Company of Heroes 2 - fascinating CPU benchmarks

[SPBHM]

I think you need your eyes checked or at least notice that the Phenom II x2/x3 Athlon II x4 has no scaling. Athlon II X2 265 has much slower caches than Athlon II X3 455, which leads to lower performance in the single core workload.

Athlon II X2 = 2MB l2
Athlon II X3 = 1.5MB l2

it's faster for ST

 

[Sweepr]

An old quad-core Core i7 920 @ 2.66GHz beats the dual-core Core i3 3220 @ 3.3GHz. The game definitely benefits from the extra threads.

 

[frozentundra123456]

An old quad-core Core i7 920 @ 2.66GHz beats the dual-core Core i3 3220 @ 3.3GHz. The game definitely benefits from the extra threads.

Yea, my feeling is that it uses four cores very well, but little or more past that.

 

[nyker96]

well we all know Intel is better for gaming, so this isn't anything surprising. I'd be surprised if opposite happens. I actually play this game on 2500k w/ a 7790, I think it's very gpu limited, cpu isn't as important for it.

 

[USER8000]

Someone explain what's wrong with this picture? Specifically the 3960X? It looks like the game, yet again, is using a single core. Gotta love it when developers are optimizing for Windows Vista and a single core. I'm optimistic that the next generation of multi platform games will address this - frankly, i'm getting sick of poorly coded games that are designed for 32 bit software/OS released 6-7 years ago. It's maddening.

It looks like the Core i3 3220 is not that impressive in that game,and the equivalently priced AMD A10,and FX4300 and FX6300 series CPUs are generally as fast or are quicker.

Also,the other problem is that with a Geforce Titan,the game is only running at 41FPS with a Core i7 4770K,and a cheapo FX4320 is not much slower it seems(35FPS). The game appears to be very GPU heavy at even 1920X1080.

Even at medium quality with less of a GPU bottleneck,it seems the AMD budget CPUs are much faster than a Core i3:

http://static.techspot.com/articles-info/689/bench/CPU_02.png

 

[DominionSeraph]

I think you need your eyes checked or at least notice that the Phenom II x2/x3 Athlon II x4 has no scaling. Athlon II X2 265 has much slower caches than Athlon II X3 455, which leads to lower performance in the single core workload.

Athlon II X3 to X4: reduction in clockspeed by 3% yet increase in FPS by 15%
Phenom II X3 to X4: increase in clockspeed by 23%. Increase in FPS of 35%.

So we have scaling past three cores.

If this was single-threaded, Faildozer would be barely above the Athlon X2.

 

[parvadomus]

This thing uses like 2-3 threads. If i3 had turbo core, and extra cache it would be really near the i5s.

 

[BigChickenJim]

CoH 2 has been widely panned by players for being a horribly optimized mess. Give the devs a few months to tinker and that may change, but for now it's probably best to avoid using a piece of questionably designed, hardware-inefficient software as a measuring stick for CPU performance or bottlenecking.

As of right now the only thing I find "fascinating" about the game is the fact that it made it out the door in 2013 with such bad optimization and no Crossfire or Sli compatibility at all.

 

[moonbogg]

Why should the 6 core CPU results surprise anyone? The game uses no more than 4 it would seem, and the clock speed along with ipc being less on the 6 cores, it should be no wonder it is slower.

 

[know of fence]

CoH 2 has been widely panned by players for being a horribly optimized mess. Give the devs a few months to tinker and that may change, but for now it's probably best to avoid using a piece of questionably designed, hardware-inefficient software as a measuring stick for CPU performance or bottlenecking.

As of right now the only thing I find "fascinating" about the game is the fact that it made it out the door in 2013 with such bad optimization and no Crossfire or Sli compatibility at all.

Their parent company THQ went bankrupt and the IP and studios were sold, Relic Entertainment was auctioned off to Sega, who probably want a return on their 26 Mio$ investment. At least the devs put a built in benchmark into the game.

At maximum quality both i7 and FX deliver 41 / 40 FPS, so much for the CPU relevance.
Maybe the Techspot guys should have made an effort to tweak the settings to get meaningful results, beyond just Medium / Maximum considering there aren't too many knobs to turn.
But that would involve something other than a by-the numbers review routine, and useful recommendations on how to tweak the game. Seriously who wants that, when you can have pretty, and probably hand-picked for clarity FPS graphs.

