CPU Bottleneck & CPU Cooling Dilemma

Discussion in 'MSI' started by ronferri, Jul 18, 2019.

  1. yrekabakery

    yrekabakery Notebook Virtuoso

    Reputations:
    924
    Messages:
    2,738
    Likes Received:
    2,582
    Trophy Points:
    181
    The lower CPU usage is because a Hyper-Threaded CPU bottlenecks at a lower total utilization than a non Hyper-Threaded CPU. Which BTW the second video was CPU bottlenecked, as the GPU usage never reached close to 100% despite it dropping below 80 FPS at points. If you turned off Hyper-Threading on that i7-8750H, it would also load all cores to 100% like on my i5, with hardly a change in performance as BF1 does not really benefit from HT on 6+ core CPUs.

    Again, you're wrong. I don't know how else to explain to you. If you let a game run uncapped, there is always something limiting its performance, whether that's the GPU, CPU, or the software itself. That's what a bottleneck is. I feel like you do understand what I'm saying, but are arguing semantics of the sake of it. And I'm not sure what you mean by not an ideal CPU for BF games, as the 8600K was the second fastest gaming CPU at the time, and only lacked Hyper-Threading compared to the top dog 8700K, which I have already demonstrated BF1 doesn't really benefit from HT (nor BFV for that matter) on 6+ core CPUs.
     
  2. rinneh

    rinneh Notebook Prophet

    Reputations:
    851
    Messages:
    4,897
    Likes Received:
    2,186
    Trophy Points:
    231

    When people talk about a bottleneck they in general talk about when loading up the GPU as much as possible, when the CPU cannot run the game logic adequately enough, people start to talk about a bottleneck. Every game bottlenecks when you eliminate the GPU as much as possible, that is not a bottleneck. Now you are just testing the performance of your CPU. Which is hampered by the way because it is a i5. Frostbite is optimized for 8 threads currently thanks to the consoles. My i5 desktop which has been OC-ed to 5ghz still performs slower in BF1 than my 7700HQ laptop which cannot be overclocked. 3Dmark also shows a higher physics performance on i7s vs i5s. You are just lacking threads. I dont get why you think that HT doesnt hel pin both BF1 and BFV because it is the sole game series that made me ditch my desktop with an i5.

    But yeah it is semantics. But in that case there is always a bottleneck and that is not what the general gamers are talking about. But the 6700HQ/7700HQ is nowhere near a bottlenecking CPU for BF1 and already outdated for 2 years. BFV still runs fantastic on it together with a 1070GTX on high settings. most of the time near 80fps.
     
  3. yrekabakery

    yrekabakery Notebook Virtuoso

    Reputations:
    924
    Messages:
    2,738
    Likes Received:
    2,582
    Trophy Points:
    181
    Uh, I think you're OOTL. You're talking about old i5. My 8th gen i5 has 6 real cores. That's 50% more performance than old quad-core i5 on the basis of core count alone, while Hyper-Threading gives your i7 at most 30% more performance over quad-core i5 in synthetic tests like Cinebench and less in actual games. If you think 6 real cores is slower than 4 Hyper-Threaded cores in gaming, then you've been completely misled. And I never said HT doesn't help in BF1/BFV; it helps a lot on quad-cores. But with 6 or more real cores, the benefit of HT is much less pronounced in Battlefield.

    And if you'll excuse my doubt, I would really love to see proof of your 5GHz i5-7600K performing worse than your i7-7700HQ in BF1 when the CPU is the bottleneck.

    If you have a 120Hz display, and your CPU is limiting the frame rate to below 120 even with the settings turned down, then that's a CPU bottleneck, plain and simple. Not sure why this is so difficult for you to comprehend.
     
    Last edited: Jul 30, 2019
  4. rinneh

    rinneh Notebook Prophet

    Reputations:
    851
    Messages:
    4,897
    Likes Received:
    2,186
    Trophy Points:
    231
    Like I said, it is what I experienced in BF1. 6 cores jsut isnt enough for frostbite engine games. Well it is enough just you have a tad lower performance if it can call up to 8 threads. It is what it is. A laptop 7700HQ was quicker at 3.4hghz than my 5ghz 4core i5 without HT. That is what I said. Your CPU is performing fine, the i5 that is but could be a tad more with HT. When I disable HT on my 6core Blade 15, 5h3 performance drop in Frostbite games is still there. 6cores in laptops arent 50% faster than quadcores. Because barely any laptop can hold full boost clocks. Most are stuck sub 3ghz while 6700HQ/7700HQ laptops are more prone to holding 3.4ghz at all times.

