I'd love to see a Rome:Total War Intel vs AMD CPU comparison

[yacoub]

I'm extremely curious how an Intel chip at, say, 3.2GHz runs a given battle compared to an AMD chip equivalent of 3200MHz would run it. That is to say a 3200+ of some sort. I would almost guarantee that the chip actually running at 3.2GHz (the Intel chip) would run the game better, but I want to see that backed up with testing.

R:TW's real-time battles are a great gaming stress test for CPU comparisons since they really show which CPU can crunch through the thousands of calculations faster for a given battle.

I remember a year or two ago when the game was used for Time Commanders (TV show in the UK), they were running it on some crazy-fast machines just to make it run smoothly for the show. (Point being it's a known CPU hog of a game.)

Currently my Barton 2500+ (oc'd to 2700+ speeds) crawls in big battles. My 9800Pro sits there twiddling its virtual thumbs waiting for the CPU. I am considering upgrading my system but have yet to see R:TW used in any CPU comparison and I'm very curious to see if actual raw clock speed is more important than fancy instruction routing and caching when it comes to this game (and others that are CPU intensive).

Thoughts?

 

[yacoub]

from: http://www.anandtech.com/guides/showdoc.aspx?i=2426&p=3

There are always people who continue to prefer Intel processors for whatever reason. We feel that AMD has the upper hand right now - both in price as well as performance - but there are merits to the Intel platform. We've said it many times before, but let us reiterate that anyone interested in serious 3D gaming should just forget the Intel chips for now. It's really that cut and dry, unfortunately.

I'd like to see that proven with R:TW and other CPU-intensive games.

Yes, AMD chips are great and I currently run one (and have for my last two PCs as well), but I'm starting to wonder if the games that most appeal to me right now would run even better with an Intel upgrade over an AMD upgrade for my next system since they are CPU intensive.

 

[CTho9305]

Currently my Barton 2500+ (oc'd to 2700+ speeds) crawls in big battles. My 9800Pro sits there twiddling its virtual thumbs waiting for the CPU.

I'm just curious... how do you know that? Did you try playing at very low resolutions?

 

[yacoub]

Yes - I only even play at 1024x regularly and don't even crank the graphical options up very high and I can still bring my system to a slideshow level of play, given enough units in a battle.

It's common knowledge to those who play the game that it has the ability to bring pretty high-end systems to their knees if you load up a Custom Battle with Huge troop size and large numbers of units. In fact, during a Custom Battle setup, the game will actually warn you when you are crossing the threshold of unit numbers that will start to impact the performance of your system.

I mean it's AWESOME to be able to play massive battles, but at the same time it could be expensive to get it running smoothly. I just want to figure out the most cost effective upgrade path that will net me the best results - meaning that when I upgrade I might spend more on an Intel CPU if that means measurably better performance in R:TW, while cutting cost in another peripheral that isn't as important to it.

 

[Turkey22]

Playing Empire earth 2 multiplayer with my friend in the same room does show the difference a new proc can make. I have a 2000+AXP and he has a 3400+ 754 A64. We both have 9800 pros. His still runs into issues with so many units on the screen, but he has his settings turned up while I have mine pretty much all the way down. The difference even with his settings higher were definitely noticeable as mine crawled and his would stutter occaisionally.

Obviously not the same game, but similar situation.

 

[Regs]

Memory bandwidth is most important for Rome Total War.

 

[Snapster]

Try Football Manager 2005 with all leagues, all forground. It's basically a number crunching game.

 

[yacoub]

Regs: How/why is that?

In my experience, it's the sheer volume of units on screen that the CPU has to calculate that bogs it down. Textures aren't an issue, as all the units of each type use the same texture and the textures are relatively light weight (file size). FWIW, I have a GB (2x512MB) of pc3200 DDR400 memory in dual-channel mode so I seriously doubt that is the issue, especially since R:TW players commonly speak to the game's ability to bog even high-end processors when a full allotment of troops with Huge unit sizes is put into play on a map and the camera is scrolled around or battle is begun.

It really does come down to number crunching - calculating all the of statistics for the units in play - combat modifiers such as fatigue, morale, attack and defense bonuses, etc, the impact of their position on the field relative to other units and how that affects those modifiers, which animations are to be played depending on what they're doing, etc.

And that's why I'm curious to know if Intel with their legit MHz ratings can outperform AMD with its fancier instructions (or whatever) at a given clockspeed/AMDrating in a given scenario in Rome: Total War.

I wish Anandtech would do comparisons like that - real world tests that would help RTS/strategy gamers like us determine exactly how much processor speed we need to run a game smoothly and which processor performs better when it comes to actual gameplay in a large battle. Even stepping back a bit and proving that it is indeed CPU-limited and then stepping forward into comparison testing.

