Sticky

Ultimate Bottlenecking Guide Rev. 2.0

Welcome back good people to revision 2.0 of the ultimate bottlenecking guide! It has been overdue for a long while now...but I was finally able to find time to bring it together!

I have streamlined the new spreadsheets tremendously vs. the old bottlenecking spreadsheets. Several improvements include grouping up all non overclockable chips into one name, putting all graphics card models into one spreadsheet, and making everything easier to read.


Difficulties In Measuring Bottlenecks:

The reason why bottlenecking is so confusing is because it's on a game to game basis and a frame rate basis. Games A, B and C bottleneck, but X,Y and Z don't but if you have a FPS of 200fps or more on those games, the results could be the complete opposite. This is why narrowing down which CPUs bottleneck which GPUs can get extremely difficult.
****************************************************************************

I’ve broken down the type of bottlenecking in 4 colors:

Black = No bottlenecking issues.
Magenta = CPU bottlenecking GPU only in a worst case scenario.
Blue = CPU bottlenecking GPU only in more advanced/CPU intensive games (like Crysis 3).
Red = CPU bottlenecking GPU in all gaming applications.

Resolutions (In the spreadsheets, each category presume a specific resolution):

GeForce GTX 1080 Ti Category = 1440P
GeForce GTX 1080 Category = 1440P
GeForce GTX 1070 Category = 1440P
GeForce GTX 1060 Category = 1080P

****************************************************************************

Intel Bottlenecking Chart

AMD Bottlenecking Chart

****************************************************************************
What is Bottlenecking?

Bottlenecking is where one component is hindering another components performance/efficency.

Truth be told, there is ALWAYS a bottleneck in a computer. Like a CPU being bottlenecked by a GPU -- yeah that’s legit -- but that isn't always a bad thing. When one component is not at 100% utilization, that means it’s being bottlenecked by something whether it’s temps, fans, software utilization etc.


Why Do GPUs Get Bottlenecked by CPUs (When Gaming)?

In a gaming oriented computer, the CPU is the 2nd most important component in your system. The CPU’s job is to send pre-rendered frames to the GPU. The contents of pre-rendered frames are basically anything not related to what the GPU will render. A good example of this is the positioning/location of AI and the positions of your teammates and enemies.

While pre-rendered frames aren’t as hard to render as fully rendered frames, it still takes quite a bit of power from the CPU to render them out --of course this depends on the game engine--. This is why you still need a powerful central processing unit for any kind of modern or advanced game you want to play.

Bottlenecking is Also Affected by Frame Rate:

Supahos, another member here at TH, describes this subject very accurately:
Supahos said:
Every CPU is capable of only a certain number of frames in every game. The resolution and settings have very little effect on CPU sided FPS limitations. This is why if you drop graphical settings in simple games your FPS doesn't go up. The CPU was the hold up and the lowered settings don't help your CPU.

As an example, if you are playing League of legends at 720P with a GTX Titan XP, your CPU is holding you back from more frames. (I don't care if you're running a 7700K overclocked at 5.2GHz, it's still the weakest link.) That same system playing Battlefield 1 at 4K ultra settings would make the Titan "bottleneck" the 7700K.

Every GPU has a maximum number of frames it can render at any given setting/resolution. Higher the settings or resolution the less FPS it can possibly render.

Okay so where does the bottleneck come in?

In any given game and setting whichever is lower for maximum possible frames (CPU or GPU) is the fps you'll get roughly.

A GTX 1070 at 1080P in League of Legends will most likely get the same FPS as a GTX Titan XP, since the CPU was the hold up to begin with.

Conversely, a 7600K at 4K max settings paired with a GTX 1070 will have the same FPS as a 6950X paired with a 1070. Because the GPU is the bottleneck in that instance.
26 answers Last reply
More about ultimate bottlenecking guide rev
  1. I would LOVE your feedback on my new charts. Specifically I need info as to whether my educated guesses on each CPUs bottleneck is true or not. If you can put your feedback in this thread, that'd be extremely helpful!
  2. I'd take it back to Ivy-Bridge, or even Sandy-Bridge, there's many whom still use them.

