NP 9262 9800M GT Sli Crysis Bencmarks

This is the only thing I can find for now but I am working on it.Hold on

 

Reran at settings you asked because I cant find the Benchmark tool in the actual game,look I know most people do not believe or trust these numbers but It Is what It Is.I have NO reason to post false info.These test were done using a Crysis benchmark tool I found on the web at www.guru3d.com. If anybody has a better benchmark for this game let me know, I will run it at any settings.
Overall this game runs very smooth with almost no slowdown even on big firefight scenes.I fell like I have to defend myself from all you non-believers.Look I was also surprised on some of the results, but THESE ARE the results.Take them any way you want.The game is very playable on my system on high settings and I am very,very satisfied with my purchase.

 

i am considering a reinstall, i do have 2 drives

 

Are they both the same size and speed?You have to set up RAID first before you install the OS.If I can do it anybody can, very easy.If you need help let me know!

 

I would not sweat it. You have some incredible numbers and I am happy you are taking the time to post them. You have been more than upfront with your info and I appreciate.

Now, once I start benching mine tommorrow, I am hesitant to post results, as I may be getting called on the mat like you are currently. aaah. Heck, I'm posting em. If people dont believe em....meh..

 

My two drives are completely different. I wanted a bigger 320GB drive for storage. The single drive I use for my OS is 200GB 7.2k RPM. Your numbers are destroying mine.

IKAS V, in the Guru3D tool you're using to benchmark, which "Timedemo" are you using. If you are using the default one, that is just a rendered video and then a playable level. Which folder is the looping island in? Is that the one you're using?

 

I think they at least have to be the same speed.If they are you can install both hard drives, but if you use Raid 0 they will be only recognized as the 200GB HD, so total you would only have 400GB with those two HD's.That's one reason I picked them all to be the same speed and size to have no problem if I wanted to change the RAID levels.

 

About the Crysis benchmarking, though. Whenever I choose the island benchmark_gpu demo, it gives me the intro and a demo level. It doesn't loop or anything. I can never get any results. Can you help me out?

 

No, please post them, especially since you'll be testing the 9800 GTX which we're all curious about. But, please also remember that what you post serves to provide guidance to others who are contemplating what to buy (why else would you post them), so it's important the numbers aren't misleading. No one disputes IKAS V's 3DMark06 scores or dx9 Crysis scores because they are consistant with what we've seen before. But, he might be the first person in the world to claim that Crysis runs nearly twice as fast in dx10 as it does in dx9 and that has naturally attracted attention.

 

Oh I will. I am as curious as everyone. Just for fun, while I wait for my lappy to arrive tomm, I ran the crysis_gpu demo (from bin64), on my desktop. The desktop is an E6600 (2.4 c2), 4gb, 8800 GTX 768mb (which I thought was zippy). Here is what I got:

TimeDemo Play Started , (Total Frames: 2000, Recorded Time: 111.86s)
!TimeDemo Run 0 Finished.
Play Time: 143.40s, Average FPS: 13.95
Min FPS: 9.49 at frame 1942, Max FPS: 16.44 at frame 1757
Average Tri/Sec: -5621053, Tri/Frame: -403027
Recorded/Played Tris ratio: -2.27
!TimeDemo Run 1 Finished.
Play Time: 140.31s, Average FPS: 14.25
Min FPS: 8.78 at frame 1943, Max FPS: 16.51 at frame 1750
Average Tri/Sec: -5561342, Tri/Frame: -390161
Recorded/Played Tris ratio: -2.35
!TimeDemo Run 2 Finished.
Play Time: 140.49s, Average FPS: 14.24
Min FPS: 8.62 at frame 1936, Max FPS: 16.51 at frame 1750
Average Tri/Sec: -5531236, Tri/Frame: -388536
Recorded/Played Tris ratio: -2.36
!TimeDemo Run 3 Finished.
Play Time: 140.26s, Average FPS: 14.26
Min FPS: 8.55 at frame 1940, Max FPS: 16.51 at frame 1750
Average Tri/Sec: -5556182, Tri/Frame: -389642
Recorded/Played Tris ratio: -2.35
TimeDemo Play Ended, (4 Runs Performed)

IKASs machine spanks my desktop...

 

The quad-core and the SLI 9800M GT's are what give the combo.

 

Another thing that I was wondering. I see benches in Dx9 and Dx10. How do you get to specify which Dx version to use? I am sure I have both installed. Total nuub here. Can you tell crysis which Dx to use? or is this handled somewhere in a control panel applet?

