far of being single thread limited as each Logic Processor is working separated from the other adjacent thread, so only two "heavy"threads are being used with no penalty from HT so no CPU bottleneck you can be I/O Bottleneck by HDD/SSD and/or RAM but is not related to CPU bottleneck.. anyway, what SomeGuy 133 refers to is the use of taskmanager in the CPU tab with "per core utilization" selected to see every thread utilization. is easier to check with bar and graph each thread utilization..

75720-lightroom_cpu_comparison.jpg


in a game, when you see contantly threads hitting over 90% that's a CPU bottleneck there, and I say threads because due the way Windows by itself assing resources, so their usage can bounce from thread to thread. as you can see in the left picture above, several times 100% utilization was reached by several threads, however still the general CPU utilization is just 46%. that's what SomeGuy is referring to. you can see 12.5% CPU global utilization if a typical 8 thread CPU is being hitted at 100% on a single thread. of course the easiest way it's always to check task manager on per core thread utilization.
 
Araxie said:
You are not CPU limited there.. the max utilization of a single CPU thread is 58.8% .
Click to expand...

and that where you are wrong because you don't get that a thread can be over multiple cores over the time measured interval, like i showed above,
In the above example i was running 7-zip benchmark with 1 thread so i was very much CPU limited in corespeed and nothing else
But your way can't show it because it not optimal for the purpose of showing Thread utilization/corespeed limitation.

Also at no time can you see thread utilization on the info so saying I have a thread running at 58.8% is entirely wrong. That is a (logical) core. NOT a thread.
Don't confuse threads with logical cores. Threads are part of software.

So to sum it up: The 58.8% usage on on logical core + the 44% on the other logical cores. Was from the same single software thread that ran at 99-100% and in fact was limited by corespeed.
Your method is not correct and you are confusing technical terms



To addressee the above pictures:t
In the first picture you can very much have a thread running at 100%
it can be 50% on logical core Y and 40% on logical X and the finnalyy 10% on logical core z ( or any other combination).
Becaues you are measuring over time and the thread can move around the different logical cores over time.
but you don't know because you cant' see it form that info.


you can see 12.5% CPU global utilization if a typical 8 thread CPU is being hitted at 100% on a single thread.
Assuming this load is only from one thread but you method has no way to show that so still you don't know that. and you could be running 2 threads that are only using 6.75% each.
Again don't confuse threads with logical cores.


let me show you what threads are in the next post
 
Last edited:
let me show you the Threads CPU utilization from the 7-zip 1 threaded benchmark
image.png

as you can see here thraed TID 5696 are in facts using around 12.5% of the total CPU or what is 100% of a core. But your way did show it, by your own words, that there was not CPU bottlnec because you thought i had 2 threads running at 58% and 44%.
When it in facts was this one single threads running at near 12.5% of a CPU (100% of a core) but over the time measure intervall was split over at least 3 logical cores. THIS is why its important not to confuse threads with logical cores.



Now this is the poster child exameel of how you way of measuring, also break down and can't give correct information
image.png

As you see here there are 2 CPU heavy threads. One is 10% and one i 7%
if i just looked for 12.5% as you say. Then I would think i was limited by Coresped. When it facts this usage is split over 2 threads and is NOT hitting the corespeed limitation

So again. don't confuse threads with logical cores. and your method is inaccurate to measure CPU bottlenecks in regards to corespeed.
 
allow me to demonstrate just once more.
I started a game and took some test

let me see by your method
image.png



Hmmm not really that information full. Am i bottleneck by core speed or not. no cores are hitting 100%
this could just eaisly be 6 or more low usage threads


leet me look at my method
image.png

Nope definatlt not 6-7 threads running with low usgage. This is in facts a single thread running at 12.4% of my full aka near 100% of a core.
so i am definatly corespeed limited.

Can you see the difference now?
And how its better to look at the actually threads usage of a CPU?
It gives you a way more exact pictures of what is limiting your program CPU wise
 
I use threads for process so forgive my laziness not being exact with terms. Processes can bounce between threads and cores but the end result is still 100% used core for X% of time.

Also what is that window you have open? I am on 10 right now and not sure where that is. It looks like a win7 thing.

C0% means the CPU was at 100% 58% of the time which means that CPU is single thread limited. Threads can bounce from core to core to core so if you see:

C0% 58%
C0% 32%
C0% 10%
C0% 0%

That means 3 cores are maxed at that % per refresh

The first screen shot shows 1 core maxed at 100% and another core maxed at 100% at 20% of the time per refresh.

