The TWC Total War Benchmark thread - compare your real TW performance

[Biggus Splenus]

hi there, benchmark submitted via mail..hope everything is ok! thanks

Nice to see you've finished your build

[DramaBelli]

thanks mate, the last effort was your suggestion about the gpu. I've paid all my attention to the gigabyte one pursuing it. I'm curious to see what you think of my bench, hoping they could be useful for your analysis. It's a shame that shogun 2 seems bugged with ultra settings (that this gpu can handle). If Rome II will be concretely well optimized and stable with not heavy bugs, despite of unit sizes, visual approach to unit cards, and other moddable not incisive elements, we really are going to gain a good tribut for all this passionate fatigue.

http://www.youtube.com/watch?v=12HXnYNA6Cg enjoy

[osros]

This would make a nice Sticky on the forum, I would be going over this after I get the game playing.

[Mimoxx]

I usuallly dont zoom in the combat melee , but i play zoom out but with plenty 20 x 20 units fighting , so how much would i probably get with a cPu 42 fps + gpu=84 fps= ??? and cpu 42fps + gpu 42 fps =???????

[SamueleD]

I usuallly dont zoom in the combat melee , but i play zoom out but with plenty 20 x 20 units fighting , so how much would i probably get with a cPu 42 fps + gpu=84 fps= ??? and cpu 42fps + gpu 42 fps =???????

I can't give you numbers, but here is where you want the 84 fpd gpu. Outside of combat and zoomed out your CPU is not slowing things down, so we can consider it as giving you infinite fps, zoomed out and outside of combat the GPU decides your speed, so a 84fps gpu gives you 84 fps, but a 42fps gpu will give you only 42fps.
Let's now hypothetically assume that not zooming in during combat grants you, say 10 fps of CPU power, so your CPU can deliver 52 fps in combat while zoomed out, if your GPU can deliver 84fps then you'll be running at the 52fps determined of the CPU; if instead you have a 42fps GPU, you will be running at 42fps, even if the CPU can deliver 52 in that situation.

The CPU benchmark monitors a thick melee fight, which is one of the worst situations for your CPU in TW, so basically that number is the minimum fps your CPU will get you, but if the fighting is less intense or you are less zoomed in, the CPU will deliver more fps, and in that case, if your GPU is strong, it will be able to give you higher fps, if it's weak it won't.

Also, a faster GPU will allow you to get increased graphics quality, while a faster CPU will allow you to get more soldiers on screen.

[Mimoxx]

Many Thanks Samuel I understand at my view : the cpu is lot more importante than gpu !

[SamueleD]

Many Thanks Samuel I understand at my view : the cpu is lot more importante than gpu !

Not really: both are important. A crappy GPU will bottleneck a good CPU. Usually, having a really good CPU and a decent GPU will get you better minumum fps than an awesome GPU and an average CPU, but if you want to experience good quality and the full beauty of the game without lag, you will need a good GPU (even if you might not need the hyper high end ones or multi gpu setups), not just a good CPU.
So yeah, if you just aim at running the game with ultra unit sizes and everything else turned down then, you are better off picking an i5 and an entry level GPU, but, since we presume that you might want a compromise of both, the builds in the HW advice thread tend not to make that kind of paring, and be more balanced.

[Biggus Splenus]

^
Described above is called "bottlenecking", where your PC will always be limited by the lowest figure ie. the weakest link.

[SamueleD]

^
Described above is called "bottlenecking", where your PC will always be limited by the lowest figure ie. the weakest link.

You mean "in the previous page"?

But yeah, that's the name folks! In TW the CPU will almost always be your bottleneck (unless your GPU is really crap), and it will bottleneck the most in zoomed in melee, but once you zoom out or you are out of combat the CPU bottleneck disappears, as the CPU is no longer stressed, so you get min fps determined by the GPU.

[alQamar]

I appreciate your contribution but please: thread rules

[Biggus Splenus]

alQ, for your benchmarks, did you install Shogun 2 on an SSD?

