The great dual vs quad vs hex gaming benchmarkapalooza

[AdamK47]

The screenshots add substance. It was also easier to do this way.

Also note: A person could play just fine with the dual core setting in any of these games backed by the rest of the hardware in my system. The goal of the test was to show the progression from dual to quad to hex when very little of a performance bottleneck exists in any other sub component of the system (video, memory, etc.).

 

[Topweasel]

Best. Thread. Ever.

But I agree with others about Hyperthreading... back in the 1 vs. 2 core debates, those Pentium4's with HT did almost as well as a dual-core in many cases. If this means the games are benefiting from 4 vs. 8 vs. 12 "cores", that should give AMD's 6-core CPU's a benefit. (Yet the 2500k beats them?)

It depends on A.) the CPU dependency of the games. B.) how well they are multi threaded. Many of the examples here are games that the games either are not very CPU dependent even if they are multithreaded.

The other problem is multiplayer. I wouldn't worry about a single CPU you can buy today (ie not soldered) in SP. But many games have been shown (though not as well as a benchmark) to vary heavily dependent on cores. BF2BC2 is one example probably the best right now.

 

[-Slacker-]

Can you bench with just one gtx480? That would do more to reflect real world performance, since most people don't even have that much horsepower, let alone tri sli, but most importantly 2 cores could be a serious bottleneck for triple card setups, which is probably why you get those big gaps in fps between switching the no. of cores.

 

[AdamK47]

Looks like I exceeded bandwidth. Darn.

I'll try to fix that.

 

[AdamK47]

Can you bench with just one gtx480? That would do more to reflect real world performance, since most people don't even have that much horsepower, let alone tri sli, but most importantly 2 cores could be a serious bottleneck for triple card setups, which is probably why you get those big gaps in fps between switching the no. of cores.

I'll gladly do it if someone can provide the hardware and compensate for the time.

I just did these benches out of my own curiosity.

 

[Smartazz]

Awesome thread! thanks for taking the time to run these tests. Awesome rig btw.

 

[toyota]

Awesome! Looks like you really don't "need" anything more than a dual-core at the moment. Quad seems to help a bit with the min framerates, though. And hex-core appears to be downright useless... for gaming anyway.

high clock speed, high IPC , 12mb cache and HT is keeping that from being a typical dual core by any means. even if he had HT off and only used 2 cores that does not mean just any dual core can play every game smoothly. the architecture plays a massive role.

 

[Ben90]

Curious there was still scaling in the Doom 3 benchmark. I would imagine that ID Tech 4 cannot assign more than 2 threads, much less more than 4. I might do some more research to back up that claim later.

If these results are representative, it looks like the windows scheduler is helping the results a decent amount when adding cores. If that's a good or bad thing, I have no idea. Just curious there was still scaling in that old of a game. Again it always could be margin of error.

 

[podspi]

high clock speed, high IPC , 12mb cache and HT is keeping that from being a typical dual core by any means. even if he had HT off and only used 2 cores that does not mean just any dual core can play every game smoothly. the architecture plays a massive role.

I wouldn't be surprised if an i3-2100 managed. Keep in mind I'm not talking about "every game at the highest settings" I'm talking about just being able to play the games reasonably.


But yea, that would be one heck of an dual-core

 

[AdamK47]

Curious there was still scaling in the Doom 3 benchmark. I would imagine that ID Tech 4 cannot assign more than 2 threads, much less more than 4. I might do some more research to back up that claim later.

If these results are representative, it looks like the windows scheduler is helping the results a decent amount when adding cores. If that's a good or bad thing, I have no idea. Just curious there was still scaling in that old of a game. Again it always could be margin of error.

I found that odd too for Doom 3. One thing I noticed though is that the Doom 3 profile in the nVidia control panel has "Thread optimization" set to "NVIDIA recommended".

 

[sm625]

Three different browsers and I dont see anything. Whatever you are doing, you are surely doing it wrong. Please stop.

 

[max347]

Three different browsers and I dont see anything. Whatever you are doing, you are surely doing it wrong. Please stop.

I see them fine.


Also, it amazes me the lack of gratitude in this thread. The dude went and did a thorough-a$$ job of benchmarking these things and you see comments like "couldn't you just make a graph?". C'mon people. If he did that, people would come in and "pics or shens" the guy.

OP- thanks for the great info. I agree, a 4.1ghz HT dual with 12mb is not your average dual, but I don't think it is the actual numbers that you are pushing here, but the scaling. The differences between 2, 4, and 6 were the main thing here, and I think this showed what most people who clicked on the thread wanted to see.

 

[Voo]

Nice work, surely lots of work.

Three different browsers and I dont see anything. Whatever you are doing, you are surely doing it wrong. Please stop.

