*PICS* BC2 from 60fps to 120fps with i5-2500k upgrade from Q9650 Benchmark

S

SHiZNiLTi

Gawd
Joined
Jan 15, 2006
Messages
927
Q9650@4ghz vs [email protected] both systems using a single ati 5870

1600x1200
4xAA, 16xAF
DX11 All Settings on Highest
HBAO off, Bloom off, fov=95
Full 32 person server


i5-2500k... GPU usage is at 99%, CPU usage is ~70%...
2500k2.jpg



Q9650... GPU usage CPU bottlenecked at 60%, CPU is at ~85%...
9650o.jpg



i5-2500k, SS location 1...
bfbc2game20110215020858.jpg


Q9650, SS location 1...
bfbc2game20110313020921.jpg



http://www.youtube.com/watch?v=uiqS341ibbc

*make sure to watch in 480*

Both systems are using win7 64bit and have 4gb of memory installed.

1600x1200
4xAA, 16xAF
DX11 All Settings on Highest
VSYNC off, HBAO off, Bloom off, fov=95
Full 32 person server

Q9650 Info in video:
CPU usage was around 85-95%
GPU usage was around 70-82%
RAM usage was around 2.45gb

2500k Info in video:
CPU usage was around 45-65%
GPU usage was pegged 99%
RAM usage was around 2.93gb
​
 
Last edited:
wow. Thanks for posting this. its a neat thing to see. It would be interesting to see how CPU speeds decrease the bottleneck.

The thing is though, that seems like a crazy jump up in frames does it not?

Its not possible that Vsync was enabled for the other computer is it?
 
Ya I plan on on doing more tests with this 2500k, this 4.2ghz was just a quick 5 sec overclock.

I already know going from a stock speed of 3ghz to a OC of 4.27ghz on the Q9650 I gain about 20-25fps, will be interesting to see how the 2500k does. I'll keep making sure to capture results in a full 32man server since it's way more CPU intensive then an empty server.

I know for a fact that Vsync was disabled, we are both on CRT's, the Q9650 was hitting 100+ fps when I turned around and faced the water.

From just messing around on both setups tonight it looks like the average difference is around 35-65fps depending on whats going on in the map.
 
Awesome! You got my support, I will be watching this thread closely. This thread will help me decide on what to buy.
 
thanks. sending thread link to a friend with an E6850@stock+GTX570.
 
Here's a few more at

1600x1200
4xAA, 16xAF
DX11 All Settings on Highest
Vsync off, HBAO off, Bloom off, fov=95
Full 32 person server


SS.2 i5-2500k...
bfbc2game20110215035748.jpg


SS.2 Q9650...
bfbc2game20110313035810.jpg



SS.3 i5-2500k...
bfbc2game20110215035424.jpg


SS.3 Q9650...
bfbc2game20110313035446.jpg



The next batch of screenshots will be done at a very low rez to truly see the CPU difference.
1024x768
noAA, noAF
DX11, lowest graphic settings
Vsync off, HBAO off, Bloom off, fov=95
Full 32 person server
 
Last edited:
1024x768
noAA, noAF
DX11, lowest graphic settings
Vsync off, HBAO off, Bloom off, fov=95
Full 32 person server

The point to benchmark the game at a lowest settings 1024x768 is to put more focus onto the CPU, just like how 3dmark CPU tests, they run at ultra low resolution.

Here's some info...

http://www.techarp.com/showarticle.aspx?artno=493

"I have to ask "Do you know that CPU benchmark are always done with low resolution?".

Why? In order to eliminate the GFX bottleneck at high resolution!

In fact, the higher the resolution you are at, the lower the impact from CPU.

Hence, it is best to compare the gaming performance between CPUs at low resolution."

"Both CPU and GPUs work harder at higher resolution than at lower resolution.

However, GPUs are far more dependent on the resolution than CPUs.

