CPU&GPU Benchmarks

[Nightbringer]

I had run quick benchmarks and I think that they are miss-labeled. In GPU section, my GPU section, there was one character with a lot of effects, particles and my CPU struggled to keep up. In CPU section my software did not quite managed to log in hardware load, but it was full of enemy models. I assume that it was meant to test my GPU.

[BlackRedDead]

nope, the GPU test is correctly testing light&shadow effects, as those are most demanding on the GPU - it's really just not a graphics intensive game!

The CPU test on the other Hand stresses the CPU with lots of Units that has to be calculated (AI&Pathfinding is still mostly done by the CPU, even at Nvidia Graphics with Cuda Cores wich just take some very specific calculations that are often part of AI's and solve them way more efficient than a CPU ever could!)

If your CPU fails at the pretty non CPU intensive test already and you have framerate issues in the CPU test, both issues are bound to the CPU and/or your system configuration!
(remember, CPU also has to handle resource calls from the GPU - potentially you haven't enough VRAM and thus need to utilize your System RAM - try running the tests at lowest possible settings, if it persists it's a major problem with eighter your hardware (maybe just to old?) or it's configuration! - if not, then it's really just about your settings and capabilities of your hardware - check recommendated hardware requirements! ;-)

[Nightbringer]

I thought that particle effects are CPU based. These days everything seems to be dependent on GPU. Even when CPU benchmarking with hardware which is either above or at recommended value, you bottleneck your GPU and not CPU.

https://steamcommunity.com/sharedfiles/filedetails/?id=2858606787
https://steamcommunity.com/sharedfiles/filedetails/?id=2858606581
https://steamcommunity.com/sharedfiles/filedetails/?id=2858606516
https://steamcommunity.com/sharedfiles/filedetails/?id=2858606457

Though, I cannot understand and will always write it off to insufficient optimisation cases where you do not max out your hardware's potential and it still bottlenecks you. Though, in this case it seems to me that my GPU had bottlenecked hard when in CPU stress test. This is why I was initially confused. Why I'm being bottlenecked in CPU test by GPU and in GPU test I'm not bottlenecked by GPU? Poor test cases?

[BlackRedDead]

depends on how it's done - physics calculations are split between GPU and CPU nowadays - it's really not that easy to tell how much it is CPU and how much GPU bound, as the implementation matters!

your files can't be accessed - might be visibility? (change it to public on the given content page)

Dude, you will always experience a bottleneck - depending on what a game demands the most - in general RTS games tend to be CPU intensive, while FPS tend to be GPU intensive, and Open World titles or especially Sandbox games love eating your system RAM - but how much really depends on the game!
So, you always have to take some tradeoffs - unless you waste unholy amounts of money to buy the high end stuff only creators really need! ;-)

And the "GPU-Bottleneck" is likely due to the CPU not fast enough in feeding the GPU - check how much your GPU is actually utilized during the test - even Win10 Taskmanager is sufficient enough - unless like 80-100% usage of the GPU, it's NOT GPU bound! - with the CPU this is a bit more complicated, as nowadays you'll rarely fully utilize a 6-12 core CPU, especially with HT/SMT enabled, and can still be bottlenecked by the CPU because a game might not be able to use all cores - there you really have to look out for certain cores the game runs on, and if those are utilized above 95% it's CPU bound ;-)
The Problem with CPU limited is, you can't improve much - other than closing background tasks however - if GPU limited you can just turn down your settings^^

[Nightbringer]

Can you see them now? Previously I had uploaded them privately and thought that people could view them with a link. This clearly is not a case and now I just made all of screenshots I made public in order not to mess up which ones I posted and which ones I did not. Hopefully, they will be available via same links and if not, I'm going to repost them again.

[BlackRedDead]

Nope, you have to go onto each page and change visibility to public individually - the links won't change, it's (content) ID based ;-)

[Nightbringer]

Hopefully now everything works fine. Max out pictures and there you will see MSI Afterburner giving real time performance.

https://steamcommunity.com/sharedfiles/filedetails/?id=2858606516
https://steamcommunity.com/sharedfiles/filedetails/?id=2858606457
https://steamcommunity.com/sharedfiles/filedetails/?id=2858606581
https://steamcommunity.com/sharedfiles/filedetails/?id=2858606787

What gave me red flag was that during CPU and GPU test, my GPU was bottlenecked. In GPU test, we see a massive improvement in framerates. In CPU test it is drastically reduced. Despite CPU not being throttled, in both cases GPU throttles the performance of these benchmarks. In GPU test, my GPU is far less used than in CPU test and it gives me far better game performance...

