Crysis GPU Bench (1280x1024 Very High 2x)

[Eddie3dfx]

Since this is probably the most looked at benchmark, I think it's important to find out what people are getting in vista 32 and 64 at 2x and 4x very high settings (1280x1024 resolution), single card or sli, also include what all your settings are and drivers)

Quad at 3520, 2 gigs of 1100mhz memory (5-5-5-15), 8800gt 753-1955-1021, 169.04 drivers

Crysis in Vista 32 - 1280x1024 with 2x antialiasing on, all very high settings

15.36fps avg


To those that are unsure of how to bench, go to C:\programfiles\Electronic Arts\Crytek\Crysis SP Demo\Bin 32 (Benchmark_GPU.Exe)

[disruptfam]

man i so won't be able to play this @ high settings...

How much diff will i get with a gtx @ 650/1120 compared to your gt in fps??

same setting's as you...

[Eddie3dfx]

With a gtx, probably close to 17-18fps.
At high you can play it with no problem. These are the max settings (very high) that people are drooling over and essentially what we want to get over 30fps. Add in a new penryn cpu, fast 1600+mhz ddr3 and maybe closer to 19-20fps.
It will take sli, driver tweaking, and a gpu doubling the performance of a gt, gtx to go passed 30fps, i think.
But in the meantime I would like to see what others are able to achieve so we can play with this

[disruptfam]

With a gtx, probably close to 17-18fps.
At high you can play it with no problem. These are the max settings (very high) that people are drooling over and essentially what we want to get over 30fps. Add in a new penryn cpu, fast 1600+mhz ddr3 and maybe closer to 19-20fps.
It will take sli, driver tweaking, and a gpu doubling the performance of a gt, gtx to go passed 30fps, i think.
But in the meantime I would like to see what others are able to achieve so we can play with this

cheers mate.... hopefully i can play on high no aa no af with my setup @ 30fps

[Kain665]

turn off aa and af and see teh fps double... seriously.

[XS Janus]

no real need for aa anyway in the game...

[Soulburner]

no real need for aa anyway in the game...

Completely untrue. This is one of the most jaggy infested games I have seen in a long time - even 2xAA does wonders for cleaning it up. Unfortunately with this engine we are going to have to take a large hit for that luxury.

[adamsleath]

i hate jaggies.

BUT I RECKON YOULL GET ANOTHER 10 FPS WITHOUT AA AT THESE SETTINGS ():

Crysis in Vista 32 - 1280x1024 with 2x antialiasing on, all very high settings

i think the eye candy of v.high settings is worth it at this point in time.

im getting 25-35 fps min (depending on the scenery/view) with modded dx9 "high" textures ('v.high')@ 1280x1024 0xaa 8xaf.......18-24fps with 2x or 4x AA at this res.@ these settings.

[Toysoldier]

Did some testing in DX9 with the 163.75 WHQL driver. All tests was done with an 8800 GTX running at stock speed with no aa and all settings at high.

DX9 32 Bit HighSpec HDR DevMode

1024 x 768
cpu : 47.21 FPS
gpu : 52.86 FPS

1280x1024
cpu : 35.12 FPS
gpu : 38.64 FPS with 2xAA ~ 33.42 FPS

DX9 32 Bit Custom HDR DevMode

Where custom means running with the following

r_UsePOM=1
r_sunshafts=1
e_water_ocean_fft=1
q_Renderer=3
r_colorgrading=1
r_UseEdgeAA=0

placed in an autoexec.cfg file located in Crysis\Game\Config directory which means that in advanced options shaders quality, postprocessing quality and water quality will all be listed as custom and not high.The above tweaks will give DX10 like graphics.

1024x768
cpu : 42.29 FPS
gpu : 46.93 FPS

1280x1024
cpu : 31.03 FPS
gpu : 33.81 FPS with 2xAA ~ 29.09 FPS


It would be interesting to see single vs dual vs quad core scores

[Eddie3dfx]

@ 1280x1024 and your config file, gpu batch (xp pro)
4x 29.89fps
2x 32.63fps
0x 39.85
I wasn't sure if you had 2x or nothing.

[skynet666]

Hmmm...

I think my fps seems a bit low for what Im using but anyway here are my results

using vista 32 bit and dx10 mode

All settings medium 1280 x 1024
Avg fps 59.44

All settings medium apart from textures, object detail and water quality high
Avg fps 41.56

All settings very high
Avg fps 11.09

All setting very high +2Xaa
Avg fps 8.61

[The0men]

Hmmm...

I think my fps seems a bit low for what Im using but anyway here are my results

using vista 32 bit and dx10 mode

All settings medium 1280 x 1024
Avg fps 59.44

All settings medium apart from textures, object detail and water quality high
Avg fps 41.56

All settings very high
Avg fps 11.09

All setting very high +2Xaa
Avg fps 8.61

It is because you only have 320 MBs on your card, this give a poor preformance score, you would find the 640 would be a bit closer to the GTX than to the 320mb GTS.

[Lestat]

Completely untrue. This is one of the most jaggy infested games I have seen in a long time - even 2xAA does wonders for cleaning it up. Unfortunately with this engine we are going to have to take a large hit for that luxury.

its only jaggy at resolutions below 1680x1050
anything above that the jaggies are hardly noticeable if at all

AF is different, higher resolutions dont fix the AF woes

i play 1920x1080 all on the max settings but with AA and AF off.
alot of explosions on screen are the only times i see any sloppy frame rates.

no game at 1920x1080 or higher needs AA people who claim they do need to pull their controller out of their anal retentive sphincter.
but jaggies are part of the doom/unreal/crytek engines because of the nature of their graphics/textures. they demand high resolutions to eliminate the jaggies.