So yep, CPU bottlenecks can be sniffed out using lower resolutions and a powerful gaming GPU. "


i5-2500k SS.1...
bfbc2game20110215055410.jpg

Q9650 SS.1...
bfbc2game20110313055432.jpg



i5-2500k SS.2...
bfbc2game20110215055328.jpg

Q9650 SS.2...
bfbc2game20110313055350.jpg



i5-2500k Settings...
bfbc2game20110215054805.jpg

Q9650 Settings...
bfbc2game20110313054722.jpg
 
Last edited:
i've been saying this about BC2 for a long time.

in sp there is not much difference, but mp is a whole different story.
 
Last edited:
Wow. I guess there IS a huge difference in multiplayer. In the next 5 days its settled, i am ordering my 2500k and a new board. Hopefully my phenom II sells. This way I dont spend a lot out of pocket.
 
Hmm, very interesting.
It would be better if both were run at the same mhz, though. But I'm sort of shocked at the FPS being *DOUBLE*. Because in superPi, going from a Yorkfield to a 2600k (both at 4 ghz) gives about 30-35% higher fps in Morrowind and Vietcong, and seems to relate directly to SuperPi: superpi 1m (11.7 to 9.406; I estimated that a 2600k @ 3.5 ghz would give 11.7 1M, and a Q9650 @ 5 ghz would give around 9.4 1M), and in Vietcong (Intro level), which is purely cpu limited and a rather old game (enabling supersampling has no effect on FPS), I got about 35% higher FPS. And at 5 ghz, my FPS wasn't "quite" double (I think the minimum FPS was 53 (QX9650 @ 4 ghz) to 100 or 98 (2600k @ 5 ghz). This is a single threaded game, though.

Yet you're getting more than double the FPS in serious sam 2. That's rather shocking...and shouldn't be possible at those speeds...

There's gotta be something else at play, here. Going from 90 fps to 200 fps is basically meaning your 2500k @ 4200 has the same horsepower as a Q9650 @ 8 ghz!
And that DEFINITELY is impossible.
 
Falkentyne said:
Hmm, very interesting.
It would be better if both were run at the same mhz, though. But I'm sort of shocked at the FPS being *DOUBLE*. Because in superPi, going from a Yorkfield to a 2600k (both at 4 ghz) gives about 30-35% higher fps in Morrowind and Vietcong, and seems to relate directly to SuperPi: superpi 1m (11.7 to 9.406; I estimated that a 2600k @ 3.5 ghz would give 11.7 1M, and a Q9650 @ 5 ghz would give around 9.4 1M), and in Vietcong (Intro level), which is purely cpu limited and a rather old game (enabling supersampling has no effect on FPS), I got about 35% higher FPS. And at 5 ghz, my FPS wasn't "quite" double (I think the minimum FPS was 53 (QX9650 @ 4 ghz) to 100 or 98 (2600k @ 5 ghz). This is a single threaded game, though.

Yet you're getting more than double the FPS in serious sam 2. That's rather shocking...and shouldn't be possible at those speeds...

There's gotta be something else at play, here. Going from 90 fps to 200 fps is basically meaning your 2500k @ 4200 has the same horsepower as a Q9650 @ 8 ghz!
And that DEFINITELY is impossible.
Click to expand...

the difference does get smaller when he increases the resolution. Im having a shocking feeling too when I look at the screen shots. But could it be this game benefits THIS much during multiplayer?
 
woot, sweet or something... I cant wait to upgrade my C2D rig with one of those SB cpus and mobo
 
Could very well be something related to memory bandwidth. It can't be raw CPU horsepower by itself, because if it was, you would be getting sub 8 second spi 1m scores at 2500k@4 ghz, which isn't possible even if you gave your left kidney for it...

I like using the vietcong intro level, because when Vietcong first came out, that level was the laggiest level ever; high end P4's could barely break 30 fps on the landing pad (starting spot), so that was a good indication of it being heavy cpu limited.