 

I don't know how to change this either, when using the Crytek batch file benchmark. I do not know how to get the Guru3D benchmark tool to get the same demo as Crytek's batch gpu demo. IKAS V?

 

If you use the 3rd party benchmark you can set either dx9 or dx10 and modify all other settings just before you run it. If you use the in-game benchmark you first have to run the game, choose all the settings you want, then close the game and run the benchmark. That's why the 3rd party benchmark was made... so you could simply do all your settings and then run the benchmark. Either way the results should be the same.

What were the settings for your Crysis scores... dx9/dx10, resolution, quality settings... ? I believe the desktop 8800 GTX is zippy too, so I don't understand those low scores.

The 9262 with a single 8800M GTX gets 17-19 FPS in the Crysis benchmark at 1920x1200 high settings depending on the OS (CPU speed doesn't make a huge difference). We're all hoping the 9800M GTX beats that by a big margin.

 

This what I do.First you pick weather you want to run your demos in DX9 or DX10.Then pick what OS you are running 32 bit or 64 bit(I use 64 bit) then under add runs I pick the resolution(1900x1200)then right under you pick what AA level you want or if you want any, then you click on the arrow next to those to boxes to load it in the box that says( Queue) follow the same steps to load the different levels you want to bench mark your run in (Yes all my demos were run on the Island level)
Click Begin Benchmarking and then wait for the results.
Simple.

 

-this is where the gpu file should be. the only difference with you is you have 64 bit so it may be under the bin64 folder.-

the problem with that 3rd party program is that it may not keep the correct settings during tests. i experienced this myself and decided to stick with the internal benchmark tool which forces you to run the game and make video settings before you run the test. this garuantees an accurate run.

.

 

Dammit!
I have gone through and looked at all 35 files in Crysis and not one says GPU benchmark or anything else like that on it.Maybe somone out there knows were the file is on Vist 64?These are what my files are.

 

Here is what my crysis install folder looks like:

 

you are browsing the game disc!

browse where you instsalled the game on the hard drive.

 

Here's an excellent comparison of dx9 and dx10 in Crysis: http://www.gamespot.com/features/6182140/index.html

The author uses a single desktop 8800 GTX, but, since it's almost as fast as our mobile SLI rigs, the information should be valuable to us.

Note that, in the graph on page two, he shows that dx9 (XP) and dx10 (Vista) run the game at roughly the same speed in low, medium, and high settings.

 

Thank you IKAS for your efforts. Unilike some people might think his results aren't at all unbelievable (otherwise they have to justify it with data other then 3DMark06).

Indeed the 9800M GT in SLI is more powerfull then the 8800M GTX in SLI. It may be due to drivers are it may be due to HW changes but they are about 10 FPS faster (33%) in Crysis, and people are reporting in general. The 9800M GT is not just a rebrand of the 8800M GTX.

Who says otherwise are using the 3Mark06 as the base for comparing performance. Indeed the 3DMark06 is worth next to nothing to measure the performance of these new cards. I already had mentioned it several times in other posts. So it is indeed unbelieveble that the experts don't know better already.

So basically we might be getting in reality far more marks then the 3DMark06 actually states. Let's see what kind of a monster the 9800M GTX is.~

If 3DMark06 was worth anything I could play Crysis almost as well as one 8800M GTX. I get 10600 Marks with 2x8700M GT in SLI ... but I cannot (I don't play Crysis).

Trance
PS: I would like to see some VERY HIGH marks. Also the same tests with just one 9800M GT working, that would be dandy (just the crysis tests) as it would show the benefits of SLI (or lack of it).

 

I don't know how you came to this conclusion, at least from the information in this thread. Carefully compare IKAS's dx9 results with those of Argh's. Discounting Argh's XP advantage, they're the same numbers!

What am I missing?

 

Well last time I've read with DX10 IKA's (Vista 64x) is trashing ARGHs DX10 (XP hacked) score by 20 FPS and the same for ARGHs DX9 but by 10. We all know also that Crysis was mostly done for DX10 rather then DX9.
You may just focus your comparison over what you see fit to make it similar, that is DX9. Anyway make your pick, either is IKA's Vista x64 is bogging down DX9 (who cares) or is XP bogging down the DX10 hack.

But I agree with you that the only way one can really compare scores with the data in this thread is when someone with the 8800M GTX in SLi and Vista posts some scores with no OC as IKA seams to be doing. Meanwhile here are some with the DELL XPS M1730 and Vista (with no overclocking):

http://www.notebookforums.com/thread211727.html

If the CPU doesn't play much in Crysis you should get more or less the same scores.