 

[Mallibu]

At maximum quality both i7 and FX deliver 41 / 40 FPS, so much for the CPU relevance.

The big difference, is that you turn down a setting or two and get the desirable 60 fps on the i7, whereas you're stuck with 40 fps with the FX.
CPU bottlenecks is the last thing you want in gaming.

 

[USER8000]

The big difference, is that you turn down a setting or two and get the desirable 60 fps on the i7, whereas you're stuck with 40 fps with the FX.
CPU bottlenecks is the last thing you want in gaming.

It seems for people on a budget playing this game,every Intel CPU under a Core i5 is not that great.

 

[galego]

Techspot's review is out; salient points:

"Effectively what this means is that the Core i7-4770K can push the GTX Titan to its limit in CoH 2 at just 2.5GHz while the FX-8350 needs to be clocked 80% higher to achieve the same level of performance."

"Now let's see how a range of CPUs handle CoH with the GTX Titan and max quality settings. As we've already seen, the Core i7-4770K is capable of 41fps and the FX-8350 just 37fps, while the Core i7-3770K matched the i7-4770K and the i5-3570K was just 1fps slower. The Core i7-3960X was slower than the i5-3570K which is very surprising given how CPU demanding this game is and makes us regret using it as the primary test chip."

"The lower-end Athlon II X4 quad-cores are useless in this game, as are the dual-core parts including the Core i3-3220."

"With the more playable medium quality settings . . . . . Here the FX-8350 was 31% slower than the Core i7-4770K and 22% slower than its competitor, the Core i5-3470. The old i7-920 was able to match the FX-8350, while the Phenom II X6 and X4 processors showed their age here."

Source: http://www.techspot.com/review/689-company-of-heroes-2-performance/page4.html

I know that AMD's chips are in next gen consoles, so you might think they'd perform better out the box, but it really doesn't look like it. Also hexa-core results are odd given the CPU grunt this game needs.

Well looks at their results

Difference between i5-3570k (4 threads) and i7-3770k (8 threads) entirely explained by clock speed alone.

The same for FX-4320 (4 threads), FX-6350 (6 threads), and FX-8350 (8 threads). The 8350 is performing like the 4320 because the game is ignoring one half of the extra cores.

This is a game that scales with more cores/threads

This is still far from what we will see from consoles (games will be optimized for more than four threads)

 

[PPB]

Running the icc_patcher program tells CoH2 is in fact compiled with ICC, so nothing really surprising (besides yet another RTS who is poorly optimized on CPU threads).

PD: Just to clarify, this is one of those games that even 30 FPS is completely playable, unlike shooters where the game experience is completely crappy. At least that was my experience during the first CoH2 beta phase (the one before the UI rework).

 

[galego]

Running the icc_patcher program tells CoH2 is in fact compiled with ICC, so nothing really surprising (besides yet another RTS who is poorly optimized on CPU threads).

PD: Just to clarify, this is one of those games that even 30 FPS is completely playable, unlike shooters where the game experience is completely crappy. At least that was my experience during the first CoH2 beta phase (the one before the UI rework).

Interesting, besides ignoring cores beyond 4, there is about a 15% less performance due to the Cripple_AMD function. Moreover, they use W7 without the FX patches. The bad W7 scheduler produces the biggest performance drop for medium-threaded games. I would consider another extra 5% less performance on 8350 due to that.

 

[parvadomus]

Does this have a built in cpu benchmark? Maybe someone with an FX can bench it with 1C per module to look for some gain. I have seen a lot of CPU benchmarks at GameGPU, and with poor threaded games, intel CPUs use just 1 thread per core (like disabling HT), however AMD CPUs dont do this, they will just throw more threads into the same module, when there are free ones...
It looks to me like its all cause of the wintel scheduler built in windows..