I remember this is what uncleweb told me on techpowerup when i was trying to figure out a way to tell if you are single thread limited.

Found unclewebbs post. Tell me if i misunderstood him.
https://www.techpowerup.com/forums/...leases-throttlestop-8-30.226664/#post-3569400

I stated task manager. I never said the graph. The process is the better way to tell but you can still tell from the graph if you know what your looking for. It doesn't have to be 100% in the graph on a single core. tasks bounce between cores. If you open a program and see a massive spike anywhere and it isn't all 4 cores maxed and the task at hand takes more than a half second to complete and load seems consistent it is 100% single thread limited.

If you stay at the graph enough you learn what to look for. I do this daily and I can pick out from a live feed of the graph and tell you if anything is singlethread limited assume it takes longer than .5 seconds to complete and the computer has mostly idle load.

But TS and process menu are easier ways to tell. Alone with other methods. Liek playing a game. If GPU utilization drops and CPU is static...its CPU limited.
 
Last edited:
SomeGuy133 said:
I use threads for process so forgive my laziness not being exact with terms. Processes can bounce between threads and cores but the end result is still 100% used core for X% of time.
Click to expand...

Exactly but you way dont show it because it døsn measure the bounching back and forth. aka a thrread runnig 100% of a core over time might be 25% on 4 different cores."
So you cant see if its really on thread on 100% bocunging over 4 cores. or if it is 4 threads that are 25% each bounching over the same 4 cores.
and there is a differencen. if its 1 its a corespeed boltle neck. but if its 4 threads its not cause each threads still have 75% unnused time.


Also what is that window you have open? I am on 10 right now and not sure where that is. It looks like a win7 thing.
Click to expand...
It is from Process explorer which is a tool you need to download from microsoft
https://technet.microsoft.com/en-us/sysinternals/processexplorer.aspx


C0% means the CPU was at 100% 58% of the time which means that CPU is single thread limited. Threads can bounce from core to core to core so if you see:
Click to expand...
Correct


C0% 58%
C0% 32%
C0% 10%
C0% 0%

That means 3 cores are maxed at that % per refresh
Click to expand...
I would have to disagree. in this refresh cycle we know that ther is one c ores at 58% and 32% and 10% and 0 % its a total of
so all we know is that there is a total usage of 100% of a single core. But its spread across 3 cores.

we dont know if that 100% of a core laod is from 1 threads running 100% or if it 100 threads runnigs 1% each. ergo we dont know if we are limited or not.
if it 1 single threads running 100% we are limited because this threads are using all potential avaible resrouce.
1 thread can only use 100% of a core so there is no breaks. its runnig constantl trying to keep up doings it calculations.

If instead it was 4 threads of 25% these threads can utilzia op to 400% of a cores (100% of a core per threads) so since its only using 100% and not 400% each threads are only runnigs 25% of the time in average aka they are idle and wiatings on something else 75% of the time in avaerge.
That could be I/O or GPU we are waiting on. so these numbers does not tell us if we are limited by the core speed or not. because we don't know if the actually threads are wating on other parts of the system or its solely limited by corespeed.
We need to know how the threads are runnings for this


The first screen shot shows 1 core maxed at 100% and another core maxed at 100% at 20% of the time per refresh.
I'm no following here where you are seeing these 100% 100% and 20%


I remember this is what uncleweb told me on techpowerup when i was trying to figure out a way to tell if you are single thread limited.

Found unclewebbs post. Tell me if i misunderstood him.
https://www.techpowerup.com/forums/...leases-throttlestop-8-30.226664/#post-3569400
Click to expand...
What I am getting from here is that unclewebb is talking about hyssical cores vs logical cores
so lets say you are using 50-60% of a 4P/8L core CPU, the he says its would run better on a 6cores. well its a slighty wrong but not all bad. if we go in a theory that Hyperthreadin does not help at ALL. the yes he would be 100% correct.
Here is what he is saying. on a Hyper threading CPU you actually have 200% CPU usaage because you have 2 logical cores on one physsical core.
So 100% load on one logical core can be 100% load of a physsical core loading both logical cores to the same physsical cores would show a 200% core utilization even though we only have 1 physsical core do the work

This is also what i call a SMT conflicts and it looks somehow like this ( the illustration is basedon a ryzen cpu but its doesnt matter

SMT4_Tworstcase.png


If we look at CPU utilization what we see is that we are using 50% of 8 lgoical cores aka 4 logical cores. but behind those 4 logical cores are only 2 physsical cores.
So lets say we hit 50-60% usage of this kidn of CPU we are actually having around 5 threads runnings (4 threads cna bring us to a max of 50%.
unclewb the says so if you have a game with 5 threads of heavy usage (50-60% CPU usage) you aer better of on 6 core CPU because instead of having to on shared physsical cores. you know have a physsical core for each threads

This is however asumming that
1: HT doesnt do anything
2: That all threads are running near 100%.