My GPU is one which is officially recommended by the developers. In my view, these are just poor benchmarks to test hardware on as they tell next to nothing about your hardware. This is why I was confused. We have here an GPU test which does not test GPU and an CPU test which tests GPU more than an GPU test...

[Nightbringer]

There is a case to be made with hardware response time as in CPU test, CPU spends more time calculating than GPU drawing the image. Faster CPU with higher frequencies, more cache, faster memory would perform better. I'm still struggling to understand if this avg delay time means how much time it spends to render one or all FPS images. There is a connection obviously as three times longer waiting period resulted in X less framerate.

However, this raises a question, why my hardware is not fully utilised in a first place? If game's code would properly divide all the work through cores, processor would be fully utilised and we would then could blame hardware for not being strong enough. I do not have problems when my hardware struggles.

For example, I'm almost done playing Stalker Call of Pripyat. This game stutters, because its application is heavy on a single core and sends only some of its work to others. Considering that my early quad core processor weren't much better or even weaker in single core performance than old mono-core processors, that is not surprising. That is down to development practices of that time and poor Directx emulation which spends a lot more resources to do the same task and can't properly divide that task through the processors. That is a matter of getting a stronger hardware. In this case it seems that stronger hardware would have low correlation with game's performance as game is simply not optimised to utilise CPU efficiently.

[BlackRedDead]

Originally posted by Nightbringer:

Hopefully now everything works fine. Max out pictures and there you will see MSI Afterburner giving real time performance.

https://steamcommunity.com/sharedfiles/filedetails/?id=2858606516
https://steamcommunity.com/sharedfiles/filedetails/?id=2858606457
https://steamcommunity.com/sharedfiles/filedetails/?id=2858606581
https://steamcommunity.com/sharedfiles/filedetails/?id=2858606787

Holy cow, that CPU is old! - i thought i hold the record for longest usage with my Athlon 64 X2 6000+ over ~8 years till Ryzen came out xD
Idk if that's actually the problem cause, but you should consider upgrading your system mate - the GPU should still be able to run most games reasonably settings wise (not the lowest specs, but definitively also not ultra)
If you consider Ryzen, be aware that Windows 7 only runs on Zen 1! - so till 1800X and APUs 2400G - every 2000er without G is Zen+ and only runs with Windows 10 (maybe also 8, but who cares about that?) and beyond! - with intel you have to research yourself, i only know that the 6700K does still run with Windows 7! (and is indeed still a decent CPU - but i would recommend going more modern, especially if you want to use it longer than 2 years! ;-)
It's not about the Gigahertz, it's about newer and more efficient architecture! - so a recent 3Ghz CPU can outperform an old even 5Ghz CPU - but it's a question about at wich task! ;-)
But you can guess that the more modern the game is, the more modern the hardware was it is developed upon, and the more relied it is upon what that hardwares capabilities was! ;-) (tough, it's not strict - the question in detail is what of those capabilities the developers actually took advantage off)

You might also want to upgrade to at least an SATA SSD, tough, if you upgrade to a board with an M.2 slot, you should get an NVMe drive with 120-150GB as System drive (and for some very Storage load heavy programms like some games (sandbox ones especially), and a 500GB-1TB SATA SSD as Game Storage, and use your HDDs for Data like musik, pictures and even videos till 4K (compressed however!) ;-)
You can cheap out on capacity of the system drive ofc, but the price per capacity just goes up below that! (and also the theoretical lifespan, tough we mostly still talk above 25TB TBW - i have like 10TBW by now on my 2year old Sytem Drive, wich i use to store multiple instances of my 32GB RAM too! - i think, everything that offers 100TBW is fine for average users - it's really more the time and propably heat that determines an SSD's lifespan! ;-) - so if you intend to write larger amounts than ~4GB at a time onto the NVMe drive, consider investing a few bucks into a passive cooler - wich, depending on design, offers eighter more thermal mass or more radiation surface, and thus eighter longer writing intervals or faster cooling & a lower overall temperature, once it's thermal mass is satisfied! ;-) - but all that matters is that you have a proper heatsink and not that heattraps most motherboard manufacturers strap onto their boards to hide SSD's - srsly, don't use them unless you really only store small data on that SSD and nothing that could be write intensive at all!)
Oh, and don't bother with PCI-E Gen.4 NVMe, just use the regular Gen.3 NVMe drives, but with at least NVMe revision 1.3, as that's the one they introduced drive selftest features and secure erase, wich is quite importand if you buy a used one ;-)

