Bottleneck test?

[WillOfTheLand]

We know there are multiple benchmark softwares out there and from them you can extrapolate if you have a CPU or GPU bottleneck, but is there any benchmark software out there for just straight up testing to see if you have a bottleneck, and if so just how bad it is?

[Fasauceome]

No

 

The word bottleneck seems to lose meaning by the day. Every pc has a weak point, a "bottleneck".

[Helpful Tech Witch]

You have a bottleneck. What are your parts? Do you have a 10 year old CPU and a brand new GPU?

[Guest]

https://pc-builds.com/calculator/

Here's one if you just want a number, but I can't vouch for how reliable it is

[Fasauceome]

2 minutes ago, itswillum said:

https://pc-builds.com/calculator/

Here's one if you just want a number, but I can't vouch for how reliable it is

Here's another good one

 

https://www.random.org/

[Lurick]

1 minute ago, Fasauceome said:

Here's another good one

 

https://www.random.org/

That's probably the best one out there and just as accurate

[WillOfTheLand]

3 minutes ago, TheTechWizardThatNeedsHelp said:

You have a bottleneck. What are your parts? Do you have a 10 year old CPU and a brand new GPU?

I have a 10 year old CPU and 10 year old GPU lol..... I was just curious if anything like it had been made. 

[Haro]

10 hours ago, itswillum said:

Here's one if you just want a number, but I can't vouch for how reliable it is

It isn't. 

 

10 hours ago, WillOfTheLand said:

if you have a CPU or GPU bottleneck

Every system has a bottleneck, any site that claims it's a calculator for that is fake. 

[Helpful Tech Witch]

Just now, WillOfTheLand said:

I have a 10 year old CPU and 10 year old GPU lol..... I was just curious if anything like it had been made. 

Then the bottleneck will be small.

[Fasauceome]

Just now, WillOfTheLand said:

I have a 10 year old CPU and 10 year old GPU lol..... I was just curious if anything like it had been made. 

A bottleneck doesn't even matter when dealing with hardware this old.

[WillOfTheLand]

1 minute ago, Fasauceome said:

A bottleneck doesn't even matter when dealing with hardware this old.

You are right, but I wasn't worried about my hardware having a bottleneck, I was just curious on the subject.

[Hairless Monkey]

1 minute ago, WillOfTheLand said:

You are right, but I wasn't worried about my hardware having a bottleneck, I was just curious on the subject.

The best bottleneck benchmark is the software you use and the games that you play. Monitor the usage of your components and try to determine for yourself which are holding back performance, if any.

[Zenith_X1]

Now what WOULD be really neat is if there was a bottleneck calculator that worked on a user basis, because really your bottleneck depends on your use case.  For example:

 

1) Is this a workstation computer or for personal use?

 

2) Personal use.  Do you use this computer for gaming?

 

3) Yes.  What genre of games would you like to optimize for?

 

4) First person shooters.  Input your PC parts here.

 

5) Done.  What resolution do you use for gaming?

 

6) 1440p.  Ok, here is how our benchmarking software estimates your system will perform under these 5 modern titles.

 

Because no one will ever have a perfect system with perfectly balanced CPU and GPU performance across all application.

[WillOfTheLand]

33 minutes ago, Zenith_X1 said:

Now what WOULD be really neat is if there was a bottleneck calculator that worked on a user basis, because really your bottleneck depends on your use case.  For example:

 

1) Is this a workstation computer or for personal use?

 

2) Personal use.  Do you use this computer for gaming?

 

3) Yes.  What genre of games would you like to optimize for?

 

4) First person shooters.  Input your PC parts here.

 

5) Done.  What resolution do you use for gaming?

 

5) 1440p.  Ok, here is how our benchmarking software estimates your system will perform under these 5 modern titles.

 

Because no one will ever have a perfect system with perfectly balanced CPU and GPU performance across all application.

My thoughts were more of a piece of software you could download like 3dmark, cinebench, heaven, whatever have you, and you'd run it and it would spit out a estimation afterwords like "your GPU is bottlenecking your CPU by 15%" or "your CPU is bottlenecking your GPU by 45%". It would run 2 tests, one to try and bottleneck your CPU and 1 to try and bottleneck the GPU. For the CPU this would be really easy, just give the GPU a really easy scene to render and have it pump out as many frames as possible as fast as possible and see where the CPU starts having problems, for the GPU you would just have the the CPU send it as many instructions as it can and just see how long it takes the GPU to chew through them.

[Zenith_X1]

5 minutes ago, WillOfTheLand said:

My thoughts were more of a piece of software you could download like 3dmark, cinebench, heaven, whatever have you, and you'd run it and it would spit out a estimation afterwords like "your GPU is bottlenecking your CPU by 15%" or "your CPU is bottlenecking your GPU by 45%". It would run 2 tests, one to try and bottleneck your CPU and 1 to try and bottleneck the GPU.

