weak quad vs strong dual

[sm625]

So boredom and curiosity got the best of me and cost me about $300. Snagged a used 2600k and Z77 board to use as a test bed. I'll be running the specific test in the OP. Running it with 2 cores HT on at 3GHz and dropping it to 2GHz with all 4 cores going but HT off.

Wow now that's the enthusiast spirit. Can you also run passmark?

 

[2is]

Nice!

Well i guess it depends on games you got?

Modern (DX11) multi threaded games i know about are:
BF3 - up to 8 cores
Company of heroes 2 - don't know
Wargame airland battle - up to 4 cores, possibly more
Civilization V - don't know
Crysis 3 - don't know

The choice for older stuff or games that only use two cores is pretty extensive, popular ones are:
Fallout3/new vegas
Skyrim
Left 4 dead 2
Bioshock infinite

Out of the games you listed I have BF3, Crysis 3, Fallout 3, L4D(original)

Other notable mentions are:

Dota 2
Arma 3 (beta)
Borderlands 2
Far Cry 3 (as well as 1 and 2 but I don't think those are relevant)
Max Payne 3
Deus Ex: HR
Dirt 3
Grid 2
Black Ops 1 & 2 (1 was a CPU killer)
All of the Assassins Creed games (minus the DLCs)
Mirrors Edge

Wow now that's the enthusiast spirit. Can you also run passmark?

I'll do that. No gym plans or females to entertain tonight so I should be able to geek out and run many of these tests when I get home from work.

I would have had some BF3 results last night but when I rebooted to do my 2nd run battlelog was down for maintenance it's probably for the better though, I didn't like the map choice (siene crossing) and later I discovered the 650Ti has a nice bit of OCing headroom, so I'll be applying that to all the tests just to further eliminate a GPU bottleneck. And I think I'll go with Caspian Border. It's very easy to find a nearly full 64 player server, it's a big map and it has a good mix of air and ground battles going on.

 

[Maximilian]

Theres also the 64 player operation metro servers to consider (if those are still around), they might stress the CPU in a reasonably consistent fashion as everyone usually gets bogged down around the same escalator/hallway area and the throw nade/die, come back throw nade/die clusterfuck cycle begins :awe:

 

[Blastman]

If you're going to run CPU gaming benchmarks, make sure you are not GPU limited. This means running the benches at 800x600 or even 640x480 like computerbase.de does with no AA. If you are GPU limited the benchmarks will be useless as CPU benches.

 

[2is]

Yep, I know to make sure I'm not GPU limited, but I will not be running the games at that low of a resolution since no one actually plays at that resolution. It's not that hard to run at 16x10 and not be GPU limited in most games. Not to mention, lowering resolution isn't exclusive to lowering GPU load but also puts less load on the CPU as well.

 

[crashtech]

Inspired by 2is, I did some informal testing of my own last night with my 2700K and Tomb Raider at 1080 and High quality. The 2GHz quad and GHz dual+HT were very close, but the quad had a tiny edge at that res, about 0.5% . The synthetics of course showed multi going to the quad and single going to the dual. I'll be interested in seeing more tests.

 

[Concillian]

Inspired by 2is, I did some informal testing of my own last night with my 2700K and Tomb Raider at 1080 and High quality. The 2GHz quad and GHz dual+HT were very close, but the quad had a tiny edge at that res, about 0.5% . The synthetics of course showed multi going to the quad and single going to the dual. I'll be interested in seeing more tests.

Which is my point. When the 2C/4T wins, it will win by a large margin thanks to the +50% clock benefit. When the 4C/4T wins, it will win by a much smaller margin... so the 2C/4T will be the better all around choice unless you have a usage profile that is virtually 100% 4 threads.

 

[crashtech]

I happen to agree with you, but I am dubious that the chosen benchmarks will properly represent the kind of overall usage pattern we envision.

 

[2is]

ok first benchmark result is in:

BF3 MP Caspian Border
2C/4T @ 3GHz
Avg: 61 Min: 23 Max: 105

4C4T @ 2GHz
Avg: 60 Min: 39 Max: 92

The intangibles:
For THIS game, I preferred the gameplay of the lower clocked quad core. Both were more than playable but 4 real cores provided for smoother and more consistent performance.

More to come!

EDIT:
Passmark Scores:

2C4T @ 3GHz

4C4T @ 2GHz

Cinebench

2C4T @ 3GHz

4C4T @ 2GHz

Handbrake Blood Diamond Encode (AppleTV 3 Preset)

2C4T @ 3GHz Avg FPS: 56.3
4C4T @ 2GHz Avg FPS: 61.1

Black Ops II TDM
2C4T @ 3GHz
Avg: 174 Min: 124 Max 197

4C4T @ 2GHz
Avg: 160 Min: 106 Max 198

NFS Shift 2
2C4T
Avg: 75 Min: 39 Max: 98

4C4T
Avg: 64 Min: 31 Max: 84

Far Cry 3
2C4T
Avg: 69 Min: 49 Max: 103

4C4T
Avg: 80 Min: 48 Max: 118

Fallout 3
2C4T
Avg: 96 Min: 14 Max: 114

4C4T
Avg: 90 Min: 14 Max: 112

Assassins Creed III
2C4T
Avg: 68 Min: 35 Max: 212

4C4T
Avg: 73 Min: 35 Max: 210

Dirt 3
2C4T
Avg: 118 Min: 95 Max: 144

4C4T
Avg: 106 Min: 89 Max: 124

 

[2is]

bump for updates

 

[crashtech]

The lone surprise is Black Ops II. Kudos for posting this interesting info!

