CPU core count important?

[atlau]

Not at all hijacked Daddy Ace... I'm actually curious following your questions as im wanting to make sure the 1060 will suffice. I basically want to build a system that can run full graphics without buying more than I need.

[DaddyAce]

Oh good atlau, you likely saw this, but in case you didn't, this was what I came up with:  http://bbs.hitechcreations.com/smf/index.php/topic,382567.0.html

My goals were similar to yours, although I went with a more expensive case & Mobo than I need to, because I wanted lots of USB ports, etc.

Best of luck with your build; I'll be interested in what you do.

 

[38ruk]

Should easily run with everything at maximum, except environmental mapping.  Still need that left at the first notch.  Basically, hit the "Default" button at the Options->Graphic Detail page.

Go offline and use CM God's Eye mode to get above the terrain a good bit.

If you look up, with no clouds in the sky, you should be getting close to a 1000 FPS (theoretical maximum).  I know a 980Ti will do that and the 1060 is close to, if not faster, than the 980Ti.

From what ive seen the 980ti and 1070 are neck and neck .  The 1060 and 980 (non ti ) are within 2- 5 fps in a bunch of the benchmarks ive seen .

[atlau]

Oh good atlau, you likely saw this, but in case you didn't, this was what I came up with:  http://bbs.hitechcreations.com/smf/index.php/topic,382567.0.html

My goals were similar to yours, although I went with a more expensive case & Mobo than I need to, because I wanted lots of USB ports, etc.

Best of luck with your build; I'll be interested in what you do.

 

Yes I actually had found that thread in my search. I'm hoping to find some  good deals after Thanksgiving! Tired of playing on my laptop:)

[Bizman]

Based on another thread where monitoring CPU usage was mentioned, I did a short test.

I have an Intel E8500 Core2Duo @ 3.8 GHz and a Radeon HD 6870, relatively old hardware as you can see...

So I opened the Task Manager, started AH3 DX11 online, Mission arena, joined the ongoing mission for maximized action, followed the AI group home and while they were landing I flew low over the field and town which always drops my frame rate. Alt-tabbed several times during various stages and the CPU usage was 0% all of the time! AH3 was in the top 5 of CPU usage, though, so instead of looking at the percentage table I opened the graph tab and redid the test. It appeared that during being being minimized AH3 doesn't stress the CPU, but the graph showed that during playing the CPU usage was between 5 and 50 % on Core 1 and 5 to 30 % on Core 2.

So it seems that the CPU rarely is the bottleneck. However, this test was very short. Maybe tonight when I have time to play a little longer I'll do some further testing using the graph method. Hints for a better way to check this are welcome.

[DaddyAce]

Sorry Bizman, have no hints for you on better ways to monitor cpu usage, but I did learn something from your post.  I've been using the Task Manager graph to monitor cpu usage overall, but did not realize you could look at each core separately.  I've since found that now for future use thanks to you mentioning it.   

[Hungry]

HWinfo may come in handy

[Pudgie]

Provided in attachments below are snippets of Resource Monitor showing the CPU core usage and affinity between AHIII Patch 10 using Dx9 vs Dx11..........

Note how the game when using Dx9 is working my CPU much more than when using Dx11, both per core and total usage %.
The game has tested w\ the OS to follow the CPU core affinity consistently as shown for being run under Dx9 vs Dx11. Under Dx9 the OS assigns CPU Core0 and Core1 to the game, under Dx11 the OS assigns CPU Core0 and Core3.

GPU usage\load% numbers pretty much follow this same pattern as the CPU when running AHIII under Dx9 vs Dx11.....GPU is loaded much more heavily under Dx9 vs Dx11.

I get the occasional screen freeze when running under Dx11 but no screen freezes when running under Dx9.

 

[DaddyAce]

My puter engineer son recommended MSI Afterburner for hardware monitoring (CPU and GPU):   https://www.msi.com/page/afterburner

Afterburner is a free overclocking tool, but also provides great detailed graphing options on all kinds of CPU & GPU processes, temps, usage, frame rates, etc, and seems like a great tool to identify whether GPU or CPU is more limiting.  I just ran it and flew offline, a lot of high speed close to the ground, graphics setting full on, and was getting frame rates dipping into 50's, sometimes a bit lower.  When I looked at the graphs I could see that my GPU (3 GB GTX 1060) was flat at 100% when FR's dropped.  My i3 6100 processor did not seem to be quite as taxed, although spiking more when FR's dropped......first impressions, and I may do some more testing but the i3 6100/ GTX 1060 seem like a pretty balanced combo with 1060 maxing out first.... so if I decide I want more performance in the future would likely upgrade GPU first.