The GPU usage you listed, however, is rather perplexing. Going from 60% to 100% gpu usage plus the difference in cpu horsepower can give you the double FPS, but I'm trying to figure out how that is possible. I don't know the GPU usage in vietcong, but in a game that old, it's probably extremely low. Going from 8X FSAA to 24x FSAA (edge detect)+supersampling gives a 0 FPS drop...so the GPU is totally out of the picture here (but I don't know the usage %).
 
Last edited:
I saw a huge upgrade going from my Q9550 @ 3.6GHz to my i7 2600k @ 4.4GHz in BC2.
 
Good effort but this post wasn't done with concrete facts or standardized testing. You didn't capture frames in the same identical spots, and theres other variables that might be at play here. But good job nonetheless. :eek: Definitely put the sandy at 4 ghz as well, why clock it 200 mhz higher when it's already a superior platform/cpu?
 
Threads like this make hard for me not to upgrade my aging Phenom II 940BE build:p
 
[edit] I thought you were cheating, but I see now it looks like you were playing in the same server on each computer, and using the same class...correct? Impressive numbers in that case.
 
Last edited:
Quick question: How come when I see people posting Task Manager shots, not one single person seems to select "Show Kernel Times?"
 
devil22 said:
[edit] I thought you were cheating, but I see now it looks like you were playing in the same server on each computer, and using the same class...correct? Impressive numbers in that case.
Click to expand...

Yeah I'm guessing that the character on the "left" of him in one of the screenshots, that is holding the same gun, is the other system.

And I got mixed up, thinking "SS2" meant serious sam 2. oops.

And, the lack of anisotropic filtering in those screenshots makes my eyes bleed....
 
DualOwn said:
i've been saying this about BC2 for a long time.

in sp there is not much difference, but mp is a whole different story.
Click to expand...

+1. I love how people post benchmark for bc2 and say a dual core is fine. Benchmark ar done in SP. MP is a whole nother ballgame.
 
I think the best way to do this would be exact ram exact ghz and exct graphic, but one thing is you can face one way and have a better FPS, so that second system beside the guy might run less with more on the screen, really who knows only way to know is to have the positioned excatly in the same spot.

What I would have done is have one system not run beside the same side, one might hinder with the guy ahead of him or something different in the field of view.

Why not if you run both systems have the other guy log out in the same spot you test this will most likely give you a better FPS vs eatchother, but if stuff moves and etc kinda hard.
 
Unsurper said:
I think the best way to do this would be exact ram exact ghz and exct graphic, but one thing is you can face one way and have a better FPS, so that second system beside the guy might run less with more on the screen, really who knows only way to know is to have the positioned excatly in the same spot.

What I would have done is have one system not run beside the same side, one might hinder with the guy ahead of him or something different in the field of view.

Why not if you run both systems have the other guy log out in the same spot you test this will most likely give you a better FPS vs eatchother, but if stuff moves and etc kinda hard.
Click to expand...

The first couple of screen shots have the guy in view on the Sandy Bridge system, and it's faster, while the C2Q is looking at no guy just the background in the same spot and has half the frame rate. That's probably as good as it's going to get in a multiplayer benchmark of this game, since it doesn't have demo recording/playback. And if the Sandy overclocks better, then that's fair imo. difference is only 200mhz anyway, i seriously doubt the C2Q would even get 5 more FPS from 200mhz more overclock. Graphics card was the same in each system, and since in task manager, you can see that neither system is even close to 3GBs usage, then the 3GBs vs 4GBs memory probably makes no difference.
 
he could turn on fraps and log the average FPS, would give a far better ideam but you cant bench multiplayer, SP maybe.
 
MrGuvernment said:
he could turn on fraps and log the average FPS, would give a far better ideam but you cant bench multiplayer, SP maybe.
Click to expand...