 

[2is]

Just added NFS Shift 2. Similar results as BO2 with the higher clocked dual edging the quad.

 

[SPBHM]

your test is showing no big advantage to any side imo, can you measure the power usage while running MT benchmarks?

also, your CB11.5 result is a little lower than what I would expect, considering a i3 2100 gets 3.00 (with 100MHz more, 3MB of l3, but CB doesn't seem to care about l3 much anyway.)

thanks for the testing

 

[2is]

your test is showing no big advantage to any side imo, can you measure the power usage while running MT benchmarks?

Yes, I have a KaW. What would you like to see?

 

[SPBHM]

Yes, I have a KaW. What would you like to see?

power usage comparison while running CB, to compare score vs power used for 4c vs 2c4t (with the clock difference)

 

[2is]

Ok, coming up. btw I just tried CB at 3.1GHz 2C4T and got 3.03 so it sounds like it's just where it should be. Power consumption at 3.1 was almost exactly 80 watts. All voltages set to auto.

 

[2is]

2C4T @ 3.1 79-81 watts
2C4T @ 3.0 77-80 watts
4C4T @ 2.0 73-75 watts
2600k @ stock w/turbo enabled 130 watts

Btw, this just being a simple test bed, I'm not using the fanciest PSU. Just a 350 watt Antec Smart Power. Doesn't even have APFC so it's probably not terribly efficient.

 

[SPBHM]

thanks again, not a huge difference but, overall it looks like 4C is a slightly better combination, and if you combine with turbo for lighter loads it should make the choice easier.

 

[taltamir]

Back when this discussion was between the E8400 vs Q6600 the decision was pretty clear (for me) that the Q6600 was the better choice. E8400 only had a 25% clock speed advantage but NO hyper threading. In this case, we are talking about a 50% speed advantage for for dual core, and it also has the benefit of HT so the performance hit from >2 threaded games isn't as big as a dual core without HT.

I upgraded my Q6600 to an E8400 and for a time it was good...
Then 10 months later I upgraded back to a Q6600... after I found out microstuttering in mass effect 1 was being caused by my CPU being too weak (at any resolution and graphics quality setting), the Q6600 fixed it.

The vast majority of modern games in 2013 fall into one of two categories:
1. Requires insignificant amount of CPU so it doesn't matter
2. Requires at least a quad core modern CPU (so, generally intel, although maybe latest gen AMD can compete)

they just don't make games anymore that are both heavy AND unable to use quad core.

And all of the above was back in the days before turbo... turbo changes the game. With turbo your quad core basically turns into a higher clocked dual core when it is needed

 

[2is]

And all of the above was back in the days before turbo... turbo changes the game. With turbo your quad core basically turns into a higher clocked dual core when it is needed

I'd agree with you except the OP states no turbo. I'll be testing some games that are not so well threaded a little later.

 

[taltamir]

I'd agree with you except the OP states no turbo. I'll be testing some games that are not so well threaded a little later.

those games are not that demanding

 

[2is]

But they exist and people play them and a 2ghz quad that isn't being taken advantage of isn't that powerful so it'll be perfect. Again, this is not only with no turbo, but the added benefit of HT for the dual. So basically taking away the advantage quad cores have received over the years (turbo) and allowing the dual core its advantage it has received over the years (HT) if it weren't for those stipulations, it would be a clear choice for the quad. As it stands now they are quite close in everything. They're trading blows but neither one is dominating the other. Most of the stuff I've tested so far has been reasonably well threaded.

 

[taltamir]

But they exist

when I say not demanding I mean they will not be limited by a MODERN quad and as such are not relevant. Who cares if a game clocks 3000 fps on setup A and 4000 FPS on setup B. What matters is when a game clocks somewhere under 120 fps, because that monitors are limited to that (typically limited to 60fps actually)

And the vast vast majority of games which are demanding do support quads. for example, every single unreal engine 3 game out there runs better on a Q6600 than on a E8400

Also, an HT CPU is not a true dual core. HT might be less agressive than AMD's "double" core design (what did they call it?), but it is still a hybrid between 1 and 2 cores. That is, certain components of a CPU's core are wholly duplicated in an HT core, it doesn't merely lie to the OS saying it has an extra core, it actually has some extras (not a fully extra core though).

 

[2is]

None of the games I'm testing are getting 4000 fps, many are hitting minimums in the 30s and one dropped to 20. Not only are the processor comparisons not as you described but the games aren't either. I understand what HT is, not sure what your point is though. The question asked was 3ghz dual core with HT vs 2ghz quad core, no HT no turbo and that's what's being compared. I didn't arbitrarily choose it.

 

[taltamir]

Can you name those games you are testing? and can you explain why you are running this test?
And it wasn't "a processor" it was specifically a sandy bridge processor.

not sure what your point is though.

I made a variety of points.