    Just looking at all those red Haswells, especially the i7's I find misleading. My 3770k never goes past 55%, my gtx970 never gets past 63% unless I drop a 4k DSR. So since the i7-4790k is bright cherry red, it must mean my 3770k is being bottlenecked by everything. Yet I'm still way beyond 60fps in everything I play, and can still get 300fps in CS:GO. Granted I'm not playing BF1 or gta5 or Witcher 3, so maybe a little clarification as to the boundaries might be warranted. With a 970 basically not much different to a 1060/3Gb, 1080p isn't much of an issue to most games, only the uber new and intense graphical/cpu monsters. Even at ultra, gta5 should be getting 60-80fps on average at 1080p. And that's what ppl will be concerned with.

    Maybe I'm getting too deep into it, but saying that i7 is going to 'bottleneck everything' is gonna freak ppl out. Especially when there's a hundred+ posts telling ppl to go ahead and wait, they don't need to upgrade, the i7-4790k is still a very good cpu that's really just as good as the i5-8600k for most things.

    For AMD, the Ryzens have equitable IPC to Haswells, so on anything 4c/8t or less, a Ryzen 5 1600 should run roughly equitable to a i7-4790k, yet a Ryzen 5 16xx is magenta, only bottlenecked in extremes, but the i7-4790k is red, bottlenecked by everything... Now I'm really confused.
  3. People need to realize that there are other factors that can make one think there is a bottleneck.

    What other programs are running in the background that's taking up memory or CPU utilization?
    Is Windows or any other program trying up/download any information that is taking up networking bandwidth?
    Is the CPU heat sink properly installed to prevent overheating and CPU throttling?

    This is just what I came up with off the top of my head. Feel free to add any more.

    I do thank you for posting this. If people read it, it should stop the, "I've got an Intel 8700K processor and a GTX 1080Ti. Will my system bottleneck" questions. :heink:

    -Wolf sends
  4. Karadjgne said:
    I'd take it back to Ivy-Bridge, or even Sandy-Bridge, there's many whom still use them.

    Just looking at all those red Haswells, especially the i7's I find misleading. My 3770k never goes past 55%, my gtx970 never gets past 63% unless I drop a 4k DSR. So since the i7-4790k is bright cherry red, it must mean my 3770k is being bottlenecked by everything. Yet I'm still way beyond 60fps in everything I play, and can still get 300fps in CS:GO. Granted I'm not playing BF1 or gta5 or Witcher 3, so maybe a little clarification as to the boundaries might be warranted. With a 970 basically not much different to a 1060/3Gb, 1080p isn't much of an issue to most games, only the uber new and intense graphical/cpu monsters. Even at ultra, gta5 should be getting 60-80fps on average at 1080p. And that's what ppl will be concerned with.

    Maybe I'm getting too deep into it, but saying that i7 is going to 'bottleneck everything' is gonna freak ppl out. Especially when there's a hundred+ posts telling ppl to go ahead and wait, they don't need to upgrade, the i7-4790k is still a very good cpu that's really just as good as the i5-8600k for most things.

    For AMD, the Ryzens have equitable IPC to Haswells, so on anything 4c/8t or less, a Ryzen 5 1600 should run roughly equitable to a i7-4790k, yet a Ryzen 5 16xx is magenta, only bottlenecked in extremes, but the i7-4790k is red, bottlenecked by everything... Now I'm really confused.


    Actually, Ryzen's IPC is in-between to equal with Skylake, it's just the clocks speeds are very similar to haswell/skylake.

    Thank you for your input, I have made a few modifications.
  5. Got a question about the Intel chart, specifically about i5-6600K (which i'm also running in my Skylake build).

    How come that there's bottleneck under worst case scenario with GTX 1070 but no bottleneck with GTX 1080 / GTX 1070 Ti?