A dozen people having absolutely no problems seeing the pictures, you having one - yep obviously the poster is doing something wrong! I mean there can't be any other explanation for it.

high clock speed, high IPC , 12mb cache and HT is keeping that from being a typical dual core by any means

Yep as is well known high clock speeds and IPC are things totally impossible to achieve with dual cores - you really need those extra four cores for that.

Too bad there weren't any tests without HT to see how only two cores fare, but still nicely done.

PS: About all those people whining about graphs - never thought about doing that yourself if it's that important for you?

 

[sm625]

There isnothing wrong on my end. I can surf the web just find. I can go to half a dozen tech sites and look at all the graphs I want. But this thread is borked and I cant see anything. Whatever he is doing, he is obviously doing it wrong. Maybe his stupid screenshots are blocked... but who cares. That's why somebody said dont post them. They dont work for everyone. I wonder how many people tried loading this page on their smartphone...

 

[AdamK47]

Three different browsers and I dont see anything. Whatever you are doing, you are surely doing it wrong. Please stop.

I originally had them on Photobucket, but my bandwidth to the site was quickly exceeded. Moved all of the images to pics.bbzzdd. There shouldn't be anything wrong with that since the image hosting site was spawned from these forums. Many members here have been using the site for years.

 

[VirtualLarry]

Screenshots load fine for me, on both a FIOS connection and a Comcast connection.

THANK YOU, AdamK47, for doing all this work.

 

[poohbear]

ofcourse 6 cores are'nt much of a step, no games have been made that use it! at the most are quad core: Battlefield company 2, shogun 2, witcher 2, farcry2.

 

[toyota]

ofcourse 6 cores are'nt much of a step, no games have been made that use it! at the most are quad core: Battlefield company 2, shogun 2, witcher 2, farcry2.

there are games that use 6 cores but you dont really need 6 cores to play them obviously. most games certainly use more than 2 cores now though. and again cpu architecture plays a huge role. overall a 125 watt Phenom X4 is slower than a 65 watt dual core i3 with HT in gaming.

 

[AdamK47]

ofcourse 6 cores are'nt much of a step, no games have been made that use it! at the most are quad core: Battlefield company 2, shogun 2, witcher 2, farcry2.

I don't think you can say no games have been made that use six cores. If you take a look at the shots you can see that Crysis 2, Mafia 2, Far Cry 2, STALKER CoP, Battleforge, Civilization 5, and the Capcom games show an increase. It may not be as profound when going from 2 to 4, but the increase is there. The games that didn't show much of an increase (or any at all) are Metro 2033, Call of Juarez, AvP, GTA 4, Shogun 2, Just Cause 2, Company of Heroes, Batman, and Doom 3 (that's a curious one anyway). AvP, Heaven, and the game tests in 3DMark didn't show any increase going beyond dual core. If Battlefield BC 2 had a benchmark, I would run that as well. I can see some good utilization of all 6 cores in task manager.

 

[Phynaz]

There isnothing wrong on my end. I can surf the web just find. I can go to half a dozen tech sites and look at all the graphs I want. But this thread is borked and I cant see anything. Whatever he is doing, he is obviously doing it wrong. Maybe his stupid screenshots are blocked... but who cares. That's why somebody said dont post them. They dont work for everyone. I wonder how many people tried loading this page on their smartphone...

Go thread crap elsewhere.

 

[Avalon]

Adam, thanks for the fantastic effort! I love it when AT users take the time to do real benchmarking work and share with everyone else. That's what a community is all about.

 

[sm625]

It is not thread crapping to point out the fact that at least 5% of the people who load this thread arent going to be able to see wtf is supposed to be here. Pointing that out reduces the possibility of seeing more of it in the future. Stuff that is posted here should *just work*. Period. No ifs and or buts. It's not my problem if it dont work. Stick with what works.

 

[Rvenger]

Three different browsers and I dont see anything. Whatever you are doing, you are surely doing it wrong. Please stop.

pics.bbzzdd comes up with 'provacative attire' at my work and posts Red X's everywhere. If I switch on to my other laptop with an outside connection, they show just fine.

 

[sm625]

Yep, it is blocked. IE7 actually shows me the red boxes. The other browsers dont. Note that simple text reads much faster, and is never blocked! Now that I've been able to actually look at these pics through a proxy, the big question is: what about 3 cores? Hasnt that been shown to be the critical cutoff point for good performance?

 

[toyota]

if you are building a pc for gaming and want the best bang for buck then you go 2500k. four fast Sandy Bridge cores will give you all the cpu power you need in games and will drive any high end gpu setup no matter how many cores a game effectively uses. Phenom X3, X4, and X6 stand no chance all while using more power. so again it is the cpu architecture that matters most here and you cant determine how many cores are needed just from looking at one type of cpu.