    Regarding bottlenecks. Relieving the GPU of as much stress as possible and then complain it bottlenecks isnt really bottlenecking in a sense that your GPU is mismatched with your CPU. And other factors are at play. Frostbite performs best at 120fps and becomes iffy over that. Also load spikes because it is currently being horribly optimized doesnt tell the story of an underpowered CPU. Our semantics differ, but I think it is fair to say that your semantics arent aligned with what reviewers and most users consider bottlenecking.
     
  5. yrekabakery

    yrekabakery Notebook Virtuoso

    Reputations:
    924
    Messages:
    2,738
    Likes Received:
    2,582
    Trophy Points:
    181
    6 real cores is better than 4 Hyper-Threaded cores for gaming period, Battlefield or otherwise. Unless you can show proof of say an i7-7700K beating an i5-8600K or 9600K at the same clockspeed with the same RAM, then I will continue to regard that claim as BS, same as your claim about the i7-7700HQ beating the 5GHz i5-7600K. My CPU can hold 4.7GHz in games all day long thanks, and higher if I cared less about fan noise. If your Razer Blade is stuck at sub-3GHz in games when all 6 cores are loaded, then it is a junk gaming notebook. Even a budget tier Lenovo Y530 can hold its full 3.9GHz all-core boost clock on the i7-8750H all day long in a CPU intensive title like BF1 as I showed

    It’s called playing at competitive settings for lowest latency and highest visibility. BFV literally has a preset called Lowest Latency that sets everything to the lowest setting. If your frame rate is limited by the CPU, then that is a CPU bottleneck, no ifs or buts. Saying Frostbite performs best at 120 FPS and iffy over that is another arbitrary BS statement, but you’re getting quite good at that.
     
  6. GrandesBollas

    GrandesBollas Notebook Consultant

    Reputations:
    197
    Messages:
    237
    Likes Received:
    310
    Trophy Points:
    76
    As a NOOB, I am extremely interested in this topic. Found this forum thread on the net:

    https://forums.guru3d.com/threads/what-is-the-best-way-to-check-for-a-cpu-bottleneck.421643/

    "If the GPU doesn't reach 95%+ but at least one of the CPU cores does, the CPU is the bottleneck.

    If neither the GPU nor the CPU reach 95%, then the bottleneck is probably the game itself* or the RAM. Or you made the mistake of having vsync enabled, in which case you can't tell. Always test with vsync off.

    Keep in mind that there's almost always a bottlneck. It's quite rare for a game to fully utilize both the GPU and the CPU. In general, it's better to have a slight CPU bottleneck rather than a GPU one, as GPU bottlenecking will usually increase input lag."


    From the videos, the first video clearly shows a CPU bottleneck. All cores at or near 100%. The GPU well underused. The second video also shows the GPU not being fully utilized. Not clear if it is a CPU bottleneck or the game itself bottlenecking something.
     
    Papusan and yrekabakery like this.
  7. yrekabakery

    yrekabakery Notebook Virtuoso

    Reputations:
    924
    Messages:
    2,738
    Likes Received:
    2,582
    Trophy Points:
    181
    I would disagree on a CPU bottleneck being better than a GPU bottleneck. Yes, GPU-bound input lag sucks if you're playing a game competitively, but at higher Hz/FPS it's not as bad, and can be improved by setting maximum pre-rendered frames to 1 in the Nvidia Control Panel. A CPU bottleneck however can result in terrible stuttering, although this definitely seems to be game/engine dependent. BF1 for example still has relatively smooth frametimes when my CPU is the bottleneck, while BFV and Witcher 3 have terrible stuttering. Having Hyper-Threading can sometimes improve that stuttering.

    The smoothest overall experience is achieved by using a CPU-based frame limiter (as these have the lowest input lag outside of the game engine) that ensures consistent frame pacing (many in-game FPS limiters have godawful frametime consistency), and setting an appropriate FPS cap so that neither GPU nor CPU are bottlenecking. RTSS is the best limiter for doing this.