Originally posted by Nightbringer:

What gave me red flag was that during CPU and GPU test, my GPU was bottlenecked. In GPU test, we see a massive improvement in framerates. In CPU test it is drastically reduced. Despite CPU not being throttled, in both cases GPU throttles the performance of these benchmarks. In GPU test, my GPU is far less used than in CPU test and it gives me far better game performance...

This is indeed strange, please report back to the devs!

Originally posted by Nightbringer:

My GPU is one which is officially recommended by the developers. In my view, these are just poor benchmarks to test hardware on as they tell next to nothing about your hardware. This is why I was confused. We have here an GPU test which does not test GPU and an CPU test which tests GPU more than an GPU test...

First of, they are not meant to benchmark hardware - there are other tools for that! - they are meant to give the user two predefined&predictable scenarios, to simply test their settings for the game! ;-)
Second, it still isn't as you perceive it - the GPU test is meant to stress the GPU without utilizing the CPU to much - as lighting and shadows are the biggest GPU hogs, it does pretty good at it!
The CPU test on the other hand is meant to test the opposite, and there it is indeed strange that your GPU runs at 99% despite there aren't as many intensive effects on screen - but i don't see that behavior with my hardware, so it's still eighter a problem with your system (i still bet on the CPU), or with that specific model of GPU - needs further investigation!


Edit:
to clarify, the tests are made in a way that achieves their purpose - they aren't made with worst case, but with bad case szenarios that a player will encounter at some point within the game, and really are only there to give players a tool to tailor their settings to their hardware capability - period
(you could do that with safegames too, this is just a shortcut and much more consistent and predictable than doing it manually)

Edit 2:
try that test again with lowering lighting&/shadows one and two steps down - it might has to do with raytracing - both scenes sport some complex lighting & shadows - try both independently, so 8 tests in total - and compare them - make sure you label them correctly, so reducing lighting settings gpu&cpu test 1&2 setting steps lowered, and the same with shadows - then you might see wich of those are heavyer on the performance - tough, you might need to run 8 aditional tests with a combination of lighting and shadow settings lowered xP
Have Fun :D (but if you find that tedious, try to overclock and undervolt your RAM xP - 2 weeks minimum, if you hunt the sweetspot, and a countless number of stability tests! - and the calculations inbetween if your combination really is faster, or just an inefficient downgrade to a previous but "slower" setting xP (again, it's not only about the Gigahertz!))

[Nightbringer]

I had disabled shadows completely. The only lighting option was raytracing which was disabled by default.

https://steamcommunity.com/sharedfiles/filedetails/?id=2860780089
https://steamcommunity.com/sharedfiles/filedetails/?id=2860780115
https://steamcommunity.com/sharedfiles/filedetails/?id=2860780063
https://steamcommunity.com/sharedfiles/filedetails/?id=2860780033

With this monitoring software, CPUs seem to be throttled. What however is strange to me is that performance barely had changed. GPU loaded images slightly faster and CPU was slightly slower. This signifies less work being done by GPU and thus CPU have to do more work. This had translated into slightly higher frame rate and had made CPU a bottleneck.

https://steamcommunity.com/sharedfiles/filedetails/?id=2860786422

This is the same benchmark, but with different monitoring software. It seems the issue is with MSI Afterburner and its whacky averaging of processor loads.

https://steamcommunity.com/sharedfiles/filedetails/?id=2860786422

I had put individual CPU usage into a graph:

https://steamcommunity.com/sharedfiles/filedetails/?id=2860788848

I did same test again with previous software:

https://steamcommunity.com/sharedfiles/filedetails/?id=2860790546

It seems to me that MSI Afterburner is an unreliable software in detecting bottlenecks. It have around 10% margin of error when calculating CPU workload. Thus if your hardware is at 90%, it can be considered to be bottlenecking performance of a game.

