Alleviate CPU Bottlenecking?

[Live2sculpt]

I'm mostly focused on NTW at the moment, but these same issues plague me in my Shogun game as well.

Whenever I'm trying to enjoy LARGE battles, I encounter the notorious stutters in my unit animations. My PC is no great shakes (Intel Duo 3.8MHz/8800GTX/6GB), but when my FPS clocks in at 35-40 during a 40v40+ battle, it gets rather frustrating to watch my units still moving as though they were going at 12-16.

So, assuming that this unit animation is getting throttled at the CPU level, it then becomes important to deduce what I may do to free up CPU real-estate.

I already know that SHADOWS are drawn using the CPU, and not in the GPU. What else might be handled in there?

And if anyone has come across any other tips or clues regarding the mysterious stutter effect, please chime in.

I've never seen a conclusive assessment in all of the internet, as of yet. Some have said that particle effects setting makes a difference, others have pointed to their sound drivers as the culprit.

But I'M getting convinced it's about the CPU (and poor muti-threading optimization).

[ChewieMuse]

You can always try un-parking the cores For your cpu.. I think I got maybe a 10 fps boost doing that with my 8150

[Live2sculpt]

You can always try un-parking the cores For your cpu.. I think I got maybe a 10 fps boost doing that with my 8150

The Zambezi? LOL
Thanks, but you're talking about drawing in more resources from an 8 core cpu to run a program originally designed for just 2 cores, which is what I've got. Don't worry, I'm monitoring those two cores, and they are both pulling their weight. No parking here.

Glad you have a nice cpu. Eight cores sounds sweet.

[Live2sculpt]

self-removed

[Live2sculpt]

We REALLY don't have anyone left here who understands the guts of this games software better than myself? That's sad.

I have a follow-up question on whether upgrading my GPU could somehow improve this problem at all.

Upgrading an old 8800 GTX with a GTX 560 ti is an obvious improvement, but I need educated comments on whether that upgrade could enable more troops on a battlefield without this stuttering. If I thought it would, I'd spend the $150. But I'm getting 29-45 FPS on high settings with the 8800 and have no desire for more if it doesn't take that load off the CPU in a meaningful way.

Past that, the clearest path would be to replace the mobo and cpu, and gtfo of the old 775 socket restrictions. And I'm not spending that money just yet.

Is there ANYONE with enough experience who can tell me what they know about this??

Someone who's answer doesn't involve the ubiquitous 'advice' to just spend money.

[A Barbarian]

A new GPU probably won't help you with that problem. S2 is very much CPU limited in many situations.
Regarding the stuttering, there is another possibility. Is your operating system 32 bit or 64 bit ?

[Live2sculpt]

A new GPU probably won't help you with that problem. S2 is very much CPU limited in many situations.
Regarding the stuttering, there is another possibility. Is your operating system 32 bit or 64 bit ?

It is a 64 bit Win 7. So that old tip on how to clear this up on 32 bit isn't applicable for me.

[Live2sculpt]

Would it really be worth it to switch down to 32 bit OS instead of 64?

I'm not overly attached to Win 7 64 bit, but expect I'd loose more than I'd gain in making the switch.

[Lord Uxbridge]

It may be best for you to ask this in the basement because more people who may be able to help you will probably see it, rather than those who are looking through the S2 Technical Support area.

[Live2sculpt]

It may be best for you to ask this in the basement because more people who may be able to help you will probably see it, rather than those who are looking through the S2 Technical Support area.

Thanks. I made a thread there.

But it seems like the basement is more for the kids who are excited about rig-building and not too interested in maximizing performance, unless it involves rig-building.

As my query is specific to how this software and the drivers and the os work to render for THIS specific title, I think it's well to talk about it here.

[A Barbarian]

64 bit is fine. 32 bit could a possible problem.

Hmmm, maybe there is a VRAM limitation. Can you monitor the VRAM usage with something like GPU-Z or MSI Afterburner and check whether it's over the graphics card memory. Your card probably has 512MB VRAM.

[Live2sculpt]

64 bit is fine. 32 bit could a possible problem.

Hmmm, maybe there is a VRAM limitation. Can you monitor the VRAM usage with something like GPU-Z or MSI Afterburner and check whether it's over the graphics card memory. Your card probably has 512MB VRAM.

Close. The 8800 GTX has 768 MB GDDR3 RAM. I have Afterburner installed, it's providing a stable oc and maxing the fan to keep it under 70 degrees. I also specified the vram in kb on that line in the preferences script.
I'll look when I get off work. Exactly what would I be looking for when using MSI AB to see if it's actually a VRAM issue?

