Crysis Benchmarking - Methodology and Mythbusting

[jas420221]

First, let me say excellent post about benchmarking Crysis!!!! Great job!

But I have a problem with this statement below.

Today on market there are NO multicore optimized games! There are only dualcore optimized titles, and Cryris is one of them! Only use of multicore CPU's with this games is that you'll be able to run some parallel tasks (preferably on another HDD), during the gaming session!

Doesnt Supreme Commander use 4 cores? Lost Planet? Also, I have seen Crysis use 4 cores as well. They were not all maxed out, but there was significant CPU usage on cores 3 and 4.

There are a few games out already that can utilize more than 2 cores.

Let me see if I can find those threads.
Some Valve games do

I guess the key word is OPTIMIZED. There are a few games that can use more than 2 cores, but none are OPTIMIZED to do so.

[Governator]

Test was done with 169.28 driver set & no graphic card oc'ing...

Beginning Run #1 on Map-island, Demo-benchmark_gpu
DX9 1680x1050, AA=16x, Vsync=Disabled, 32 bit test, FullScreen
Demo Loops=3, Time Of Day= 9
Global Game Quality: High
================================================== ==========
==
TimeDemo Play Started , (Total Frames: 2000, Recorded Time: 111.86s)
!TimeDemo Run 0 Finished.
Play Time: 70.37s, Average FPS: 28.42
Min FPS: 18.33 at frame 1960, Max FPS: 38.77 at frame 974
Average Tri/Sec: -19611028, Tri/Frame: -690041
Recorded/Played Tris ratio: -1.33
!TimeDemo Run 1 Finished.
Play Time: 66.23s, Average FPS: 30.20
Min FPS: 17.84 at frame 1936, Max FPS: 41.83 at frame 91
Average Tri/Sec: -20788814, Tri/Frame: -688420
Recorded/Played Tris ratio: -1.33
!TimeDemo Run 2 Finished.
Play Time: 65.67s, Average FPS: 30.46
Min FPS: 17.49 at frame 1943, Max FPS: 43.32 at frame 90
Average Tri/Sec: -20983054, Tri/Frame: -688942
Recorded/Played Tris ratio: -1.33
TimeDemo Play Ended, (3 Runs Performed)
================================================== ==========
==

Completed All Tests

<><><><><><><><><><><><><>>--SUMMARY--<<><><><><><><><><><><><><>

2/11/2008 11:37:56 AM - XP
Run #1- DX9 1680x1050 AA=16x, 32 bit test, Quality: High ~~ Overall Average FPS: 30.33

************************************************** ************************************************

The, for kicks tried the 171.16 drivers, needless to say pretty ugly...

Update: Turned out SLI was disabled, though I could've swore it was enabled but anyhow below this is the result with it enabled ...

2/13/2008 9:27:24 AM - XP
Beginning Run #1 on Map-island, Demo-benchmark_gpu
DX9 1680x1050, AA=16x, Vsync=Disabled, 32 bit test, FullScreen
Demo Loops=3, Time Of Day= 9
Global Game Quality: High
================================================== ============
TimeDemo Play Started , (Total Frames: 2000, Recorded Time: 111.86s)
!TimeDemo Run 0 Finished.
Play Time: 101.13s, Average FPS: 19.78
Min FPS: 14.11 at frame 1952, Max FPS: 25.37 at frame 980
Average Tri/Sec: -18566552, Tri/Frame: -938784
Recorded/Played Tris ratio: -0.98
!TimeDemo Run 1 Finished.
Play Time: 99.21s, Average FPS: 20.16
Min FPS: 14.11 at frame 1952, Max FPS: 25.37 at frame 980
Average Tri/Sec: -18694888, Tri/Frame: -927371
Recorded/Played Tris ratio: -0.99
!TimeDemo Run 2 Finished.
Play Time: 99.53s, Average FPS: 20.09
Min FPS: 13.27 at frame 1958, Max FPS: 25.37 at frame 980
Average Tri/Sec: -18635028, Tri/Frame: -927349
Recorded/Played Tris ratio: -0.99
TimeDemo Play Ended, (3 Runs Performed)
================================================== ============

Completed All Tests

<><><><><><><><><><><><><>>--SUMMARY--<<><><><><><><><><><><><><>

2/13/2008 9:27:24 AM - XP

Run #1- DX9 1680x1050 AA=16x, 32 bit test, Quality: High ~~ Overall Average FPS: 20.125

************************************************** *******************************************

SLI Enabled now...