If I could read an article one day that states something simple like "If you want to play games like R:TW or -some modern RTS game name here-, you want to focus on purchasing a single-core processor with the highest actual clockspeed your budget can afford." (That's not a confirmed statement, it's simply an example of how specific and clear a statement should be.) Something simple like that would make it much easier to upgrade a system that will be used to play such games - games that are often extremely different in their taskings and stresses than your average FPS game that primarily hungers for a strong GPU.

If I knew for a fact that I should focus on CPU clockspeed, I'd take the time to research Intel solutions and lock onto a 3.2-3.6GHz chip that was best affordable for me and then build the rest of the system around that. If it turned out that AMD's rating sytem was true for RTS/strategy game performance, I'd continue to plan to build an AMD system for the better cost and greater overall performance.

 

[DaFinn]

Hmm, my dual xeon box has no problem handling large battles at 1280x1024 resolution with all eyecandy on. I have a geforce fx5900ultra 256Mb/ dual 2,4Ghz xeon@3Ghz/ 1Gb (2x512) RAM on Asus PC-DL deluxe board. I've also played Rome on my other machine which is a P4 3,06Ghz/1Gb/5900ultra, also without any slowdowns...

 

[Schadenfroh]

Originally posted by: DaFinn
Hmm, my dual xeon box has no problem handling large battles at 1280x1024 resolution with all eyecandy on. I have a geforce fx5900ultra 256Mb/ dual 2,4Ghz xeon@3Ghz/ 1Gb (2x512) RAM on Asus PC-DL deluxe board. I've also played Rome on my other machine which is a P4 3,06Ghz/1Gb/5900ultra, also without any slowdowns...

can i have some money?

 

[yacoub]

Hah, yeah me too. =P

 

[Malladine]

R: TW runs well on my AXp1700+, 9800np, 1gb pc2700 system EXCEPT when fire arrows are used. Damn those things slow the game to a craawl, i'm talking 5-10fps whenever they're used by just 1 unit of archers. I'm thinking that is definately cpu based.

For some kind of comparison, what 3dmark2005 score you all get? 2202 here.

 

[yacoub]

Wow that test really shows what a worthless piece of crap my system is today for modern games:

2490 is my score and it was an absolute slideshow the entire time.

3D Mark Score: 2490
GT1: 10.8fps
GT2: 7.1fps
GT3: 12.9fps

CPU Score 2966
Test 1: 1.3fps
Test 2: 2.9fps

Now I REALLY wanna upgrade but dunno what to get.

Intel? AMD? Regular ATX? Shuttle SFF? Yarrr...

 

[gobucks]

i've never had an issue with rome: total war's performance on my Athlon 64. Of course, it is at 2.55GHz (practically an FX-55, lol), and is accompanied by an overclocked 6600GT. I do get the occasional stutter (maybe 1 time an hour), but i think that's a bug in Rome, because I've run across it in many people's computers.

Anyways, Athlon XP gaming performance is hardly an indicator of Athlon 64 performance. The later models of the Athlon XP clearly were outclassed by the Northwood Pentium 4, since the 512KB cache of northwood and its rapid ratcheting of clockspeed made it just impossible to keep up. However, the A64 is a much more sophisticated chip. Its memory bandwidth is unsurpassed - it consistently outscores the P4 in this area, because 1) AMD's memory controller is integrated 2) AMD is not stuck on a FSB, whereas intel is and 3) Intel's FSB is not keeping up with its memory, so even though it supports dual channel DDR533, if it only has an 800MHz FSB, you're not gonna be able to use all that memory bandwidth. Having rapid memory access is very important for games, so this is a big advantage for AMD. Also, since game software is not currently multithreaded, and multithreading is Intel's forte, I don't see how a P4 would perform better in Rome. Throw in the fact that AMD has a big time performance advantage in the actaul graphical area of gaming, and Intel would have to absolutely trash AMD in the CPU calculations area in order to put up better performance in the game overall, and I just don't see that happening.

 

[Concillian]

Originally posted by: yacoub
I would almost guarantee that the chip actually running at 3.2GHz (the Intel chip) would run the game better.

Why?

 

[yacoub]

Gobucks: Nice!

After reading your system's specs, I'm like, "Okay I enjoy researching and buying the hardware and I can put together a system but I wish I had someone who would mess around with o/c'ing it for me so I don't have to struggle through that myself." =P

I always hated when you build a new system and try to o/c it and it doesn't like the settings you're trying and it won't display anything on the monitor until you find the BIOS reset jumper and use that to reset it to defaults. And so on. Over and over until you get a good o/c out of it. I guess it'd be more fun if I wasn't so impatient these days.

I wish there was a geeky high school kid in my neighborhood who knows the ins and outs of o/c'ing an NForce4/Athlon64 system and would be willing to mess around with a new system to get the most performance out of it.

Jsut watch him zoom through the BIOS flipping settings and rebooting and within like an hour, BAM, safest stable overclock.

That would rock!


edit: Concillian - because I don't have a clue? lol. I am just under the impression that the chip that runs at the highest clock speed is able to process the most requests in a given clock cycle and thus can handle a higher load of requests.