    Oh, there's error in Intel chart as well in Skylake -> Core i5 section. There is no i5-6700K CPU in the world. What is, is the i7-6700K. Though, it looks like you have a typo there and it should be i5-7600K and not i5-6700K.
  6. Can someone elaborate on the "worse case scenario" and what that might be?

    "CPU bottlenecking GPU only in a worst case scenario."
  7. unknown100 said:
    Can someone elaborate on the "worse case scenario" and what that might be?

    "CPU bottlenecking GPU only in a worst case scenario."


    Worst case scenario are for extremely CPU demanding games like Ashes of the Singularity. Typically large RTS games and to some extent, some of the massive open world games (Star Citizen & World of Warcraft come to mind) take up lots of CPU, to the point you'd probably be better off going with say a core i7 over a core i5 and a lowering your graphics card to a 1070 if you were thinking of a 1070 ti (as an example).
  8. Anything that uses scripted mods, from skyrim to minecraft to gta:v can get seriously cpu bogged especially games that run with things like scripted graphics mods like ENB's. I have enough mods + ENB that even though Skyrim is predominately single thread heavy using 2 threads vanilla, it's currently seeing @50-55% usage on 6 threads on my i7. Same thing on my i5 I had to drop the ENB and a bunch of mods simply to keep fps 50+
  9. I still like some clarification on my last question. Here's it as simple as possible:
    i5-6600K + GTX 1060 = no bottleneck
    i5-6600K + GTX 1070 = bottleneck under worst case scenario
    i5-6600K + GTX 1080 = no bottleneck

    How come GTX 1070 produces minor bottleneck while much weaker GPU (GTX 1060) doesn't? Same with i5-7600K too.
    And since i'm planning to upgrade my GTX 1060 to GTX 1070, i'd like to get some heads-up in this matter.
  10. Thank you Aeacus, that was a glitch.

    If you want, I can give editing rights to you aswell if you have a google account so you can make changes.
  11. TechyInAZ said:
    Thank you Aeacus, that was a glitch.

    If you want, I can give editing rights to you aswell if you have a google account so you can make changes.

    While i do have a Google account, i don't have access to the data which was used to create this chart. And without that, i can't make corrections based on my logic alone.

    Just saw that you corrected the i5-6600K and GTX 1080 where there's minor bottleneck (pink) but the correction isn't done with i5-7600K which still shows:
    i5-7600K + GTX 1060 = no bottleneck (black)
    i5-7600K + GTX 1070 = bottleneck under worst case scenario (pink)
    i5-7600K + GTX 1080 = no bottleneck (black)
    i5-7600K + GTX 1080 Ti = bottleneck under worst case scenario (pink)

    Though, when looking closer, there are some inconsistency in "Haswell / Ivy Bridge-E" section as well. For example:
    i5-446x + GTX 1060 = bottleneck under worst case scenario (pink)
    i5-446x + GTX 1070 = bottlenecking in everything (red)
    i5-446x + GTX 1080 = semi-bottlenecked in CPU demanding games (blue)
    i5-446x + GTX 1080 Ti = bottlenecking in everything (red)

    As far as i know about CPUs and GPUs, if you upgrade GPU, you'll create bigger gap between the two and bottleneck becomes more severe. And that's why it doesn't make any sense for me why better GPU (GTX 1080) offers less severe bottleneck over lower performing GPU (GTX 1070) while in reality, it's vice-versa.
  12. I have a question about this chart, what technical process was used to compile the data that confirms all this "excel sheet", or is this "collected data from hearsay ?
  13. Yes it is all by hearsay.

    How this works, is that I semi guess what stuff is going to bottleneck, then i have you guys correct me (hearsay) so that the chart gets more accurate.

    For a 100% true bottlenecking chart, I'd need every single one of those CPUs and GPUs to test with which would take forever.
  14. There are plenty of videos on the Youtube where most of these setups are tested within games. Since each game reacts differently, it's hard to compose such chart that covers all games.