Basically only in therotical a minimum threaded situation.

still this does not really prove core speed limitation as it about Core scaleabilty. We basically just know the software has a potential for scaling over more physical core. but we dont know if its limited by the current corespeed.
and its a very indirect way to see it rather than just you know... looking at the actually threads and see how many threads are cpu heavy


In short: its about potential capability. and not a accurate measure of corespeed bottleneck. it is based on a lot of assumptions








I stated task manager. I never said the graph. The process is the better way to tell but you can still tell from the graph if you know what your looking for. It doesn't have to be 100% in the graph on a single core. tasks bounce between cores. If you open a program and see a massive spike anywhere and it isn't all 4 cores maxed and the task at hand takes more than a half second to complete and load seems consistent it is 100% single thread limited.

If you stay at the graph enough you learn what to look for. I do this daily and I can pick out from a live feed of the graph and tell you if anything is singlethread limited assume it takes longer than .5 seconds to complete and the computer has mostly idle load.
Click to expand...
This is not correct because again you don't know what the CPU usage you see is for

But please tell me what you are looking for then. Tou haven said anything of what to look for.
How do you see with taskamanager if a game is CPU corespeed or not with taskmanager?
You don't tell you are just claiming you can do it. but yet you couldt when i asked into it.


But TS and process menu are easier ways to tell. Alone with other methods. Liek playing a game. If GPU utilization drops and CPU is static...its CPU limited.
Click to expand...
It still doesnt tell you the right info
and remember you did in fact fail to indentify a corespeed bottlenck when i gave you a TS screenshoot.

You don't know if a 100% core/time usage is comming form a single thread and is corespeed limited or 4 threads and it is not corespeed limited.
Please tell me how you know the difference.
Click to expand...
Click to expand...
 
SvenBent said:
Exactly but you way dont show it because it døsn measure the bounching back and forth. aka a thrread runnig 100% of a core over time might be 25% on 4 different cores."
So you cant see if its really on thread on 100% bocunging over 4 cores. or if it is 4 threads that are 25% each bounching over the same 4 cores.
and there is a differencen. if its 1 its a corespeed boltle neck. but if its 4 threads its not cause each threads still have 75% unnused time.



It is from Process explorer which is a tool you need to download from microsoft
https://technet.microsoft.com/en-us/sysinternals/processexplorer.aspx



Correct



I would have to disagree. in this refresh cycle we know that ther is one c ores at 58% and 32% and 10% and 0 % its a total of
so all we know is that there is a total usage of 100% of a single core. But its spread across 3 cores.

we dont know if that 100% of a core laod is from 1 threads running 100% or if it 100 threads runnigs 1% each. ergo we dont know if we are limited or not.
if it 1 single threads running 100% we are limited because this threads are using all potential avaible resrouce.
1 thread can only use 100% of a core so there is no breaks. its runnig constantl trying to keep up doings it calculations.

If instead it was 4 threads of 25% these threads can utilzia op to 400% of a cores (100% of a core per threads) so since its only using 100% and not 400% each threads are only runnigs 25% of the time in average aka they are idle and wiatings on something else 75% of the time in avaerge.
That could be I/O or GPU we are waiting on. so these numbers does not tell us if we are limited by the core speed or not. because we don't know if the actually threads are wating on other parts of the system or its solely limited by corespeed.
We need to know how the threads are runnings for this
Click to expand...
according to unclewebb if a core is in C0% it is being maxed out.

So C0% at 58% means that core is maxed 58% of the time which is a core bottleneck.

You will never see a core at 100% if it is 100 threads unless all other cores are also loaded. Windows will never overload a core if another core is still available....unless windows is that badly designed...i would highly doubt that.

If a process used 100% of a core 25% of the time and than waits for another process it is still CPU limited 25% of the time. A faster single core speed will speed up the process.
EDIT:
BTW looking at that program and its pretty cool. It might come in handy for a project of mine.
 
Last edited:
SomeGuy133 said:
according to unclewebb if a core is in C0% it is being maxed out.