Heads up on Afterburner, you will need a zip file extraction software, I used 7-zip.

 

[Pudgie]

My puter engineer son recommended MSI Afterburner for hardware monitoring (CPU and GPU):   https://www.msi.com/page/afterburner

Afterburner is a free overclocking tool, but also provides great detailed graphing options on all kinds of CPU & GPU processes, temps, usage, frame rates, etc, and seems like a great tool to identify whether GPU or CPU is more limiting.  I just ran it and flew offline, a lot of high speed close to the ground, graphics setting full on, and was getting frame rates dipping into 50's, sometimes a bit lower.  When I looked at the graphs I could see that my GPU (3 GB GTX 1060) was flat at 100% when FR's dropped.  My i3 6100 processor did not seem to be quite as taxed, although spiking more when FR's dropped......first impressions, and I may do some more testing but the i3 6100/ GTX 1060 seem like a pretty balanced combo with 1060 maxing out first.... so if I decide I want more performance in the future would likely upgrade GPU first.

Heads up on Afterburner, you will need a zip file extraction software, I used 7-zip.

 

Hi DaddyAce,

Thanks for posting this as I use Afterburner exclusively but I had completely spaced out that AB can also graph individual CPU core usage % as well.

Got em enabled now. Gonna make some game runs to check on this..............

 

[DaddyAce]

You're welcome Pudgie,

I enjoyed building this puter, and am enjoying learning more about it.  I am happy if I can share something in this that helps you and others.  In that regard I'd like to figure out how to post a plot of my AB output.  The AB software doesn't seem to provide for printing....any ideas on how I can grab a shot of my graphs to share?

Thanks!

 

[Hungry]

If you can, screen print it, crop to where you want it, paste it into windows paint then save it as a picture

[Pudgie]

You're welcome Pudgie,

I enjoyed building this puter, and am enjoying learning more about it.  I feel the whole process is more worthwhile to me if any part of if it helps you and others.  In that regard I'd like to figure out how to post a plot of my AB output.  The AB software doesn't seem to provide for printing....any ideas on how I can grab a shot of my graphs to share?

Thanks!

 

Hi DaddyAce,

I use the built in Snippet program in Win 7 to "snip" the AB graphs as follows:

Provided are 4 attachments of snippets of the CPU individual core usage along w\ the GPU clocks\usage of my box, 2 as recorded in Resource Monitor and 2 as recorded in AB of playing AHIII Patch 10 under Dx9 vs Dx11.

Again please note how much more the CPU is utilized when running under Dx9 vs Dx11..............

These graphs indicate to me that under Dx11 API the game coding is exerting much less work across the CPU cores so the CPU is encountering a lot more CPU wait time to process game instructions while the game is running than when the very same game coding is running under Dx9 for as you can see the CPU cores are utilized MUCH, MUCH more especially Core1 and Core2 but also Core5 and Core6. Since I'm using an Intel I7 5820K CPU w\ HT disabled the OS has 6 physical cores to spread the game instructions across.

Somehow IMHO there is something within something between the AHIII game client and the OS Dx API's that is instructing the OS to assign the game threads along the CPU core affinities as shown and to cause the CPU usage to drop off relative to the Dx API being used w\ the exact same game client using the exact same settings\setup but seeing issues crop up mostly under Dx11 API vs Dx9 API. The CPU usage dropoff w\ increased CPU core wait time under Dx11 IMM has to be a contributing factor in these issues as I can see a screen pause\freeze occur if a CPU core(s) is slow in starting back up to finish an instruction thread and slow to send the interrupt to let the graphics card know there is work in the cache and to flip finished frames in sequence to display from being idle for too long a time from the OS moving game threads across this many idle cores vs what I see being utilized under Dx9.................

But to date I haven't found anything yet to expose this as the smoking gun for the issues being seen...............

Anyway, hope this helps you out.