    I'd split the chart up for 2 parts; one for CPU bound games and another for GPU bound games for better classification. But then again, not all people know which games are CPU or GPU bound and listing all games takes ton of time.
  15. this will help you a LOT http://thebottlenecker.com/
  16. @Amitsandhu943 that website is complete garbage
  17. amitsandhu943 said:
    this will help you a LOT http://thebottlenecker.com/


    It helps me a lot, in knowing who to not listen to.
    That site is worse than useless.
  18. amitsandhu943 said:
    this will help you a LOT http://thebottlenecker.com/


    Yeah never ever use this steaming pile of crap.
  19. It's hilarious. My 3770k is a perfect match stock @3.5GHz with a gtx970, yet my 3.4GHz i5-3570k is a 10% bottleneck, I'd better replace it with a 7820HQ. So who wants to rip apart their laptop to donate its guts to my pc? What a freaking joke.
  20. I got a question:
    When i had a Q6600 at 3GHZ with GTX 660 OEM 1.5GB GDDR5, it gave me in GTA Online very high settings 30-40 ish FPS.
    The CPU load was 95-100% and GPU at 60-70%.

    Now i have a Phenom x4 955 at 3.8GHZ with GTX 660 OEM 1.5GB GDDR5, it gives me at GTA Online at very high settings 50-80FPS.
    But i am concerned about the GPU load. CPU load is at max 92% and GPU load is max 40%. Why is my GPU load so low?
  21. budgetgamer12345 said:
    I got a question:
    When i had a Q6600 at 3GHZ with GTX 660 OEM 1.5GB GDDR5, it gave me in GTA Online very high settings 30-40 ish FPS.
    The CPU load was 95-100% and GPU at 60-70%.

    Now i have a Phenom x4 955 at 3.8GHZ with GTX 660 OEM 1.5GB GDDR5, it gives me at GTA Online at very high settings 50-80FPS.
    But i am concerned about the GPU load. CPU load is at max 92% and GPU load is max 40%. Why is my GPU load so low?


    Since that doesn't seem to be a bottlenecking issue (more like a driver/software bug), I recommend making your own thread regarding the matter.
  22. TechyInAZ said:
    budgetgamer12345 said:
    I got a question:
    When i had a Q6600 at 3GHZ with GTX 660 OEM 1.5GB GDDR5, it gave me in GTA Online very high settings 30-40 ish FPS.
    The CPU load was 95-100% and GPU at 60-70%.

    Now i have a Phenom x4 955 at 3.8GHZ with GTX 660 OEM 1.5GB GDDR5, it gives me at GTA Online at very high settings 50-80FPS.
    But i am concerned about the GPU load. CPU load is at max 92% and GPU load is max 40%. Why is my GPU load so low?


    Since that doesn't seem to be a bottlenecking issue (more like a driver/software bug), I recommend making your own thread regarding the matter.


    Thank you for your quick answer!
    I made my own thread.
  23. Thanks for this! I just ordered an i7-4770 and I'm planning to upgrade to a GTX 1070 soon but I was worried that my CPU may bottleneck it. Good to see it won't!
  24. I7 4770 at 4.1Ghz (so 600mhz oc over stock ) locked on all cores permanently (not throttling cus max temp is 57 Degree Celsius(liquid metal)) with 1070 (+70 mhz).. BF 1 Multiplayer 64 player server, 1080p all ultra settings and 125% resolution scale (so almost 1440p)
    99% cpu usage
    60-70% gpu usage
    It does lag and framedrop a lot less than my previous fx8370 but still.. Can't maintain solid 75hz of my monitor (drops to 50's.. With fx8370 were drops to 30s and more frequently).
    Maybe it's my 16gb 1333 mhz ram???
  25. TechyInAZ said:
    ...

    Hi,
    @TechyInAZ - are the charts prepared for Overclocked CPUs or for the base clocks ?

    Thanks for the help !
  26. Base clocks for all except the K skus, K skus I assume a very mediocre overclock.

    However, those charts are now not up to date, so don't follow them exactly like their 100% true.
Ask a new question

Read More

Games Bottleneck Graphics Cards Spreadsheet