Hardware How many CPU cores do games need?

so basically all you need is a triple core if at least your a gamer, the rest lays in the hands of the graphics cards.

 

The games featured aren't really all that CPU heavy anyway. I think you would have seen a big difference on GTA4 and Far Cry 2 - they both run a quad quite hard.

Using the COD engine twice was a bit of a poor show. My Breville sandwich maker can run that.

 

If after all this time of having multi core cpu's we are still not seeing much improvement over and above a dual core for gaming it make me wonder how far we can go with parallelising this sort of app. Obviously data parallel apps like video encoding are a different story.
Lets not forget that multi core is just the hardware manufacturers selling feature they used after the death of the GHz war, it may work for them for increased sales but there will be a wall for parallelism just as there was for the GHz war.

 

Great article but...

I felt it could have examined more than just the minimum CPU required to play games without too much extra effort. We saw what is potentially a GPU bottleneck but provided for the minimum number of cores, however it would have also been nice to see the CPU bottleneck end of things by at least turning down the graphics, just to see how far the number of cores takes each game. The would have shown what the different game engines are capable of, and further allowed us to predict performance on different systems.

For that reason, I would have also liked to have seen a Source game - just so I know how it performs as an engine. Many, many people play Source games, and while it might have been around a while, it still gets updated and used. And will be further updated and used. If I remember correctly, it used to make heavy use of only one core (don't quote me on that, I can't be sure), but it definately uses more than one very heavily now.

I would have also liked to have seen a supcom test. I still play it regurlarly. Hell, I was playing it last night. And it was the first game I noticed that took both my cores for a ride at the same time. I also would like to know if it can scale to 6 cores. Certainly, if I open up a skirmish with 7 CPU players and maximum unit counts on the largest map possible (using SC:Forged Alliance), then one 'second' on the game clock takes about two seconds of real time. And this long before the players reach their unit caps.

Anyway, great article, I just wish it could have gone that little bit further.

 

This is a good demonstration of how selectively choosing what to test and what to ignore let's you manufacture whatever conclusions you want.

BFBC2 for example, runs fine on a dual core in single player, but as soon as you go online, your frames get cut in half on a dual core because there are so many players are so much going on at once.

BFBC2 needs a tri or quad core, no question, because frankly, the single player is crap and nobody is playing it. The multiplayer is where 90% of players will be spending their time.

The same is true for most of the games tested here, which run much smoother in single player because everything is scripted, confined and heavily playtested, whereas multiplayer requires a lot more grunt because none of the computation required can be done off-line (i.e. not on the fly, as opposed to single player when things can be optimised ahead of time).

 

Any chance you could add an infamously CPU bound game to the charts? You know, like GTA IV? You'll definitely see a difference in performance depending on cores...

 

I think it was interesting to show how little the COD engine games use the CPU and compare the 2 MW games to see if anything had changed from COD4.

I would like a feature on the Source engine though, as i've played Day of Defeat source since it appeared and the move to the OB engine did show some amazing multi-threading optimisations when compared to the original engine (unfortunately it was one of the few updates it did get). This was from a while back, just after the OB update:

 

2 cores are good enough. Most games arent set up for multi-threading. Or at least to do it efficiently.

It isnt the best way to do it clearly. Using 2 cores I would reccomend an overclock of some sorts. Probably as close to 4mhz as you can get. But it works. And if you have X money to spend I would go with the cheaper processor and throw more cash at a video card or two.

I think the point of the article is to say that 4 cores are fantastic and there is a benefit, however if you are building on a budget a dual core is more than ample to get you from A to B.

 

Thanks for the highly detailed reply Bindi, and at four o'clock in the morning too! (Good God man, don't you sleep?)

OK, I think maybe I simply misunderstood the intention of the article in certain respects, and was reading it expecting something slightly different, in that I think I was expecting more of a proof of concept type of study, ie, raw scalability - something for now and the near future, as opposed to just relevant to the precise here and now.

I get your point, but as Elton said, that's not really a fair or direct comparison is it? That's comparing very slightly different flavours of the same generation, or comparing a dual GPU card to a single GPU card. Comparing a 4870 to a 5870, or a 4870x2 to a 5970 shows a very significant difference.

I think all was getting at was, while not guaranteed, the chances are quite high that the next gen of cards will again be a significant step up from what we have now - and that step up will probably come in the near future. Those cards will much more likely require the CPU to work harder to keep up, at which point the ability to spread the workload across multiple cores becomes even more important than right now.

Also, you said you were trying to give a real-world indication of a high-end gaming rig, which is why you didn't under-clock the CPU or lower the game resolution - which is fair enough - but then surely anyone likely to shell-out £800+ on a CPU for a real-world gaming rig would also be more likely to pay the £490+ needed for a 5970 than the £320+ for a 5870? At which point we still may have seen more difference between the CPU core use and the FPS achieved.

I understand your point about not being able to evaluate what doesn't yet exist, (obviously!) but with a number of your graphs already showing a significant take up of all six cores, this would surely only point to six cores becoming more important over time?