So C0% at 58% means that core is maxed 58% of the time which is a core bottleneck.
Click to expand...

Again that is actually totally wrong because you don't know what the core is maxed out WITH
it correct you can see a specific core is maxed out but not if its a corespeed limitation because you don't know of the software acutally has the potential to use more than one core but it not because it just don't need more CPU ressources.
again because you don't look at the threads. over the time interval measure that could be to slow threaded being bottlenecked by GPU or I/O and only need 50% CPU time each because they wait 50% of the time.
So the conclusion are incorrect based on a correct observation.


You will never see a core at 100% if it is 100 threads unless all other cores are also loaded. Windows will never overload a core if another core is still available....unless windows is that badly designed...i would highly doubt that.
Click to expand...
correct which is the very reason why looking at core utilization does not reveals if you are hitting a core-speed limitation.
Because AGAIN one threads can be running on a giving core 100% of the time trying to get all potential available CPU power. but since its spending time a bit here and therer on each core you will never see the limitations by your method.
We can however very easily see a thread running at 100% core utilization, which again is why we want to look at thread utilization of a CPU.

If a process used 100% of a core 25% of the time and than waits for another process it is still CPU limited 25% of the time. A faster single core speed will speed up the process.
Click to expand...
mostly Incorrect
You could argue that you are limited in that short while but that's really not the important part here. The problem to find a CPU bottleneck is to see if it holding back the time to project a taske.
And by this argument you are saying that you are always CPU limited (Because there wil always be some CPU load and on that time you are CPU limited)
The problem is that you assume as strongly serielize workload. Where its more often more of a Gatling serialization aka your CPU is working on data set 2 while storage is fetchings dataset 3 ad the GPU is working on dataset 1

So lest say you do have a thread that runs 25% of a core over 1 sek. and fecthing data from a storage that can deliver the section of data in 12cycles Those might bee 3 cycles calculations then waitt 9 cycles for data. execute 3 cycles. wait 9 cycles for the next data data.
Speeding op the CoreSpeed in this case would still have to wait for the data to come from the storage and you would still have to wait 12 cycles to complete a task. More core speed would just lower the CPU utilization but not decrease task for completion because CPU power is not the bottlenecked for this task
So in short: mostly incorrect.



Again you did not explain how you would see a core speed limitations in task manager let me again illustrate the inherited flaw of taskmanger

image.png

image.png


Please inform me which is CPU bottlenecked and which is I/O bottlenecked by your measuring method. and how you came to which conclusion.
It should be obvious that looking at task manger process view or graph or whatever is useless to see core speed limitations, because you don't know the underlying load distribution and how it reflects in according to potential available CPU power.
 
oh and yes process explorer is pretty gold in regards to a lot of stuff.
I use it to adjust settings in my games for optimal performance. because i can easily see when im starting to hit a CPU bottleneck and can adjust details
 
SvenBent said:
Again that is actually totally wrong because you don't know what the core is maxed out WITH
it correct you can see a specific core is maxed out but not if its a corespeed limitation because you don't know of the software acutally has the potential to use more than one core but it not because it just don't need more CPU ressources.
again because you don't look at the threads. over the time interval measure that could be to slow threaded being bottlenecked by GPU or I/O and only need 50% CPU time each because they wait 50% of the time.
So the conclusion are incorrect based on a correct observation.



correct which is the very reason why looking at core utilization does not reveals if you are hitting a core-speed limitation.
Because AGAIN one threads can be running on a giving core 100% of the time trying to get all potential available CPU power. but since its spending time a bit here and therer on each core you will never see the limitations by your method.
We can however very easily see a thread running at 100% core utilization, which again is why we want to look at thread utilization of a CPU.


mostly Incorrect
You could argue that you are limited in that short while but that's really not the important part here. The problem to find a CPU bottleneck is to see if it holding back the time to project a taske.
And by this argument you are saying that you are always CPU limited (Because there wil always be some CPU load and on that time you are CPU limited)
The problem is that you assume as strongly serielize workload. Where its more often more of a Gatling serialization aka your CPU is working on data set 2 while storage is fetchings dataset 3 ad the GPU is working on dataset 1

So lest say you do have a thread that runs 25% of a core over 1 sek. and fecthing data from a storage that can deliver the section of data in 12cycles Those might bee 3 cycles calculations then waitt 9 cycles for data. execute 3 cycles. wait 9 cycles for the next data data.
Speeding op the CoreSpeed in this case would still have to wait for the data to come from the storage and you would still have to wait 12 cycles to complete a task. More core speed would just lower the CPU utilization but not decrease task for completion because CPU power is not the bottlenecked for this task
So in short: mostly incorrect.



