Interesting myth buster :) Good article. In other words, Duel Core is fine for now, but If you game & multi task and your getting a new PC, get Quad core.
Interesting read. You should suspect that by now, after six? years of having duo/quad core technology. Games would be more multi core capable by now. Seems that they could offload much of game calculations back to the CPU instead of GPU. Or is it because more and more game related calculations are done on the GPU leaving the CPU doing nothing much?
It's more or less what I expected. I'm still on dual core because I'd seen nothing to warrant any more than that. I feel vindicated hurrah! I shall stick to my plan of leaving it another year or 2 before I go quad core.
Wasn't there a major debate a few years ago about the relative merits of coding multithreaded gaming applications? I seem to remember things such as race conditions and the linear cause/effect simulations in most games being a hindrance to developers taking up this new technology.
Does anyone have more info on this, and possibly an opinion on why the games tested don't scale beyond three cores?
Kinda makes me feel bad for the fools who've paid the i7-premium for their gaming-only rigs. Seems like a heavily overclocked Wolfdale would still be all it takes to keep gamers happy. And at the very least it seems like budget builders should just forget about the extra cores and spend more money on the graphics.. Kinda what I've thought, I just didn't know it was such a clear cut.
I would suggest returning to this test and do it again, but this time try underclocking the CPU to see the multithreading in real action.
In order to see the benefit of multithreading CPU has to be the bottleneck, but in all these tests after enabling third core GPU becomes the bottleneck. You are not gonna see much improvement from 3 to 6 cores if GPU is the limiting factor. Or you should be looking more at CPU loads with differnet amount of cores enabled.
Essentially from that your saying that the G6980 or i3 530 are the best CPUs for games at the moment/AMD equiv parts??
Id noticed a speed bump running an 530 at 4.8ghz vs at 920 at 4.2ghz myself and you have by this confirmed my thoughts.
Well, that makes me happy as I have been reluctant to sell up my x58 in favour of a cheap i3 system (1 to save on power and 2 because I dont bench anymore).
alternatively 6 cores means you dont have to faff around closing stuff down...
what is heavy multitasking btw... when working have perhaps 20 apps open at a time + 3 database servers, many active and with chrome with flash enabled you could double that... whats the actual break down....
btw I only just moved to a quad and its a clear improvement from the old dual core I had previously which was horrid.
Originally Posted by HyBry Or you should be looking more at CPU loads with differnet amount of cores enabled.
In a perfect world, yes I'd have images for all of them, but it quite long to setup it all up, play the game as it should be and then quickly alt+tab to get a good screenshot - getting 6 results for each game would have taken 6x as long. It was just impractical :/ Looking at all six together tells you where the game stops using cores though, so you can infer from that.
I'd love to have done more but it took nearly 5 days to manually play all these games and get consistent results. :( I might have another go after Starcraft 2 or Civ 5 comes out - test those, Metro 2033, Blur, ARMA II update (if I can get hold of it out here) and whatever else looks interesting.
Hamish - Supcom is ooooooooooold dude. At least people still play CoD 4 MP.
I wouldn't be so quick to say we shouldn't upgrade our CPU for a performance boost. I used to ran a Core 2 duo at 2.66ghz for a dual core game - World of Warcraft, at 25-35fps, now I get 50-60 with my new Intel i5-750 at the same 2.66GHz with the SAME graphics card from my previous system, the Geforce 260 SSC.
If anything, it could be the CPU architecture that improves things rather than the number of cores it has.
As a side-by-side kind-of comparison, it would be interesting to then add to the article the dedicated multi-threading apps (encoding, etc) that do try to use all cores. I know we have access to that info through the CPU tests in the mag etc, but to actually lay that side-by-side would be revealing. I guess it will likely show the slow take-up and usage of many cores for gaming?
