Civ5 performance analysis

[Becephalus]

Maybe most Civ players don't play graphics-heavy games, but compared to what other games can do today, Civ 5's graphics are a joke when you factor in the specifications they need. A game with such graphics should be playing top speed and full settings in 5-7 years old hardware.

BUt most graphics heavy games aren't also overloading the CPU. Look at Paradox Entertainment games...

Graphcis are 8 years old, CPU loads are 2 years into future...

 

[aziantuntija]

http://www.gamespot.com/features/6279244/index.html?tag=topslot;title;5

Why these guys are comparing old Q6600 2.66GHz against newer E8200 2.66GHz? E8200 should go against Core 2 Quad Q9450 2.66GHz.

Core2 Quad Q9650 @ 2.93GHz should be competing against Intel Core 2 Duo E8400 @ 3.00GHz.


Core 17-980 @ 3.33 vs Core i3-550 @ 3.20GHz = Hexa core wins by 11%

Phenom II X4 965 @ 3.4GHz vs Phenom II X2 545 @3.0GHz = Quad core wins by 13%

2+Cores wins this test with average of 12% better performance against Dual core.

Other CPUs in this link are uncompearable against each other, but still even old Core2 Quad Q9650 @ 2.93GHz manages to beat Phenom II X2 545 @3.0GHz and Core2 Quad Q6600 @ 2.66GHz manages to beat Core2 Duo E8200 @ 2.66Ghz

 

[tokala]

Core 17-980 @ 3.33 vs Core i3-550 @ 3.20GHz = Hexa core wins by 11%

Phenom II X4 965 @ 3.4GHz vs Phenom II X2 545 @3.0GHz = Quad core wins by 13%

2+Cores wins this test with average of 12% better performance against Dual core.

It seems you conveniently overlooked that the X4 has also 13% more clockspeed than the X2

And the 980 has 12MB vs 4MB L3 cache next to its 5% speed advantage over the 550

So, all else beeing equal, a quad will be EXACTLY as fast as a dual core within the margin of error

 

[MarkJohnson]

My i5 655k dual core outperforms my i7 920 by quite a bit. My turn time on a huge map late game was reduced from 4minutes to around 3m 45s. I now overclocked my 655k from 3.2GHz to 4.4GHz and turn times are under 3minutes. but each turn, the time slowly goes up.

btw, i get zero fps between turns and video cards goes idle. there seems to be some coding issue(s) as windows reports "not responding" and the graphics freeze up. Once they cure this, maybe turn times will improve greatly late game and probably increase fps at the same time.

 

[aziantuntija]

It seems you conveniently overlooked that the X4 has also 13% more clockspeed than the X2

So you are saying that with AMD CPUs +1MHz of clock speed means 1% less waiting between turns and number of cores doesnt have absolutely anything to do with it?

And the 980 has 12MB vs 4MB L3 cache next to its 5% speed advantage over the 550

..with the Hex core being faster than Dual core, so what are you trying to proove?

So, all else beeing equal, a quad will be EXACTLY as fast as a dual core within the margin of error

Yeah right

 

[aziantuntija]

So, all else beeing equal, a quad will be EXACTLY as fast as a dual core within the margin of error

Ok here we go:

http://www.anandtech.com/bench/Product/52?vs=58

For example, here is Intel Core 2 Quad Q8200 2.33GHz 4MB L2 going against Intel Core 2 Duo E8200 2.66GHz 6MB L2

Note that the Quad core has LESS MHz AND LESS L2 but it still beats the Dual core in 22 out of 31 tests .

EDIT: http://www.anandtech.com/bench/Product/76?vs=58 Oh and here we have a equal comparison between Intel Core 2 Quad Q9400 2.66GHz 6MB L2 going against Intel Core 2 Duo E8200 2.66GHz 6MB L2

Well it doesnt look too good for the equal MHz/L2 dual core . The quad beats it in 27 out of 31 tests.

 

[tokala]

Never contested that a quad will be faster with the appropriate applications, but what's the relevance of all that for the question how they perform in Civ5

 

[aziantuntija]

So, all else beeing equal, a quad will be EXACTLY as fast as a dual core within the margin of error

Never contested that a quad will be faster with the appropriate applications

lol

..well quad core is better in most of todays application as you can see from my links.

..but what's the relevance of all that for the question how they perform in Civ5

Well you claimed that dual cores and quad cores are equal when GHz and L2 are equal and i proved you wrong. Whats it got to do with civ5? Well your claim wasnt about civ5 now was it, so why are you asking this?


EDIT: If you are only using your computer just for civ5, then you should propably only look for civ5 CPU benchmarks when concidering what CPU you are going to buy next. In this case all other tests are irrelevant.

 

[aziantuntija]

..well quad core is better in most of todays application as you can see from my links.

And ill give you some more proof of that.

http://www.anandtech.com/bench/Product/49?vs=145

Here is much older and clearly slower clocked Intel Core 2 Quad Q9650 3.00GHz 12MB L2 beating Intel Core i5 670 - 3.46GHz - 512KB L2 4MB L3 in 17 out of 31 tests wich makes it the winner of that comparison.

 

[tokala]

Well you claimed that dual cores and quad cores are equal when GHz and L2 are equal and i proved you wrong. Whats it got to do with civ5? Well your claim wasnt about civ5 now was it, so why are you asking this?

You have paid attention to the thread title, haven't you
The purpose of this thread is the discussion of hardware performance in Civ5, NOT in general, and accordingly everything written here is supposed to refer to Civ5, even if not explicitly stated

Edit: And I would kindly ask you not to clutter up the thread with information irrrelevant to its stated porpose.