 

[PPB]

It does, sadly i dont own a FX cpu to test it and I no longer have access to the beta as it was released already. But from what I remember the bench on this game really brings your hardware to its kness as it has a part where the camera is leveled to almost 3rd person view (I mean, instead of your normal axo/isometric aerial view the camera can go as low as the very ground level), and that where the game gets REALLY REALLY demanding as far as FPS goes.

Even tho the game can get demanding at big skirmishes in a battle for a victory point or stuff, if you stay at your normal axo/isometric view things will be OK on most hardware, as I repeat this game is completely playable at 30 fps (lower and it could start to affect your APM/micromanaging of units).

 

[Durvelle27]

Running the icc_patcher program tells CoH2 is in fact compiled with ICC, so nothing really surprising (besides yet another RTS who is poorly optimized on CPU threads).

PD: Just to clarify, this is one of those games that even 30 FPS is completely playable, unlike shooters where the game experience is completely crappy. At least that was my experience during the first CoH2 beta phase (the one before the UI rework).

I could test it when i get home

 

[SPBHM]

It seems for people on a budget playing this game,every Intel CPU under a Core i5 is not that great.

true, it's a weak showing for the i3 3220 (to bad they didn't include some "pentium"), the money saved compared to the cheapest i5 ($180) is not justifiable for games like this (and many others) imo


and...

some people found a way to bring the old Crysis 3 graphic (with an old version and the only scene with this kind of result) and Intel compiler back here

 

[d3m]

Running the icc_patcher program tells CoH2 is in fact compiled with ICC, so nothing really surprising (besides yet another RTS who is poorly optimized on CPU threads).

PD: Just to clarify, this is one of those games that even 30 FPS is completely playable, unlike shooters where the game experience is completely crappy. At least that was my experience during the first CoH2 beta phase (the one before the UI rework).

What the hell?

 

[SPBHM]

What the hell?

Titan is to slow to go any higher at these settings.

 

[d3m]

What kinda code is this game using?

i7-4770K @ 2.5 GHz is giving the same FPS results as at 4.5 GHz.

 

[PPB]

true, it's a weak showing for the i3 3220 (to bad they didn't include some "pentium"), the money saved compared to the cheapest i5 ($180) is not justifiable for games like this (and many others) imo


and...

some people found a way to bring the old Crysis 3 graphic (with an old version and the only scene with this kind of result) and Intel compiler back here

About the ICC, I just tell the way it is. Wether or not you think it still detrimental into AMD CPUs performance its not really the concern here.

In my opinion, its really sad that the only think we got out of that FTC settlement was a tiny paragraph in Intel's page. And its sadder some devs still choose ICC in some games.

EDIT: About the above post, that chart just shows how GPU intensive the game is. At least on highest settings (most people wouldnt give 2 cents for eyecandy on RTS, but oh well).

 

[SPBHM]

What kinda code is this game using?

i7-4770K @ 2.5 GHz is giving the same FPS results as at 4.5 GHz.

faster CPU is not going to make the GPU go faster than its max performance.
it's a clear case of GPU limited test, GPU bottleneck... that why it's a good thing to also test with lower settings/res, because you can always turn dow a few settings for better framerate with a fast CPU, also a faster VGA (or combination), like a GTX 780/Titan SLI would show some gains with clock probably,

 

[know of fence]

For the record they test the fastest graphics card, slower cards are even LESS LIKELY to be CPU limited/bottlenecked, and only in a weird scenario when you push the frame rate very high for which the software wasn't really designed for.

So for all we know a Core i3 or a Phenom x3 coupled with a GTX 660 Ti may do just as well as a quad core would with the same card.

faster CPU is not going to make the GPU go faster than its max performance.
it's a clear case of GPU limited test, GPU bottleneck... that why it's a good thing to also test with lower settings/res, because you can always turn down a few settings for better framerate with a fast CPU, also a faster VGA (or combination), like a GTX 780/Titan SLI would show some gains with clock probably,

"Although we didn't show SLI performance in this review, we did try it and the scaling was very poor. It was much the same when testing Crossfire, and with no profiles for either technology we did a little digging into the subject. It turns out CoH will never feature SLI or Crossfire support -- at least not without some major work. Relic has confirmed that support won't be introduced and Nvidia has said that it is powerless when it comes to adding SLI." - Techspot