    The second video still shows a CPU bottleneck. I actually had my friend do that test based on my explicit instructions lol. Hyper-Threading improves core efficiency, meaning an HT CPU can do the same amount of work as a non-HT CPU, at a lower utilization %. With BF1 being a game that doesn't really benefit from HT on modern hexa-core or greater CPUs, that's the reason the total CPU utilization in the second video is relatively low despite the CPU still being the limiting factor. If he disabled HT, his CPU utilization would be maxed out like mine was or close to it. Also I believe he had a bit of a RAM bottleneck as well that was limiting his CPU/GPU utilization, since he was using 2666MHz CL15 (I have 3000MHz CL16), and the Frostbite Engine used in the Battlefield games loves memory bandwidth.

    On a side note, I recently got an NVMe drive and installed Windows 8.1 on it. I was very surprised to see a consistent ~20 FPS increase across the board in recent Battlefield games (BF4, BF1, BFV) compared to Windows 7, which I was using when I recorded that first video. No other games I have currently installed had performance improve in such a manner, if anything some of them saw a slight regression compared to W7. I did some Googling and found this Twitter log from the former rendering architect on the Frostbite Engine. It seems that DX11 and WDDM updates in Windows 8+ enable specific optimizations which improve performance in Battlefield games from BF4 onward. I believe Windows 7's lack of full DX11.1 support and WDDM 1.2/1.3 is also the reason for its lower Graphics Score in 3DMark Fire Strike, which has been known as a Windows 8/10 favoring benchmark for years.
     
    Last edited: Jul 30, 2019
  8. rinneh

    rinneh Notebook Prophet

    Reputations:
    851
    Messages:
    4,897
    Likes Received:
    2,186
    Trophy Points:
    231
    First of all, I never said that my 7700HQ beats your 6core i5. I said that the 4 hyperthreading cores boosted performance to such a level that a 3.4ghz core i7 4core was faster in games such as BF1 than a 5ghz core i5 with 2400mhz ram. DOnt create a false narration by yourself. Second, hyperthreading still matters even on 6 core CPU's if I turn HT off on my blade, the performance drop is about 25% not extremely huge but it is there.

    I never said my blade was sub 3ghz, it isnt. I have undervolted it and it is stuck at 3.4ghz at all time for all 6 cores (also in Prime95, I can guarantee that no lenovo will stay at 3.9ghz in that test). I spoke about the majority of laptops out of the box.

    Third, the so called low latency mode in BFV is nothing more than disabling future frame rendering which in the end just greatly lowers performance because your GPU is doing nothing half of the time, considering that as a test for bottlenecking is hugely flawed. If that is your measurement stick, than you should go home.....
     
  9. yrekabakery

    yrekabakery Notebook Virtuoso

    Reputations:
    924
    Messages:
    2,738
    Likes Received:
    2,582
    Trophy Points:
    181
    That is what you were insinuating with the nonsense about 8 threads and Frostbite Engine beating 6 cores etc. Don’t backtrack now, it’s not a good look. Your 2400MHz RAM speed is garbage for Battlefield and bottlenecking the i5 hard. Your Blade sees a drop when HT is disabled because you’re clocked in the low 3GHz range, while a desktop CPU clocked much higher around 5GHz with at least 6 cores and faster RAM sees much less difference from HT off.

    I was talking about clockspeeds while gaming, not Prime95. Way to move the goalposts.

    I suggest you actually check what the low latency graphical preset in BFV does. Hint, it changes settings on both the general and advanced video tabs, not just FFR. You’re the one who needs to stop posting BS and go home.
     
  10. rinneh

    rinneh Notebook Prophet

    Reputations:
    851
    Messages:
    4,897
    Likes Received:
    2,186
    Trophy Points:
    231
    No it is you failing to understand that I only said that Frostbite prefers 8 threads in the latest iterations and that is why a CPU running 8 threads is always more beneficial compared to a lower thread count even if it is hyperthreading. You failing to read is not a good look.

    Again It is just forward frame rendering what makes it a low latency preset. The rest are just graphical details which do not affect latency if you have the horsepower.

    You really lack clear understanding of what happens in game engines and GPU's. Get some basic dev classes in developing in common game engines......

    Also I aint moving the goalposts. I simply gave a baseline. Some games barely touch the CPU at all, while others do. BF1 for example is not that CPU heavy, it is thread heavy though. But it doesnt push the power usage over 40watts in the majority of todays gaming laptops. You are making insinuations based on your own bias. But to keep it in your ballpark, lets just say BF1 and BFV only fro now on.

    This thread has derailed quite a bit. But come with some counter evidence that a 6700HQ/7700HQ has bottlenecking issues at medium and high graphic presets in BF1, R6 siege.
     
Loading...

Share This Page