The thing is WE say we cannot bench multiplayer....but why not? When we purchase these parts to play online how else are we supposed to judge it. If he were to play a couple of rounds and then record the average. And then play a couple of other rounds on the other computer and do the same.

whenever we buy the parts and play multiplayer rounds, we are never guaranteed the same performance anyway. So lets allow for the error if he is willing to test it out. This would atleast give us an idea.
 
Ground looks terrible in all those shots... really, really terrible. That's somewhere we really need to improve still, regarding FPS games. Looks like bumpy brown plastic.
 
Who plays at that resolution?

Here's a idea instead of oc the crap out of the cpu how about putting more load on the gpu? A dosh of AA maybe. Disable vsync its junk.
 
Last edited:
exlink said:
It eliminates GPU bottlenecks and allows the CPU's to work to their limits. Thats why.
Click to expand...

if your gpu is the bottleneck then it means 1 thing UPGRADE.
No use trying to let a cpu make up for it. that's why we buy gpus ain't it?

But hey he's playing with a 5870 how can it bottleneck?
 
Spooony said:
if your gpu is the bottleneck then it means 1 thing UPGRADE.
No use trying to let a cpu make up for it. that's why we buy gpus ain't it?

But hey he's playing with a 5870 how can it bottleneck?
Click to expand...

apparently it just does. Look at the afterburner pics. The videocard in the core 2 system runs at lower usage than the one in the i7.
 
I'm happy to see the results. Something I've been saying for a long time. Thats a huge increase in performance and I've seen the same coming from Q9550 to i7 860 then from the 860 to this. There is definitely a difference in gaming from upgrading your CPU!


Now - Beware the haters will be out to call you a liar soon enough. They will accuse you of doing something to sway the numbers or other ridiculous shit. I believe they are jealous and just cant afford to upgrade so they will start calling you names. Happens every time.
 
Spooony said:
But hey he's playing with a 5870 how can it bottleneck?
Click to expand...

Likewise there is also a point when a CPU can no longer feed enough data to the GPU so that the GPU can run @ 100 Percent possible performance.

I think this guys test proves just that.

His Graphics card could not be pushed to 100 percent with a Q9650 in at least BC2 (maybe other games as well) it showed 60 percent usage, but with the 2500k he can now push his graphics card to the 99-100 percent limit in BC2 (maybe other games/newer games as well). See his GPU was performing at almost half of it's potential because of the Q9650 CPU being far outclassed with his 5870 GPU, so upgrading his CPU to 2500k basically gave him Crossfire scaling on his GPU because he gimped the heck out of it with that CPU (the Q9650)

I know when I went from 775 dual core to AM3 Quad Core my same 260GTX GPU performed ALOT smoother/faster/better. I didn't use the Riva Tuner to check but it felt like my GPU was @ 70 percent usage in TF2 with the 775 Dual Core, now it feels like it's @ 100 percent with the AM3 Quad Core.

The moral of the story is don't ever gimp a NEW graphics card with an OLDER CPU (if you game FPS on servers with big loads 20-30+ people etc) even if you think it's a great CPU in sum benchmark like Super Pi/Hyper Pi. Super Pi benchmarks don't tell you how well a game runs, especially on a big server.

That's my opinion. ;)
 
2wiced said:
apparently it just does. Look at the afterburner pics. The videocard in the core 2 system runs at lower usage than the one in the i7.
Click to expand...

oh ill look now its almost 6am here too lazy to get up and switch the pc on on ze phone at the moment.
But don't you mean the cpu is bottlenecking if the gpu is doing less work?
 
Spooony said:
oh ill look now its almost 6am here too lazy to get up and switch the pc on on ze phone at the moment.
But don't you mean the cpu is bottlenecking if the gpu is doing less work?
Click to expand...

Yea.

If the GPU usage is low, then the CPU is bottlenecking it.
 
Very cool results. Now you need to update your sig. This makes me very excited for Ivy Bridge.
 
You must log in or register to reply here.
Back
Top