Again you did not explain how you would see a core speed limitations in task manager let me again illustrate the inherited flaw of taskmanger

image.png

image.png


Please inform me which is CPU bottlenecked and which is I/O bottlenecked by your measuring method. and how you came to which conclusion.
It should be obvious that looking at task manger process view or graph or whatever is useless to see core speed limitations, because you don't know the underlying load distribution and how it reflects in according to potential available CPU power.
Click to expand...
Again...Just because something is GPU limited doesnt mean it isn't also CPU limited. If a core is 100% maxed for 300ms....it is still signal thread/single core limited.
If GPU and CPU loads bouce a task back and forth there is still a net gain with a better CPU...granted slight but there is still a singlethread limit.

I have never heard of a process or task that only uses 25% of a CPU consistently. It is always max on and off inregards to loads....atleast for everything I do.

Windows will never over load a core unless you can prove otherwise that is always the case. Maxed core=singlethread limited

Maybe short term isn't relevent to you but it is for me and everything I do.

I care about sub second response times in regards to snappiness and ultra fast response times. Windows and tasks taking 300ms+ is something I am interested in due to IBMs study from 1982.
http://jlelliotton.blogspot.com/p/the-economic-value-of-rapid-response.html

As you can see this is where you and I are looking at the data and perspective differently. For me I look at how a program and OS operate in the sub second field because I am interested in seeing and improving the responsiveness of a computer for productivity.

I look at per click loads and dont generally look at long term loads because there is massive improvements in productivity for day to day usage in getting response times low.

And again thanks for explaining a lot of this. I have found a new way to see this deeper and that will help me.

Again the way I can a core limitation is by seeing the graph and knowing what I am doing.

I control P in chrome and see a spike and know it takes 1 second to open and the spike correlates to the opening of print screen in chrome. I know there is no SSD/HDD activity and RAM is obviously not a limit here and nor is the internet. So I cna obviously see it is single thread limited. Also because there is a tangible different between 3.6 and 4.8GHz in terms of delay. It is just obvious.

I open explorer on a 950 PRO with a bunch and thumb nails that does not require much SSD speed (did it in RAMdisk too) and I see a consistent spike for that entire time. I know it is single thread limited for no reason. I think in Win10 they finally made that threaded but not sure. (idiotic on how much is not threaded in programs and OS).

But now I got a new way to analyze this :D thanks

Is there a graphing and logging feature so I cna single screen records and chart aspects of the OS?

Again I can tell when I am using the computer as I explained above watching when I know enough about the task. I don't know what your doing and I can't be sure.

For many programs and windows operations you can figure it out by knowing what everything on your PC is going by simply deducing it but I like that program and makes it easier :D
 
#27
When did i say you could not have a CPU and a GPU bottleneck under the same software? Its highly unlikely in games during the same instance. But over a timeframe yes it is possible
However it still does not change the fact that the method you proposed does do an optimal job at finding them.

You failed in your first attempt to use your method and have not addressed the others clearly showing the lack of ability to prrof your method is working, and it apeppas you don't want to learn another method, and are now tying to move the goalpsot to be right. So there is no real reason to further debate this.
 
Shintai said:
And downgrade for a lot of games. Sounds like a bad idea.
Click to expand...

Prove this comment Shintai?

In what qualitative measurement can you post that proves that he would be downgrading his gaming by leaving his I5 to a Ryzen 1700 oc'd to 4ghz?

I am extremely curious
 
SvenBent said:
#27
When did i say you could not have a CPU and a GPU bottleneck under the same software? Its highly unlikely in games during the same instance. But over a timeframe yes it is possible
However it still does not change the fact that the method you proposed does do an optimal job at finding them.

You failed in your first attempt to use your method and have not addressed the others clearly showing the lack of ability to prrof your method is working, and it apeppas you don't want to learn another method, and are now tying to move the goalpsot to be right. So there is no real reason to further debate this.
Click to expand...
jesus you never read my post so just stfu then.

tangoseal said:
Prove this comment Shintai?

In what qualitative measurement can you post that proves that he would be downgrading his gaming by leaving his I5 to a Ryzen 1700 oc'd to 4ghz?

I am extremely curious
Click to expand...
counting the higher clock but the IPC difference between ryzen and IB it is likely they bought have roughly the same single thread so a side grade for almost all programs and games.
 
You must log in or register to reply here.
Back
Top