[Eddie3dfx]

I tried the new xp beta drivers 169.09
@ (1280x1024 2x enabled) in xp all on high with that special batch file that was posted above that enabled some thigns with directx10 quality
It came out to 33.4fps
1fps improvement from 169.04

[xlink]

its only jaggy at resolutions below 1680x1050
anything above that the jaggies are hardly noticeable if at all

AF is different, higher resolutions dont fix the AF woes

i play 1920x1080 all on the max settings but with AA and AF off.
alot of explosions on screen are the only times i see any sloppy frame rates.

no game at 1920x1080 or higher needs AA people who claim they do need to pull their controller out of their anal retentive sphincter.
but jaggies are part of the doom/unreal/crytek engines because of the nature of their graphics/textures. they demand high resolutions to eliminate the jaggies.

or 8x AA...

but I'll largely agree with you. @ 1680*1050 I'll take mostly high/very high over AA anytime.

[Soulburner]

its only jaggy at resolutions below 1680x1050
anything above that the jaggies are hardly noticeable if at all

AF is different, higher resolutions dont fix the AF woes

i play 1920x1080 all on the max settings but with AA and AF off.
alot of explosions on screen are the only times i see any sloppy frame rates.

no game at 1920x1080 or higher needs AA people who claim they do need to pull their controller out of their anal retentive sphincter.
but jaggies are part of the doom/unreal/crytek engines because of the nature of their graphics/textures. they demand high resolutions to eliminate the jaggies.

Are you telling me what I'm seeing isn't real?

I see jaggies in Crysis @ 1920x1200. Lots of them. Not on the character models - those are superb. Look at the trees, grass, rocks, and virtually any other object in the game.

I was about to think you were still playing on a CRT, but I see you have a 24" LCD like myself in your sig, so you should very well know already that you still need anti-aliasing at high res with an LCD due to the increased amount of detail shown over a CRT. Us LCD guys don't have the benefit of an analog CRT fuzzing an image and creating a pseudo-anti-aliased image (and without a performance hit too). We see pixel for pixel the exact output of the video card - which means resolution alone can not and will not fix aliasing in these games.

[Eddie3dfx]

<I will try this in vista and report back on my original 15.36fps avg since this is the most important score with all settings enabled>

So I tried it in vista a few times and these drivers keep crashing on me, no luck so far.

Tomshardware reports that with a gtx, quad at 2.66 (1600x1200 4x, he is mustering out only 10.1fps.

Anyone with Sli with gt, gtx, ultra, or crossfire 3850/3870 do some benches in vista with very high settings and 0x, 2x, 4x?
Thanks

[Eastcoasthandle]

This will make benchmarking easier

[Computerville]

Damn, talk about low FPS for those "pretty usually" settings...
Everyone are running on at least 1280x1024 and with AF/AA today, it´s an awful way to go on those FPS, and think about that 8800Ultra in SLI is NOT ENOUGHT for good FPS when playing on the today usually settings...

[cadaveca]

8800U is enough...after benching for many hours, it seems that this game is truly cpu-limited.

I remember hearing that this game would use a quadcore...it doesn't. Single thread avg 100&#37; cpu usage regardless of speed(2.4ghz all the way to 3.8ghz, Core2Duo, or Core2Quad).


Funny thing tho...if you've had an ATI X-series running FAH before, you've seen high cpu usage, but not high cpu temps..this game behaves the same. It truly seems that texture resolution is far to high to fit within vidmem of either 8800U or 2900XT, and so cpu spends ALOT of time shuffling textures, etc to gpu...

I have many issues running this bench proggie, BTW. it will not close crysis thread properly, so I can only run one set of benches at atime...very frustrating. I'd love to just load up a bunch of different settings, hit "BENCHMARK", and walk away and come back later to see results, but this is not possible ATM. Demo or retail version does not matter.

otherwise, this tool is really good!

[Eddie3dfx]

Here is teamaqua's post on how to get some of the directx10 features in directx9 xp
http://www.xtremesystems.org/forums/...d.php?t=163687

I also read an article on crysis and benchmarks online yesterday, but forgot to post the site. I'll search again for it and post the link, meanwhile here are the stats they came up with.

1280x1024 very high, no x enabled
Gtx 20.3fps
Gtx Sli 25.4fps

We are getting very close to 30fps.
Has anyone with Ultra Sli/ddr3 been able to get above 30fps?

[Fraggle]

Aliasing is not a problem caused by low resolutions as such, what matters is the actual size of a pixel on the screen which is related to the resolution and screen size. A 28" monitor needs twice the resolution of a 14" monitor just to get the same physical resolution on the screen.

[Eastcoasthandle]

If you add:
con_restricted = 0
r_DisplayInfo = 1
to benchmark_gpu.cfg or benchmark_cpu.cfg you will notice some disparities between what the fps is in the recorded benchmark vs what the display info provides. It makes me wonder which is correct. From my observations the display info (displayed using white lettering) is on par with Fraps while the recorded benchmark (displayed using green lettering) is different from both display info and fraps.

Anyone else notice this?

[cadaveca]

Yep, I mentioned this in the patch thread. All reviews posting Crysis numbers are wrong, pretty much. The difference between real FPS and bench dispaly varies from 45-75&#37; actual FPS, partially dependant on CPU load, it seems. Not confirmed cpu bit tho.

[ak_47_boy]

So pc preformance has to quadruple for me to get good fps at qxga.