I take that you guys have chosen XP to play game due to better FPSs, even against DX10 in Vista, I know I did. So I'm puzzled about you saying that ARGHs has the same scores as IKA's. For me they are totally different, so much that people argue that something is not right.

I think people just assumed that the 9800M GT was basically repacked 8800M GTX due to 3DMark06 scores ... or because someone said so. This user data points to a different reality, but it is still arguable.

Anyway I think we all agree that we can only compare when the same OS (preferably Vista since these new cards are really made for this OS and DX10)
Trance.

 

The Dell M1730 has skewed results since it has a PhysX.... which has shown to skew results in such basic benchmarks as 3DMark06 (for the CPU score).

So the PPU can effectively make up for some CPU calculations in games and benchmarks.

If one were to disable (or even remove) the PPU and run the tests again, then it would be more interesting.

 

@Gophn,

Interesting, do you know if it has the same effect to the Time Demo? I'm simply looking at average FPS counts calculated by Crysis Time Demo Benchmark Tool.

I always assumed that we the Clevo crowd went for XP due to better FPS even in Crysis. As far as I know this is the first time someone posts much better results in Vista in terms of FPSs (never mind if it is DX10 or DX9). This since it happened the 3DMark06 score to be better in Vista then in XP, at least that is my case. So I'm inclined to think that there is some merit on the 9800M GT card series that wasn't on the previous series.

Trance

 

The 9800M GT is the near same as the 8800M GTX, but slightly modded core, with higher power rating... and coupled with latest Nvidia drivers for PhysX support.

There would be a slight difference between the two cards in and XP or Vista environment.

But there has to be a controlled setting to do the tests with the same system.

Only the Clevo M570RU/TU or D901C can do such a test with the two different videocards while keeping the other specs the same.

 

Agreed. Waiting for those scores in a Controlled Setting for an effective comparison (same everything but the video cards). Maybe will surprise us.

Trance
PS: What is a slightly modded/changed core? Changed to achieve what? Why more power? The word on the streets is that the 9800M GT is a repacked 8800M GTX.

 

Didn't Justin show that the 9800m gt score the same in crysis and dmc4 as the 8800m gtx? His scores don't just go up a little bit either in dx10 Its a increase of almost 20fps, not only that but his scores almost match some 9800gx2 setups.

 

My thoughts exactly. That is why I asked IKAS to post some scores with only card activated.

IKAS scores in DX10 are indeed fab and never seen before on a 8800M GTX in SLI. At least not posted on this forum.

Trance

 

The 9800M GT core has a higher power rating because it will be the base for the 9800M GTX with 16 more shaders and higher clocks.... which will need the higher power rating.

I think we should look at the tests for the graphics scores, not the CPU scores... in both the 3DMark06 and 3DMark Vantage.

 

Disagree, 3DMark06 is flawed. Otherwise how can you explain me getting the score of 10600 (OCd) with 2x8700M GTX in Vista and get better FPS in XP with only a score 9300? Simply does not work since with that score I should be able to play Crysis as good as any single 8800M GTX and well I can't. Indeed I get lousy FPS counts using Crysis Time Demo in comparison.

I think we sould keep focusing on Crysis Benchmark and FPS counts. This seams to be closer to reality. Furhermore Far Cry 2 is coming etc etc. For some reason even Anandtech just used 3DMark06 for decoration in the head to head contest.

As for 3DMark Vantage there is very little samples of data to be considered reliable.

Trance
PS: My bottom line is today, any system that playss Crysis well will play any other game even better.

 

ARGH, of course, wasn't running dx10 because he has XP. He simply used the hack so he could use the "very high" settings, and he posted his results. If you look carefully you'll notice that IKAS's dx10 runs were at "high" quality settings, not "very high", as he said at the beginning of his post. ("This test was done at all high quality settings,used stock drivers 176.02."). Even so, his numbers are incredibly high. If would be nice to believe that dx10 is that powerful, but, there are so many carefully managed comparisons out there that say otherwise, it's hard not to be skeptical.

I looked at the interesting link you provided, and perhaps the results there might indicate what is going on. With dx10, 1900x1200, and very high settings, the poster got an average FPS of 14.42. While it's not quite an apples-to-apples comparison, ARGH with the dx9 hack, 19 20x1200, and very high settings got an average of 21.44 FPS. When this poster switched to high settings, however, he got an incredible jump to 38.31 FPS, while ARGH, running dx9, only got 30.28, a typical increase that is confirmed in many Crysis tests. (See: http://www.extremetech.com/article2/0,2845,2222416,00.asp )

ARGH ran his tests using the Crytek benchmark , but this poster and IKAS both used the third party benchmark. This makes me think that there might be a problem with this benchmark. All the results I've seen suggest that when this benchmark is set to dx10 and very high, the results are "right". But, when it is set to dx10 and high/medium/low, I suspect it may yield incorrect results.