Several games seemed to like three cores - could it be that atually they were only using two cores and the benefit of the third is that it picks up all of the Windows background services etc? It would make sense that most game devs are only coding for dual-threading at the mo, on the basis that the proportion of gamers with quads is still comparitively low.
very nice results, shows that game developers still have untapped power to be had. also busts the common myth that you need a £3000 quad core monster machine to play PC games.
from the benchmarks, i see a common trend where 4 core machines seem to always have higher or one of highest minimal FPS. is this because game developers have now started to optimise for 4 cores? why does 6 core give less minimal FPS?
Agreed with what you said about SupCom, but what about something a bit bloatier, like GTA IV? It normally runs all four of my cores to 80-90%. Although a bit of that is probably lazy coding, it does show it's using all available resources, and I'd be interested to see if it scaled up and down in the same way.
what a fantastic article. Thankyou for writing it and all the testing.
i was wondering though, couldn't you compare duals and quads from multiple generations, so from the E2140 E4300 E5200 E6600 E6750 E8600 to the i3s, so we know if for gaming its worth upgrading. Same for quadcores?
and then a heavily oced dual vs quadcore? Also with the core scaling, how about brining in multiGPU gaming into it aswel, from x2 style cards to 4way gpu setups?
Comments 1 to 25 of 123
Reply/meme
In the Gigabyte BIOS of the X58A-UD7 mobo you can specify how many CPU cores you want it to use.
Does anyone have more info on this, and possibly an opinion on why the games tested don't scale beyond three cores?
Thanks for the article, very interesting indeed.
In order to see the benefit of multithreading CPU has to be the bottleneck, but in all these tests after enabling third core GPU becomes the bottleneck. You are not gonna see much improvement from 3 to 6 cores if GPU is the limiting factor. Or you should be looking more at CPU loads with differnet amount of cores enabled.
Id noticed a speed bump running an 530 at 4.8ghz vs at 920 at 4.2ghz myself and you have by this confirmed my thoughts.
Well, that makes me happy as I have been reluctant to sell up my x58 in favour of a cheap i3 system (1 to save on power and 2 because I dont bench anymore).
Nice results,
Cheers!
Andy
what is heavy multitasking btw... when working have perhaps 20 apps open at a time + 3 database servers, many active and with chrome with flash enabled you could double that... whats the actual break down....
btw I only just moved to a quad and its a clear improvement from the old dual core I had previously which was horrid.
I'm still using a old E6600 with a GTX280, it runs everything fine and I see no reason to upgrade any time soon.
supcom was always good at beating the crap out of my quad, i'd like to see if it can handle a hex core too :p
In a perfect world, yes I'd have images for all of them, but it quite long to setup it all up, play the game as it should be and then quickly alt+tab to get a good screenshot - getting 6 results for each game would have taken 6x as long. It was just impractical :/ Looking at all six together tells you where the game stops using cores though, so you can infer from that.
I'd love to have done more but it took nearly 5 days to manually play all these games and get consistent results. :( I might have another go after Starcraft 2 or Civ 5 comes out - test those, Metro 2033, Blur, ARMA II update (if I can get hold of it out here) and whatever else looks interesting.
Hamish - Supcom is ooooooooooold dude. At least people still play CoD 4 MP.
If anything, it could be the CPU architecture that improves things rather than the number of cores it has.
very nice results, shows that game developers still have untapped power to be had. also busts the common myth that you need a £3000 quad core monster machine to play PC games.
from the benchmarks, i see a common trend where 4 core machines seem to always have higher or one of highest minimal FPS. is this because game developers have now started to optimise for 4 cores? why does 6 core give less minimal FPS?
With my Phenom 550 in dual core mode (@3.6Ghz) i see regular frame rate drops to the mid 40's...
Play with 3 or more cores active (again @3.6Ghz) i see a constant 55-60fps....
i was wondering though, couldn't you compare duals and quads from multiple generations, so from the E2140 E4300 E5200 E6600 E6750 E8600 to the i3s, so we know if for gaming its worth upgrading. Same for quadcores?
and then a heavily oced dual vs quadcore? Also with the core scaling, how about brining in multiGPU gaming into it aswel, from x2 style cards to 4way gpu setups?
-
« Previous
-
1
-
2
-
3
-
4
-
5
-
Next »
Discuss in the forums