I had checked monitoring software CPU usage. MSI Afterburner seems to use slightly more resources with slight lag spikes (using upwards 1,1% CPU from typical 0,4%). FPS Monitor however tends to use 0,5% while being overall more stable program.

I had noticed that task manager also does not recognise bottlenecked CPU. MSI Afterburner mirrors Task Manager information which hints that they are using same source for data. When task manager says that my CPU is bottlenecked, bottleneck also shows up in game. FPS Monitor on the other hand says that my CPU is bottlenecked all of the time. In aftermath data, render latency shows that CPU is holding back game's performance. This indicates to me that source which Task Manager and MSI Afterburner uses is imprecise in benchmarking.

I had made side by side comparisons of both programs running simultaneously:
https://steamcommunity.com/sharedfiles/filedetails/?id=2860847617
https://steamcommunity.com/sharedfiles/filedetails/?id=2860847601

It appears that MSI Afterburner does not register information accurately when system starts to throttle. It always under-reports system performance.

In conclusion, tests shows CPU as a bottleneck in CPU test and GPU as a bottleneck in GPU test. After testing, I had discovered that MSI Afterburner cannot be trusted in detecting system bottlenecks and I should consider anything coming to 90% in that software to be a potential bottleneck due to margin of error from imprecise method of calculation they are using.

Never trust branded software. MSI Afterburner seems to be crap. I have similar woes with Nvidia Experience being crappy.

[BlackRedDead]

First of, i read trough all that again - my conclusion is that your CPU is just to old and doesn't feature function the game requires!
The GPU on the other hand, especially considered that you deactivated Raytracing, is still a bit strange - but not that much, it's still just old hardware that propably lacks technologies the game relies upon mate - you also running the game on QHD (1440p), try FHD (1080p) or better even just HD (720p, wich would have the benefit of your display using 4 pixels for one renderpixel, resulting in a clearer image!)
Also CPU utilization doesn't mean that much with multicore CPU's, as the workload is split between them and the % you get shown is an average between them - some cores run at 99% while others run at 0% at a given moment is totally normal - usually it's somewhere inbetween and can still be a bottleneck! (usually due to simply missing improvements in it's architecture that the game in question relies upon - it might still run, it might be done another, simply slower way without getting utilized as much, or making the cpu sweat because it is fully utilized - not that easy anymore - GPU's start to get more complicated too, your 1070 already has CUDA cores wich could be fully utilized and your GPU reporting back it's BORING, while in reality it's crunching numbers very efficient but indeed not doing much on the rest of the GPU! ;-)

But what i found during my own testing and a direct comparson between the test, is that indeed the GPU is slightly more utilized during the CPU benchmark than the GPU benchmark (in my case ~66% to ~44%) - but in hindsight it makes totally sense - there is much more to be drawn in the CPU test, and as far i can tell, they use real 3D models and not just complex 2D sprites for their entities - so there is a lot more to process when you draw hundreds of them onto the screen and calculate even basic shading, shadowing&light from&onto them ;-)
(tough my RX580 simply isn't even getting hot - despite i set it back to default settings for the test - i usually run UV profile and customize game settings to get rid of unwanted effects and tailor it as close to my preference as possible, and also limit my FPS and use Freesync - wich i have to underclock too now, just recently found out that you can UC/OC your monitor xD - 48Hz isn't low enough for me, i want at least 42Hz as lower end! :P)

If you want to shop for a new GPU, prizes got sane now - especially on the used market ;-)
i would suggest eighter AMD 5700X(T) or Nvidia 2060S as lower end, you should already see considerable performance increases with those, while not deadend yourself about resellability ;-) (like with my RX580 - it's a good GPU, but was already middleclass at release, and now is the bare minimum i would recommend for a budget gaming PC)

[Nightbringer]

You can see my photos from previous post? As it seems you were talking about old tests, in new photos you can see each core's utilisation. The problem appears to be with MSI Afterburner reporting as it differs from FPS Monitor reporting when CPU reaches 90%. Software seems to incorrectly report bottlenecks and thus it appeared like CPU being under-utilised while GPU throttles CPU benchmark.