[Live2sculpt]

64 bit is fine. 32 bit could a possible problem.

Hmmm, maybe there is a VRAM limitation. Can you monitor the VRAM usage with something like GPU-Z or MSI Afterburner and check whether it's over the graphics card memory. Your card probably has 512MB VRAM.

Well, live graphs during play seem to demonstrate that both of my cores are sharing the load equally, and that they are indeed hitting their ceilings when the battles exceed 15,000. Funny thing about the mem usage graph from MSI; it seemed to show a fixed value, either 521, 483, or 484, but never reading at it's designated max of 768, and never fluctuating to use. I wonder what gives? Is that normal?

[A Barbarian]

It's normal that VRAM values don't fluctuate much. With a game like Shogun 2, there is usually one amount of VRAM associated with one scenario (e.g. campaign map, land battle, sea battle). The memory usage might go up during one scenario as more textures have to be loaded for objects which come into sight first. Shogun 2 uses the most VRAM in siege maps with big castles, Kyoto being the largest.

That being said, VRAM also doesn't seem to be a problem for you since with your detail level it doesn't even reach the 768MB. Together with your observed CPU utilization the conclusion is that your gaming experience is mostly CPU limited just like you already suspected in the beginning.

And as you also already know, lowering/switching off shadows can help and also switching off blood effects. You could also try to test the lowest settings and see whether the animation skipping is still present. If yes, probably only a new CPU will help (or overclocking if it's possible). Which CPU and motherboard models do you have exactly?

Another interesting test would be the built-in DX9 CPU benchmark which can be started from the startup menu of S2 in Steam.

[Live2sculpt]

It's normal that VRAM values don't fluctuate much. With a game like Shogun 2, there is usually one amount of VRAM associated with one scenario (e.g. campaign map, land battle, sea battle). The memory usage might go up during one scenario as more textures have to be loaded for objects which come into sight first. Shogun 2 uses the most VRAM in siege maps with big castles, Kyoto being the largest.

That being said, VRAM also doesn't seem to be a problem for you since with your detail level it doesn't even reach the 768MB. Together with your observed CPU utilization the conclusion is that your gaming experience is mostly CPU limited just like you already suspected in the beginning.

And as you also already know, lowering/switching off shadows can help and also switching off blood effects. You could also try to test the lowest settings and see whether the animation skipping is still present. If yes, probably only a new CPU will help (or overclocking if it's possible). Which CPU and motherboard models do you have exactly?

Thanks.

I'm making the best I can with an old 775 socket Gigabyte board and a 8500 Duo 3.1 @ 3.8. I don't think this old equipment merits any cooling beyond good flowing air so I'll stick with my copper Zalman and 70 c on load and not try for 4.1 or 4.3, which this chip is well capable of if better cooled.

I went on the assumption that a decent dual core is primarily all I need, as these TW titles are notorious for not utilizing any more than two anyhow. But now I'm learning how there could be impact-full differences in the architecture of an older Wolfdale like the 8500, and some of the newer gen dual cores (even ones with much lower Mhz ratings). But that probably involves switching to a new motherboard, as the 775 socket wouldn't allow me an upgrade to those new chips. There's plenty I don't know, but I have done a little bit of research.

I have tested performance with various lower settings. That's what prompts this thread as unique from standard performance queries. This behavior largely persists SEPARATE from the graphics settings or the FPS scores. It's primarily all about troop and unit numbers.

As for your comment re my VRAM, you may be right, but I AM having to keep my settings at HIGH only to maintain just 25-31 FPS during thick engagements. With smoke involved it can quickly tank to 18 FPS. Shouldn't my v-card be pulling out all the stops to prevent that from happening? Shouldn't my 768MB be used entirely before I'm to experience such shortcomings? I wonder if there is a manual setting somewhere that may force that 100% v-ram usage, and what effect it may have.

[Live2sculpt]

A follow-up for others information:

For my particular system, I have been able to run 80 v 80 battles in campaign mode (2 armies) and have achieved a modest but playable 32FPS and an 'ANIM' FPS of 20-25. I run settings on High with Shadows, Anti-aliasing, and Ambient Occlusion turned off. The unit sizes are X3.

Though others systems may differ, the POINT of my report is that this performance indicates to me that there is an additional task being set upon the CPU when running the same number of agents on the field, but divided among more than two armies. I have run custom 4 v 4 battles with these same settings and unit sizes and those custom battles perform noticeably worse.

For now I'm going to relax and enjoy my full DEMI-CORP sized battles on my puny little scratch-built 775 PC that I made out of junk I found in a closet.