Sager 9262 8800 SLI and CF SSD Benchmark Results

Ok, so my new Sager 9262 arrived yesterday so I have been busy trying to get some quick benchmarks.

My Laptop consists of:
Core 2 Quad Q6700 (2.66 GHz)
2 x Nvidia GeForce Go 8800M GTX w/512 (in SLI)
4096MB DDR2 800 RAM
17' WSXGA+ (1680x1050) Glossy Screen

The first thing I wanted to do was benchmark my 8800M GTX cards in both SLI and non-SLI as shipped from Sager. So I booted the machine and only did the necessary steps to log into Windows Vista Home Premium x64 edition. I installed 3DMark06 and immediatly benchmarked it.

As shipped from the vendor I got the following scores:
Installed Driver Version 7.15.11.7433 Not WHQL
Results in SLI: 13003


Results with NO-SLI: 10171


So once I had these resultes, I then removed the shipped HD and added my new HD setup.

I installed 2 CF to SATA HD adapters along with 2 RiData 16GB CF cards rated at 233x speed (Read 30MB/s Write 17MB/s each tested by me). I placed the two CF cards in RAID 0 and install my OS of Windows Server 2008 64 bit (Sorry, for those looking for Vista X64 results you will have to look elsewhere. My results should be close to those of Vista X64 SP1.)

From the clean installed Nvidia Driver Version 7.15.11.6762 (avaliable from Clevo) and got the following results.
Results in SLI: 13586


No SLI: 9991


Please note that I did not tweak the system or drivers for performance. These are the stock drivers at the stock settings. I am sure I can get higher results but wanted people to see what they could expect with no work on their part.


So on to the CF SSD HD testing...
Let me start by saying there is a nasty bug with the current Intel Matrix Storage BIOS that pauses the machine for 1.5 minutes PER CF card because they do not support SMART (so every reboot I have to wait 3 minutes!). This has been resolved by Intel and I have asked Sager for a new BIOS to resolve this issue, but time will tell when it arrives.
You can look at this article if you want to know more about this issue (refrence # 2102039). This bug is VERY annoying.
UPDATE: Sager has provided a solution: "If you have non-Seagete hard drives, you need to set the SATA mode on AHCI. After the computer restarted, you set the SATA on RAID mode. You will not see the delay then." I have tested this and it appears to work.



But anyway, on to the testing.
So I have Windows Server 2008 Standard x64 edition installed on my 2 CF cards that are placed in RAID 0. I have an additional Hitachi HTS722020K9SA00 7200RPM HD (on of the fastest laptop drives) to be my storage drive.

Here are the results of CF SSD HD tests:
Without adding the Intel Drivers: CF SSD HD Tune Average read of 36.2MB/s


Without adding the Intel Drivers CF SSD ATTO Results:



With the Intel Drivers: CF SSD HD Tune Average read of 63.9MB/s


With the Intel Drivers CF SSD ATTO Results:


As you can see, the Intel Drivers add a huge amount of performance and make the system run as I believed it would. The funny thing is however, that while at times the system seems much more responsive (any time it needs to quickly read items) at other times it seems much more unresponsive. I guess the problem is that I don't know what this system is like on a HD alone, so I can't really judge it so well.

Here are some additional benchmarks for those who wish to see them.
Hitachi 200GB 7200 RPM HD: HD Tune Average read of 53.8MB/s


My old Laptop WD 80GB 7200 RPM HD (On my old laptop with XP 32 bit): HD Tune Average read of 34.4MB/s


If people would like I can try to get some Crysis and Bioshock FPS numbers as well as see what I can do with some software overclocking of the Nvidia cards.

I am also probably going to reinstall my OS onto the Hitachi HD to see how the machine feels compared to the CF SSD HD I have created.

 

I'd be very interested to actually see how the system performs on the Hitachi HDD. We are watching SSD performance closely, and even though your configuration gives something of an unfair advantage to the CF configuration (RAID 0 speeds versus a single HDD), I'm still interested to see how much faster the CF RAID array is versus a traditional drive.

 

Those are wicked results.I`m sure you could reach 15k easily with some mild OCing.
We expect gaming benchmarking also.

 

Nice Review!

but I think for your q6700 you should be getting higher scores than the m1730 in sli. but were getting par hmmmm x9000 vs q6700...

 

Is it possible that once we start to see quad-core games (Far Cry 2.. Alan Wake to name a couple) that even with nearly identical scores right now to the Dell that eventually having the Quad in the newer games would start to show an advantage against the dual-core despite the higher clock speed? The question is how much of an advantage that would be.

 

it's the fact that our GPU is underscoring in comparison. and not performing up to par at all. even when overclocked.

sure the quad can or maybe will make up the difference. but it shouldnt have to.. it should be an extra because our GPU's allready are great.