The reason I think this is important, (and I'm not just trying to be pedantic for the sake of it) is that as I said before, I think most people give their CPU's a much longer lifespan than their GPU's, (because of changing socket types necessitating changing the motherboard etc) and so most people will probably get through maybe two or even three graphics cards for every CPU they buy.

I just felt that for a buying guide for a more long-term component, a little more emphasise could have been put on the likelihood of what is probably only just around the corner - as demonstrated already in some of these graphs.

I understand too your point re: AMD Fusion/Intel Sandybridge, but again, I expect most people who bought a 6-core now would probably not be upgrading their entire system again in just 12 months. Some would, obviously, but I think a minority. Many people who bought the early quad cores had to wait a while to make full use of them in games, but many will still be running them now and will be glad they went quad at the time - as it promotes longevity and value in their system.

Anyway, I wasn't trying to be too critical, as I said it was a good article, I just felt that the conclusions drawn didn't really match the results you found or maybe look at the bigger picture in general.

 

Since almost all new games require dual-core CPUs, I think a quad-core makes for an ideal gaming processor myself, which as the article mentions, leaves you breathing space for running a music program in the background, running a virus scan on a non-game drive or for those games, such as Ghostbusters, GTA IV, ArmA II and Prince of Persia: The Forgotten Sands (not covered by this article), that do benefit from having more than two cores. The latter game can use 6 cores according to Ubi Soft.

 

I still rock my 1055t and no regrets about buying. Getting 4 Ghz on air, runing cold. The best procesor I ever had =)

 

Yeah just a few that use the quad very effectively: GTA IV, which has been mentioned here a lot already, also Dragon Age, Lost Planet (i think it even scaled to 8 cores as shown in demos), Fifa 10 uses all 4 cores - i guess there are more that I cannot remember.
So overall, I'd say there's no point in buying a CPU for a gaming rig that has less than 4 cores. (lets not go into hyper threading here as it helps, but it varies in how much it helps case by case).

 

exactly what ive been seeing. which is why i went core i3-530 for $80! although on a p55 board which is giving me constant issues...

 

My teasmade scoffs at your breville

 

Yes The same thing everytime why so shy forge alliance eats any rig because of the memory loop so why not use that as a good test of cpu. (p.s. I use the core Maximizer which allows any number of core and sockets). dig deep and lets see a rig come on.

 

BFBC2 has a single player? lol

Why did they not use X 3 Terran Conflict in this test as it is the perfect game for this test?
The chip in my watch can run COD MW 1 or 2 so why continually use it? yeh I know everyone plays it (except me)

over all great information! With people having multiple monitors and over 6 gb ram I recommend a 4 or 6 or 12 core processor so you can do more at once as I stream live TV while gaming.

bottom line it seems PC game developers are SO FAR Behind the Hardware that they may never ever catch up to utilize today's technology to it's fully optimized potential. You would think a program or a set of instructions would force the cpu to use it's cores evenly no matter what is going on.

 

The amount of time the Cyber-Luddites spend on here and other sites defending their outdated equipment could be put to better use doing some yard work for a few pounds an hour, it would make up the difference between your outdated XP running trashbin and a real gaming computer!

I do not see what benefit people get from supporting inferior equipment. There isn't a computer builder who is a gamer who would be delusional enough to place a dual core in when he could use a quad core. They are cheap as hell so what are we arguing about here?
People who need to upgrade but don't want to/can't afford to. If you can't fine, I understand that not all people have the spare cash, not that it is a lot, or the skill/patience/motivation to upgrade. But PLEASE do not make up specious arguments to support your current state of affairs with specious arguments!

Enjoy your Summer of PC Gaming,
Star*Dagger

P.S. Upgrade!

 

it's not that star.. it's the exact opposite actually- if your looking to build the fastest gaming rig from the ground up your better off knowing what works.. the new box builder usually makes the mistake of overbuying on the cpu and then cutting corners on the gpu

it should be the other way around.. if you can get higher clocks on a very cheap cpu- then look at the numbers.. you can get a better gaming rig for the same price it has nothing to do with defending anything, it's the truth that escapes some people

I know people get hot under the collar because some guy with a cheap chip oc'd to the balls makes their expensive quad look like a p4 in both load and fps.. but that's pretty much fact- your goal is high clocks and the best gpu if your looking at building a fast gaming rig

sli/crossfire is a different story.. but if your going down that road, you can probably spend more on the cpu and board

basically you can get a lot more for your money.. spending a few hundred more just because of epeen is fine.. but realize that's all it is if your only using it for games

 

Greetings!

Is it even wise for someone who buys a new system today to shop bellow an i5 750?

 

Have you got the same graphs?

Four and six cores are the only two setups which never saw any sort of core based performance loss in any of the games tested. Some games such as MW2 and DOW2 required three to plateau so obviously dual core is not always enough, yet BC2 with ATI saw three and five oddly losing performance so three is not the magic number. STALKER had very slightly less performance with three and five cores compared to dual and quad, but quad was also better than dual despite the game being single threaded. The only exception to quad being the best is Dirt 2 with ATI, in which case dual was the best of them all by a couple frames. If anything this test shows a gamer that something like an i3 won't always be enough. It certainly has shown me that an i5-750 would be pretty tasty.