 

[DaddyAce]

Wow that's pretty striking Pudgie!  I'm going to have to try that.  Apparently my Win 10 has a snippet tool too, will have to try that as well.  Tried the shift Prnt scrn, no joy, with a quick try earlier.  Thanks for the help guys will try to get something out on my system to see, among other things if I see a diff between DX 9 & 11 on my cpu usage.

[Pudgie]

Hi DaddyAce,

I've been seeing this for quite some time on my box but I was using the Windows Resource Monitor to record the CPU core usage during playing of AHIII then snipping the graphs as soon as I could to save them.

I had completely spaced off that MSI Afterburner was capable of monitoring individual CPU core usage and I can also set the graph width as wide as I want to get a good recording sequence of the graph lines.

I had disabled hyperthreading on my CPU to see if this would help and had found out that the OS had parked all 6 of the logical cores anyway and from reading up on this found that most games won't benefit from hyperthreading anyway if a CPU has more than 2-4 physical cores on die as most games won't need them and I don't do any content creation\video streaming type of activity on my box that could use the extra logical CPU cores as well as hyperthreading was invented to help a single core CPU to process more than 1 thread at a time to increase single core CPU performance. After multi-core CPU's have become the norm for consumers hyperthreading has much less practical use in application and could in some cases cause less CPU performance instead of more.

From doing a LOT of reading\studying on the subject (also conferring w\ others as well) of the way a Windows OS works to assign app\game threads to CPU cores for execution, what you see on the graph of AHIII running under Dx11 is pretty typical across a multi-core CPU when the app\game isn't coded to give CPU core affinity instructions to the OS to assign it's threads to but what is seen on the graph of AHIII running under Dx9 has the appearance of the particular CPU core affinity pattern being designated instead of being randomly assigned.....just what is instructing the OS to designate the particular CPU core usage pattern to AHIII game threads under Dx9 is what I can't isolate so far. I've done MANY game runs using both Dx9 and Dx11 and have only witnessed the OS to deviate from the CPU core usage pattern running AHIII 1 time using Dx11 but never using Dx9. Usually CPU core affinity is written into the app\game software to instruct the OS to use certain CPU cores on a multi-core CPU (the preferred method by MS) to process the app\game threads due to the developer's goals\wishes but this adds more complexity to the app\game code that most developers don't want to get into so this is usually left to the OS to do so the OS will always start w\ Core1 (usually identified by the OS as Core0) and will assign a new thread to the next idle CPU core in sequence if the prior CPU core hasn't completed executing the thread it was assigned at the time. It is this CPU core affinity assignment that I believe is what is giving AHIII the stability when it is ran under Dx9 vs Dx11 as the chosen CPU cores are far more saturated w\ thread execution thus are not near as idle thus do a better job of keeping the GPU saturated w\ rendering work and in synch flipping finished graphics frames to display thus no screen freeze\pause seen.

Taking all this w\ the fact that I've never witnessed a screen freeze\pause on my box running AHIII under Dx9 vs Dx11 is lending a lot of credence to what I've posted above but since I can't verify it to be more than coincidence (or theory if you prefer) the school is still out on this being the culprit that is causing the screen freezes\pauses that I've seen on my box. I seriously do not believe the issue is solely the AHIII game client software in\of itself or the HTC servers themselves...........could just be a function of further developing the game to be so efficient using post-processing graphics rendering that has offloaded the GPU loads under Dx11 to the point that they're becoming more idle due to moving data faster than the CPU can get it to them thus causing what we're seeing.....Dx11 IS much more efficient in post-process graphics rendering due to Shader Model 5.0 vs Dx9 w\ Shader Model 3.0.

I have isolated in times past this screen freezing\pausing to be caused by a more modern and powerful GPU simply outpacing the CPU\mobo mem\subsystem of a particular computer configuration's ability to keep the more modern GPU fed fast enough to keep the GPU busy and stay in synch so it would "freeze" until the rest of the system caught up (the main reason why I went to using Intel's X series HEDT platforms). But if something is causing the CPU to slow down arbitrarily the same screen freezing\pausing can occur if the CPU is slowed down enough so that it can't keep pace w\ the GPU...............and from the graphs it doesn't appear to be that the CPU is overloaded by threads\workload under Dx11......................... .

This is my thoughts on this issue of screen freezing and what got me to start monitoring GPU\CPU usage................

Well I've wasted enough time typing.......time for flying now.