 

@wxkid23:

In professional applications that are able to utilize quad-core architecture, they seem to be reporting anywhere from a 30-40% speed increase over a faster processor. As time goes by and multi-threaded support improves, this ratio should improve, too. Right now, most games will work better on the faster processor, but we should start seeing some multi-threaded enhancements in the near future.

 

Ok, so I have done some more testing. I removed the two CF SSD HD setup in Raid 0 and instead installed it directly to the Hitachi HD.
Sadly, the machine performs much, much quicker. My boot time drops from 72 seconds (with the SSD) to 29 seconds (with the Hitachi). For some reason the SSD seems to sit for long periods without the HD accessing (according to the lights) during booting while the Hitachi does not have this issue.

Additionally, while running from the RAID 0 SSD setup my machine seems to "freeze" at times. I can replicate this while running the HD Tune read test and trying to use the machine. It will freeze a lot instead of just running everything slower. When I do the same test from the Hitachi HD, I don't get the freezing (yes, the HD Tune scoring drops, but the system continues to respond).

Sadly, overall I have decided to use the Hitachi HD as my primary. The shipped 5400 80GB HD that shipped with the machine as a secondary, and a single 16 GB CF card as a third HD.

I cannot recommend two CF cards in RAID 0 to make a SSD. I am not sure if it this CF adapter, the RAID BIOS, or the CF cards themselves that causes the solution to be less than optimal, but one or more of those items is the cause.

I will post gaming benchmarks later this week. (I just picked up Crysis today).

 

I don't see that.

In the 1730/9262 (STOCK) comparisions I've made, where "apples-to-apples" could be assured, the two systems seem to have roughly the same graphic scores in 3DMark06 (1280x800), with the 9262 having higher CPU scores when the quad is used. This seems to be the case in both SLI and non-SLI tests. Both systems also seem to be producing the same fps results in the Crysis benchmark.

I'll admit, though, that I've had difficulty finding many "apples-to-apples" comparisons, so, if you have found solid information to the contrary concerning "out-of-the-box" performance, it would be very valuable if you could point us to it.

 

I sjust spent 2 weeks doing research on this. with over 100 posts of information dedicated to this.

I have in fact proved that the CPU scores are higher in single vs single and sli vs sli.

and I have proved that the sm2.0 score is lower in single vs single and sli vs sli. and overclocked vs overclocked.

the quad score balances out the scores making them appear to be equal but take away the quad and assume we both have x9000 processors and then the score is lowered about 1000 points with the q6700 and 700 with the q6600.

there is lots of information to support this and documentation to prove it.

read the bad sli 8800 score 9262 thread and the 8800 sli thread. those contain many posts and accuracte results to compare.

 

I never re-wrote an Nvidia Bios before...
Looks like I might have to to hit 16k..

 

you would never be able to write a bios. let alone an nvidia one..

gimme a break dude.

the best people I know myself included takes hundreds of hours just to learn what 10 bytes in 1 of the perf tables function is...

I have been working on vbioses for years.. and Nobody ever has ever written their own bios for Mobile videocards or Normal video cards..

 

Like I said I never looked at one.
I've only modded and added code to MB Bioses

 

as I have to video bioses. Hope you can add code to unlock the multipliers in the
d901c.

 

I guess he didn't really meant to write the BIOS from scratch, but rather just modify it.

 

Like doing a haircut to a curly rabbit.

 

Just another update. I think the reason the CF card is performing so poorly in the 9262 is a bios issue. Even just leaving 1 CF card (unformatted, formatted, or any way) slows my whole machine down during boot and other functions.
I have tested this on a different machine and I don't see these problems. I think this goes back to the CF cards lack of SMART support and the Intel Matrix BIOS issue I posted earlier.
Perhaps there is still some hope for the CF cards to work correctly once we get an updated BIOS.

 

Swiftnc,

Great report... thanks.

Does anyone but me think it's weird that the CPU scores went up so much when SLI was used (10% in the second test)?

 

I was also wondering about those results. Say, how much you`d say a 128 GB kind of setup would cost?

 

are you using an sd or a cf??

because cf is slow compared to a hdd...

not the same.

 

These were CF Cards. 233x rated (30MB/s READ, 17.5 MB/s Write each as tested by me.) My goal was to use RAID 0 to bring up the speed while gaining the 0.3 access times.

 

I wanted to share my 3DMark06 Overclock results
For those who don't want to look to page 1 my stock results were:
Driver Version 7.15.11.6762
Core 500 Shader 1250 Memory 799
Results in SLI: 13586
No SLI: 9991

My OC results were:
Driver Version 7.15.11.6762
Core 600 Shader 1500 Memory 950
Results in SLI: 13964
No SLI: 11625

Notice the huge jump in without SLI but the small jump in SLI mode.
It is very interesting to see.
(I can post the screenshots if people want)

 

that sux crap!

LMAO!!!!!

Can't you use 174.31 from lv2g or 174.74????

sli is supported with that.

also. try crysis benchmark tool @ 1680x1050 and 1900x1200 all setting high and post what you get.