That could also be helpful, but it would need more than two tests.  With only two tests you would have to assume those two test are perfect averages of modern CPU-intensive and GPU-intensive workloads.  The trouble is, even with a single game you can have CPU and GPU bottlenecks in the exact same system, for example, if you run a game at 1080p you might be CPU-bottlenecked, while the exact same game and same system at 4k might be GPU-bottlenecked.  This is why I included so many questions in my example.  If those 5 questions only had 2 options to choose from, that would account for 32 possible variances in use case.

[WillOfTheLand]

3 minutes ago, Zenith_X1 said:

That could also be helpful, but it would need more than two tests.  With only two tests you would have to assume those two test are perfect averages of modern CPU-intensive and GPU-intensive workloads.  Even with a single game you can have CPU and GPU bottlenecks in the exact same system, for example, if you run a game at 1080p you might be CPU-bottlenecked, while the exact same game and same system at 4k might be GPU-bottlenecked.  This is why I included so many questions in my example

I updated my last post the moment I was getting the notification of yours.

[Zenith_X1]

6 minutes ago, WillOfTheLand said:

I updated me last post the moment I was getting the notification of yours.

Likewise.  So to add to that, those two tests you described would tell you the performance of the CPU and GPU (the same thing that 3D Mark does), but I think what I haven't made clear is that your bottleneck isn't the same as my bottleneck.  If I use my computer as a workstation with light gaming, and you use your computer for professional CS:GO, then we both have very different requirements for our system.  If you need to identify a bottleneck, you can tell other people on the LTT forums what application you feel is underperforming, and we will do an okay job suggesting what to upgrade without overspending.  Important to note is the application.  That's why the very first question in my example was about determining the use case for the system

[WillOfTheLand]

1 minute ago, Zenith_X1 said:

Likewise.  So to add to that, those two tests you described would tell you the performance of the CPU and GPU (the same thing that 3D Mark does), but I think what I haven't made clear is that your bottleneck isn't the same as my bottleneck.  If I use my computer as a workstation with light gaming, and you use your computer for professional CS:GO, then we both have very different requirements for our system.  If you need to identify a bottleneck, you can tell use what application you feel is underperforming, and we will do an okay job suggesting what to upgrade without overspending.

Well no benchmark is perfect, it could have multiple settings to the 2 tests you could tweak depending on the situation. My thoughts were just a general purpose benchmark. A sorta "you can expect this part to cause you issues if you try and use it for XYZ" sort of thing.

[Zenith_X1]

7 minutes ago, WillOfTheLand said:

Well no benchmark is perfect, it could have multiple settings to the 2 tests you could tweak depending on the situation. My thoughts were just a general purpose benchmark. A sorta "you can expect this part to cause you issues if you try and use it for XYZ" sort of thing.

A general purpose benchmark score is the CPU and GPU score of 3D Mark, which I know is not exactly what you're looking for.  You could skip programming a rendering engine entirely if the program just interpreted the benchmark scores of 3D Mark, Cinebench, and other tools.  Input your scores from those other benchmarks, answer a questionnaire about your use case, and viola, time to visit newegg and spend money.

 

Hopefully you'll credit me if you code such a program and start that business venture

[Hairless Monkey]

4 minutes ago, WillOfTheLand said:

Well no benchmark is perfect, it could have multiple settings to the 2 tests you could tweak depending on the situation. My thoughts were just a general purpose benchmark. A sorta "you can expect this part to cause you issues if you try and use it for XYZ" sort of thing.

There still seems to be a fundamental misunderstanding of what a bottleneck is.

 

 

[jones177]

 

44 minutes ago, WillOfTheLand said:

My thoughts were more of a piece of software you could download like 3dmark, cinebench, heaven, whatever have you, and you'd run it and it would spit out a estimation afterwords like "your GPU is bottlenecking your CPU by 15%" or "your CPU is bottlenecking your GPU by 45%". It would run 2 tests, one to try and bottleneck your CPU and 1 to try and bottleneck the GPU.

I run games with MSI Afterburner showing CPU and GPU percentages. 

Here is what it looks like.

 

This is SOTTR at 4k so the GPU is at 99%. At 66c the temps on the GPU are good. MHz is high and the GPU is using all the power it can with the power limit at 125%. The CPU is cruising as expected at 4k.

 

The one games I like to use is Shadow of the Tomb Raider for showing bottlenecks.. It shows the GPU bound percentage in the bench.

 

Hre is what it looks like.

SOTTR is set to 1440p with the high preset but the game says I am only using 49% of the GPU. This is with a RTX 2080 ti.

If I go to the highest peset I use more GPU but I have to raise the resolution to 4k to get 100% in the GPU. 

 

If I do the same test with a GTX 1080 ti all the GPU is used.