It would be interesting, and helpful, IKAS, if you would try the Crytek benchmark and try either/both benchmarks on dx10/very high.

 

can you bench crysis with "Benchmark_CPU2.bat" please ?

 

I think that the Guru3D Crysis Benchmark Tool is just a front end for the test application used by Crytek and issued in command line using Benchmark_CPU2.bat. In other words, the core test application is the same, is just the user interface that changes, but I may be wrong. I don't have Crysis with me to verify this, I guess it is faster that IKEAS makes the test as suggested in order to dissipate some issues.

Let's be clear, an FPS is the number of framed rendered per second. It is doesn't matter what are the settings (HIGH, MEDIUM, LOW, ULTRA HIGH, VERY HIGH). So basically if one system has better FPS counts then the other, means that in that particular settings one system performs better then the other.

IKEAS you are doing a great job. People are not doubting your values, we are just researching why they are so different to what we expected, this based on ARGHs results and some previous posts.

So help us out mate

Trance
PS: Maybe Justin can help us out too.

 

crysis benchmark tool has only benchmark_cpu.bat but not benchmark_cpu2.bat.Benchmark cpu2 isn´t map island but lost paradise.

 

I'd be interested to see what the fps are at medium settings, I say that because his high setting fps are only 2-3 off from what others reported with medium settings. If there is no increase then we should know something is wrong.

 

guys i have been trying to tell you that the 3rd party program has a bug where it does not keep the game's settings properly. this is why IKAS should make a decent effort and help us out by running the built in benchmark tool that crytek supplies. i assure you once IKAS does this the performance should be on par with 8800m gtx card(s). what i am most curious at is to see if he does score better fps because he has a faster cpu by a whole 600 mhz and a faster FSB for that cpu.

bwheon, i will run the medium settings benchmarks for you now.

 

1920x1200 all MEDIUM dx9 = 55.20 fps avg. sli enabled

going to run low settings and see what that is now.

 

take into effect that vista and dx10 do not boost fps for this game and decrease them and that he has a faster processor, looks like I might be correct?

 

before i run the low settings i ran the medium settings again but this time i kept the "objects quality" to high and the rest to medium.

1920x1200 all medium (except objects quality to hig) dx9 = 48.38 fps avg. sli enabled

i noticed that when you select to place all settings back to medium it did not alter the objects quality and it remained high. maybe when i modified the cvar group to hack xp dx 10 effects it may have altered something in the game or maybe i simply forget how the game changes the settings (it's been a while since played it). what i do know is that i did indeed revert the game back to stock and ran the wuxga high settings to verify this.

 

ok low settings;

1920x1200 all low dx9 = 95.46 fps avg. sli enabled

 

In the absense of any answer from IKEAS can't ARGH run the tests with the same program that IKEAS did to compare results? I mean, one of the premisses is for both systems to use the same benchmark tool, and that can be complied by either.

Another suggestion is for now to focus just on HIGH settings (ALL HIGH). It is difficult to follow all those changes in settings and IMHO it does not make sense to test it on MEDIUM for now unless one wants to push the fps up. IKEAS pics suggests that he has set everything on HIGH.

Trance

 

ARGH's sig says he's running XP. If so he obviously can't run the dx10 tests. I can't either because I'm also using XP. I can confirm, though, that I get the same results with either benchmark with dx9.

 

Have you installed the 1.2 patch? When you do, you get the CPU2.bat file. I just noticed this today when I was installing.

Also, if you install 1.2 patch, you are also supposed to install the 1.2.1 hotfix.

See this site for downloads and details:
http://www.crymod.com/thread.php?threadid=20664

 

the 1.2.1 is just a multiplayer hotfix and it's not needed for single player mode.

 

So it looks like the Guru3d tool uses the testing core. IKEAS does not have XP, neither you or ARGH have Vista, so we need to get someone wth Vista to run the tests

Trance
PS: I don't think IKEAS can help us much more. It seams that he is using the Crysis demo and Guru3D test too because he doesn't have a full copy of the game.

 

I can run some test if u want guys. I have a full version of Crysis.

 

all high settings; 1920x1200, 1680x1050, 1440x900

use the built in Benchmark_GPU tool

 

You can run the Crytek benchmark from the demo as well.