My processor is two generations ahead of what developers are recommending - Intel i7 gen 2. So, my CPU is good enough, otherwise developers had made a mistake in recommending systems specs. GPU is also exactly on point with recommended hardware. Developers had made their game with systems such as my own in mind. Especially knowing that GTX 1060 is the most popular graphics card at the moment.

As for an upgrade, I had considered buying 1XXX/2XXX Ryzen from ebay with motherboard, cooler and RAM this weekend. Though, I decided to wait out this month or a next, because now I have to recover financially from various things.

I want it, because I had managed to damage CPU pins on my motherboard when I was taking my CPU to be delidded. I bought myself i3 gen 4 processor for testing, so there was a lot of inserting and removing of CPUs. First to see performance of CPU without liquid metal, then with liquid metal. When I practised with i3, I went with my i5. I had idea to upgrade to i7 gen 4, but they were overly expensive and I lost appetite for it after finishing with liquid metal and seeing damaged pins. I considered buying early used Ryzen bundle, because it is relatively cheap, around 150 euros. It would be an upgrade for my existing system and an interim system with whom I could experiment, adding custom liquid cooling, LCD screens, etc. My original plan was to wait until Meteor lake and then buy into it. I think it is pretty symbolic to wait out for 10 whole generations and upgrade my own gen 4 system into gen 14.

Previously I had GTX 760 card. It sadly broke down this spring and I had to replace it with most reasonable alternative. I chose GTX 1060 6gb. I previously had purchased GTX 780 Ti at the start of an year. I needed to play Space Hulk and record footage for it and my GPU was too weak for that. GTX 780 Ti could carry all the games I played. However, I did not liked that GPU. Lack of Directx 12 support. 3 GB of RAM instead of 4. Terrible base cooler design. I sold it off and stuck with my GTX 760. I thought, it could carry me until next generation, but sadly it died and I was forced to upgrade.

Since then I installed AIO cooler, refreshed system fans as they seemed to have died. Added additional SATA drive solely for video footage. I bought adapter cable for my motherboard and probably I will buy 4 pin motherboard extension cable. Other than that, I'm done with this system. I will retire it to my parents house. As for new computer, I'm already slowly looking up for parts. I purchased aorus c700 case. It is sitting empty at the right side next to me. I will be waiting for next generation. I want AMD 7000 series card, because it associates with a time when I purchased my GTX 760. It was this or AMD HD 6000/7000 cards. I like how they went full circle with naming scheme and I feel nostalgic about it. Plus, all the BS Nvidia is doing with their own pricing. I also hope that Intel still dominates gaming market with its Meteor lake, though, they are not looking so hot anymore. Ryzen 7000 would make more sense to me, buy Godlike motherboard and Ryzen 7600x CPU. Then wait until end of motherboard lifecycle and purchase their 3D processor as an upgrade. Intel really needs to prove itself as a superior gaming CPU to me. I do not like its efficiency cores philosophy which manages to throttle the gaming performance. Move those efficiency CPUs into the background, make some smart AI which automatically dedicates these CPUs to all background tasks and leave performance cores to gaming. I would like that and I cheer for Intel to gets its crap together and start dominating market once again. As it seems now, AMD is heading towards the premium brand and Intel will start producing high volume, performance processors.

As for why I'm not so enthusiastic about upgrading my system. Well, you can deduce that from my steam game review history. I do not play games which require modern hardware. The last game I played was Stalker Call of Pripyat. My system barely ran that game, but that is because of grossly unoptimised mods. Base game struggles with maxed out settings, but it still manages up to 60 FPS. All the games which I had in my lists are less graphically demanding than this one. Even my old GTX 760 was good enough for them and considering that old games aren't good lookers, you are not missing much by reducing graphical settings to medium if one or another game struggles to run. I will upgrade my system next year to high end/enthusiast level PC, but that is more of a vanity project. I do not imagine where I would need such powerful PC, outside of fulfilling one of my youth's dreams.