2/13/2008 9:41:10 AM - XP
Beginning Run #1 on Map-island, Demo-benchmark_gpu
DX9 1680x1050, AA=16x, Vsync=Disabled, 32 bit test, FullScreen
Demo Loops=3, Time Of Day= 9
Global Game Quality: High
================================================== ============
TimeDemo Play Started , (Total Frames: 2000, Recorded Time: 111.86s)
!TimeDemo Run 0 Finished.
Play Time: 69.52s, Average FPS: 28.77
Min FPS: 18.82 at frame 156, Max FPS: 40.04 at frame 976
Average Tri/Sec: -19852152, Tri/Frame: -690026
Recorded/Played Tris ratio: -1.33
!TimeDemo Run 1 Finished.
Play Time: 64.88s, Average FPS: 30.83
Min FPS: 18.82 at frame 156, Max FPS: 42.52 at frame 82
Average Tri/Sec: -21208078, Tri/Frame: -687947
Recorded/Played Tris ratio: -1.33
!TimeDemo Run 2 Finished.
Play Time: 64.26s, Average FPS: 31.13
Min FPS: 17.47 at frame 1953, Max FPS: 43.98 at frame 87
Average Tri/Sec: -21397594, Tri/Frame: -687458
Recorded/Played Tris ratio: -1.33
TimeDemo Play Ended, (3 Runs Performed)
================================================== ============

Completed All Tests

<><><><><><><><><><><><><>>--SUMMARY--<<><><><><><><><><><><><><>

2/13/2008 9:41:10 AM - XP

Run #1- DX9 1680x1050 AA=16x, 32 bit test, Quality: High ~~ Overall Average FPS: 30.98

[Nedjo]

First, let me say excellent post about benchmarking Crysis!!!! Great job!

But I have a problem with this statement below.
Doesnt Supreme Commander use 4 cores? Lost Planet? Also, I have seen Crysis use 4 cores as well. They were not all maxed out, but there was significant CPU usage on cores 3 and 4.

There are a few games out already that can utilize more than 2 cores.

Let me see if I can find those threads.
Some Valve games do

I guess the key word is OPTIMIZED. There are a few games that can use more than 2 cores, but none are OPTIMIZED to do so.

I'm glad you like this thread.
I haven't tried Superme commander, so I might be wrong on it. Will try LP. Regarding crysis, my testing with 2/3/4 cores don't show any difference, so I'm inclined to think that mulitcore support in this game is just marketing hype.
Regarding Valve's games there are no confirmation in benchmarking that they're multithreaded. hopefully I'll get Orange Box soon to check out...

[KTE]

First, let me say excellent post about benchmarking Crysis!!!! Great job!

But I have a problem with this statement below.
Doesnt Supreme Commander use 4 cores?

You're right. Some games can use more than two cores, but it has minimal benefit over a dual core at the same MHz, since the amount of the 2+ cores they use is minimal with very little optimization to effectively make use of its resources, hence dual core actually tends to lambaste quad/octa overall.

Check out this SupCom benchmark: http://techreport.com/articles.x/14052/7



The 9775 is SkullTrail with 8x 3.2GHz cores. Those results look ridiculous considering the clockspeeds and the number of cores, but I think it does show how SupCom isn't actually multi-core optimized at all, although it does sparingly use more than 2-cores.

And if you look at Crysis with Penryn dual-core (6.32GHz) Vs quad-core (10GHz): http://www.xbitlabs.com/articles/cpu...0_5.html#sect0



That shows clockspeed on two cores in games rules rather than >3 cores. Factor in that E8500 clocks much higher than Q9300 or Q9450.

I guess the key word is OPTIMIZED. There are a few games that can use more than 2 cores, but none are OPTIMIZED to do so.

Yup, I have to agree. If any software can make complete and effective use of each architecture, loading separate libraries for different processors archs, you would see the real power of each core utilized. Right now, you don't as coding is generalized.

Great work Nedjo! Now do something, turn off one core and then two cores using Windows using the same benchmark and see what results you get, if any different.

[Nedjo]

Great work Nedjo! Now do something, turn off one core and then two cores using Windows using the same benchmark and see what results you get, if any different.

You think it'll differ from core disabling with in BIOS?