[alQamar]

Yes but it is running on a seperate SSD. But not in other cases. It does not influence your fps in the tests. Tried it with a WD green aswell. Just loading times improved.

[alQamar]

added new results for Intel i7-3770K and AMD X4 465 and a EVGA GTX 670 FTW (MLAA) mode

[alQamar]

no we have not. As I said it is a mod then (blood pack DLC is a mod) but that do not explain his also poor performance in SuperPI.

[baptistus]

I will run the benchmarks tomorrow with my i7 3930k and HD 7970.

[A Barbarian]

Ok, first CPU result from me:
47.7FPS: DX9 CPU@1920x1080, no blood effects, 16 threads, i7-4770K@4.3Ghz, uncore 3900Mhz, HT on, 1866Mhz CL9 memory, CF disabled
50.2FPS: DX9 CPU@1280x1024, no blood effects, 16 threads, i7-4770K@4.3Ghz, uncore 3900Mhz, HT on, 1866Mhz CL9 memory, CF disabled

I also noticed that the recommended '16 thread' setting is quite effective (about 10% plus) with blood effects disabled, but with blood effects enabled, using 0 (=default) yields actually somewhat better performance.

GPU benchmark: High benchmarks from this thread + 16 thread settings + blood effects disabled (probably doesn't matter since there is no battle in the benchmark)
83,1FPS: 1x7970@1060core@1500mem , CPU like before, 13.8 beta drivers
144,6FPS: 2x7970@1060core@1500mem , CPU like before, 13.8 beta drivers, framepacing on

[Biggus Splenus]

@ Sam - thanks I did do that, but I wasn't sure if it worked.

@ - barbarian - yep, I've specifically set the uncore and core speed together, and I knew how to check that with cpu-z Also, thanks for figuring out my problem for me, I can't believe blood causes so much more stress! Ill try it out once I manage to get out of bed

EDIT: After a couple hours of overclocking and benching, I got these results:

1 - stock settings; 34.3 fps
2 - 4.4Ghz core overclock; 38.6 fps
3 - 4.4Ghz core and 4.4Ghz uncore overclock; 39.2
4 - 4.4Ghz core and 4.4Ghz uncore on 129.5 CPU BUS; 40.2

So, as you can see, in my setup there is not much improvement by increasing the uncore memory. Could be different for others though. There was an improvement by leaving the multipliers for the core and uncore on default, and just increasing the BCLK to the desired speeds, though.

The only thing I can think of to increase my performance is to buy some beastly RAM... I might look for a good 2x2GB kit at 2400Mhz later on, I'll see how much RAM Rome 2 uses first.

[Stixz Awesome]

Is there a possibility that 4670k and 4770k owners could do benchmarks on stock clock speeds, and agains on 4.2ghz? Me and a friend are speculating on which cpu would be the best choice for us considering the only form of OC we will be doing is using OC genie to get them to 4.2ghz.

Additionally. Could someone figure out why the results on this site varies so much from the results in this thread? They´re using fraps as an fps counter, but that shouldn´t be it?
http://www.bit-tech.net/hardware/201...l-cpu-review/5

[alQamar]

Thank you all for your contribution. And currently most of all barbarian to point out the blood dlc issues. I can literally recall the preferences.script.txt file in sleep but I never thought about the impact in fps. I will contact all participants to verify that and rerun benchmarks if needed. Before any pm or email.


[B [All those who participated yet who owns the blood dlc? Please pm me.[/b]


Splenyi the blood dlc may impact but don't explain your performance issues also confirmed in super pi.

Mate if you need to waste money for ram be sure not to get more than cl11. Btw next year ddr4 I won't do that. You have a basic performance issue. Please replace your mobo as talked. The c2 stepping seems to easy oc haswell aswell.

Btw you perfectly know how much ram rome 2 uses not more than 2 GB at all for the application as they at least need 1-2 GB VRAM and then exceed the 32 bit border of 3.6 GB

[alQamar]

▲▲▲Please thread rules! I will upload new preferences files with blood disabled that's the safest way for comparable results.