 

[aziantuntija]

So, all else beeing equal, a quad will be EXACTLY as fast as a dual core within the margin of error

Well you should propably also stop writing this kind of garbage to this thread.

I just toke my space to prove you wrong and thats why you propably got so upset.



EDIT: OMG I just realized that you are the same guy who claimed (in this thread:http://forums.civfanatics.com/showthread.php?t=379787&page=2) that civ5 can be maxed out on 1680x1050 High settings with Full Tessellation and 8xAA icluded with just a HD 5670 and that it would be 'perfectly playable' .


http://www.techspot.com/review/320-civilization-v-performance/page5.html

Well i wouldnt concider 19 fps on average a playable frame rate .


BTW, do you remember this:

EDIT: Im willing to bet that HD 5670 is not capable of producing steadily 30 fps or over framerate in civ5 with all settings maxed out.

 

[bossconian]

And ill give you some more proof of that.

http://www.anandtech.com/bench/Product/49?vs=145

Here is much older and clearly slower clocked Intel Core 2 Quad Q9650 3.00GHz 12MB L2 beating Intel Core i5 670 - 3.46GHz - 512KB L2 4MB L3 in 17 out of 31 tests wich makes it the winner of that comparison.

i5 670 has only 2 cores and it's very impressive it beats high end 4 core processor in almost half of benchmarks.

 

[aziantuntija]

i5 670 has only 2 cores and it's very impressive it beats high end 4 core processor in almost half of benchmarks.

Well it IS very impressing to see even OLD Quad core with less GHz beating a MODERN Dual core with higher GHz and Hyper Threading in over half of the benchmarks.

I just happen to have Q9450 wich is the same chip as Q9650 is but ive clocked it to 3.2 GHz so it is even faster than Q9650! It is old CPU but it still beats most (if not all) of the modern Dual cores

 

[PieceOfMind]

Moderator Action: Guys, please stop spamming up the thread. The topic is civ5 performance analysis. Keep to the topic.

 

[Gorey]

My PC handles the graphics well (as it does for any new game out).. it's just the wait between turns sometimes takes forever.

Most turns it's 3-5 seconds.. but every now and then it's like it hangs for a good 15 seconds. Really saps the fun out of it. I spend more time waiting then playing mid game on.

 

[Venger]

It seems the multi-threading in Civ5 pretty much only shows up in the interface - it seems to handle moving and clicking during the AI turns much better than Civ4 did.

 

[Akka]

I couldn't care less about graphics in any Civ game, what's interesting is how fast the AI finishes its turn.

I don't particularly care for graphics as long as they are at least "acceptable", but the amount of ressources eaten to display something extremely NOT awesome is simply staggering.

I just don't understand where all the processing power go. The graphics are not better than the ones of a game that is five years old, and they still require a computer that is much more powerful. What the hell ?

 

[routehero]

OLD cpu = retail circa 2 years ago?
MODERN cpu = retail circa today?

Thanks. What a pointless argument.

To whomever said it's difficult to do parallel programming, it's not and it shouldn't be for this type of game. Each civilization would get its own mutex with multiple child threads for their individual AIs. You interrupt the remote diplo thread for AI-to-AI diplo, but besides that, there's nothing that really requires interaction between the two.

The OS will take care of assigning the threads to the various cores on the system. I would guess that there is a single mutex and each AI uses each thread (diplo, city building, etc) sequentially. It's a bummer and could have been designed better without much developer time.

 

[VirtualMark]

What annoys me is that we have been lied to by Firaxis. I remember reading that the game would make use of up to 12 cores. I was very impressed with this, as it had seemed that they've really done a lot of work. And now we get the game and it barely makes use of my i3, which shows 25% usage!

Surely it can't be that hard to compute some of the AI's moves while we're making turns. Thats how chessmaster does it, its constantly thinking. And as Civ is turn based, theres only a certain amount of moves our units can make, so why couldn't the AI compute answers for every possible move we make, so that by the time our turn is finished most of the AI calculation is done and turn times are much quicker.

Its 2010, and intel are pushing more and more cores. quad core is the norm now, with 8 cores and 16 threads in the high end chips. And most new systems have 4gb+ of ram. Yet civ v is 32bit! Am i the only one who thinks it would be nice to have a 64bit civ, where the maps can be huge and use as much ram as they need?

Perhaps civ 6 will be the killer game i was waiting for. Oh well, at least i'll get more work done in the next few years.

 

[PieceOfMind]

To whomever said it's difficult to do parallel programming, it's not and it shouldn't be for this type of game. Each civilization would get its own mutex with multiple child threads for their individual AIs. You interrupt the remote diplo thread for AI-to-AI diplo, but besides that, there's nothing that really requires interaction between the two.

The OS will take care of assigning the threads to the various cores on the system. I would guess that there is a single mutex and each AI uses each thread (diplo, city building, etc) sequentially. It's a bummer and could have been designed better without much developer time.

I was one of the people who said it's difficult to make do parallel programming (in context to this game anyway). Because it's a turn based game with lots of rules that need to be followed by each civ on each turn in exactly the same order, it seems it would be harder to send off various tasks that may finish at undetermined times. Yes it can be done to a limited extent, but things like what order to process the collection of gold and the completion of buildings (like banks) need to be the same every time.
A lot of decisions about unit movement can't be done in parallel if that would result in lots of units trying to move to the same square. (oops, I mean hex )

I admit I'm no expert on this, and I can absolutely agree that limited multi-threading is not difficult and from what evidence we have appears to have been done already, but optimising it to near equal use of all cores seems a very challenging task for what are essentially a small games studio with probably only a handful of people working on the code that can be optimised in this way.

routehero, you speak on this topic with probably a lot more experience than I have so I am interested in reading your views, but your brief explanation has not done much to convince me.