I am getting 11500 in xp wit 3d06

or that proves my point that there is a wall with clevo laptop bus and you can only get said performance and that's it.

even with 2 cards you can only get 13xxxx.....

that's why you get more with 1 card.....

so you know the OC makes a diff......

so looks like the bus might be very restricted and we can't get better with 2 cards or 1 for that matter....


or maybe it downclocks to stock with overclock in SLI???

that's why it shows no real increase with SLI.

 

meh....how aobut gaming benchmarks instead of 3dmarks? Can you show crysis fps results @ 1900x1200 all high?

 

He can´t he has a 1680x1050 screen. But at that res you will surely play Crysis very very good with that SLI setup

 

Is very high on everything but shaders on high achievable with 1680x1050?

 

I will aim to update my drivers and get some crysis numbers for you all tomorrow.

 

bhattsan definitely I can run that easily all Very High DX10 and Shaders on High and this at 1920x1200. So surely his Clevo can

 

Ok. Here are my Crysis Benchmarks with settings set to Crysis "High".

==============================================================
Single Card NO OC Driver 7.15.11.6762 dated 1/20/2008
TimeDemo Play Started , (Total Frames: 2000, Recorded Time: 111.86s)
!TimeDemo Run 0 Finished.
Play Time: 105.13s, Average FPS: 19.02
Min FPS: 10.99 at frame 139, Max FPS: 23.20 at frame 104
Average Tri/Sec: -17993100, Tri/Frame: -945848
Recorded/Played Tris ratio: -0.97
!TimeDemo Run 1 Finished.
Play Time: 100.02s, Average FPS: 20.00
Min FPS: 10.99 at frame 139, Max FPS: 24.58 at frame 62
Average Tri/Sec: -18678464, Tri/Frame: -934078
Recorded/Played Tris ratio: -0.98
!TimeDemo Run 2 Finished.
Play Time: 99.63s, Average FPS: 20.07
Min FPS: 10.99 at frame 139, Max FPS: 24.75 at frame 75
Average Tri/Sec: -18764536, Tri/Frame: -934782
Recorded/Played Tris ratio: -0.98
!TimeDemo Run 3 Finished.
Play Time: 99.55s, Average FPS: 20.09
Min FPS: 10.99 at frame 139, Max FPS: 24.75 at frame 75
Average Tri/Sec: -18783668, Tri/Frame: -934983
Recorded/Played Tris ratio: -0.98
TimeDemo Play Ended, (4 Runs Performed)
==============================================================
==============================================================
SLI NO OC Driver 7.15.11.6762 dated 1/20/2008
TimeDemo Play Started , (Total Frames: 2000, Recorded Time: 111.86s)
!TimeDemo Run 0 Finished.
Play Time: 71.47s, Average FPS: 27.98
Min FPS: 16.64 at frame 155, Max FPS: 38.74 at frame 853
Average Tri/Sec: -19202956, Tri/Frame: -686206
Recorded/Played Tris ratio: -1.34
!TimeDemo Run 1 Finished.
Play Time: 60.36s, Average FPS: 33.13
Min FPS: 16.64 at frame 155, Max FPS: 46.00 at frame 76
Average Tri/Sec: -22740762, Tri/Frame: -686362
Recorded/Played Tris ratio: -1.34
!TimeDemo Run 2 Finished.
Play Time: 59.85s, Average FPS: 33.42
Min FPS: 16.64 at frame 155, Max FPS: 46.00 at frame 76
Average Tri/Sec: -22910772, Tri/Frame: -685561
Recorded/Played Tris ratio: -1.34
!TimeDemo Run 3 Finished.
Play Time: 60.43s, Average FPS: 33.09
Min FPS: 16.64 at frame 155, Max FPS: 46.00 at frame 76
Average Tri/Sec: -22725354, Tri/Frame: -686677
Recorded/Played Tris ratio: -1.33
TimeDemo Play Ended, (4 Runs Performed)
==============================================================
==============================================================
Single Card. NO OC Driver 174.74
TimeDemo Play Started , (Total Frames: 2000, Recorded Time: 111.86s)
!TimeDemo Run 0 Finished.
Play Time: 101.41s, Average FPS: 19.72
Min FPS: 8.31 at frame 145, Max FPS: 23.79 at frame 992
Average Tri/Sec: -18659988, Tri/Frame: -946124
Recorded/Played Tris ratio: -0.97
!TimeDemo Run 1 Finished.
Play Time: 94.37s, Average FPS: 21.19
Min FPS: 8.31 at frame 145, Max FPS: 25.79 at frame 84
Average Tri/Sec: -19797954, Tri/Frame: -934204
Recorded/Played Tris ratio: -0.98
!TimeDemo Run 2 Finished.
Play Time: 94.63s, Average FPS: 21.13
Min FPS: 8.31 at frame 145, Max FPS: 26.07 at frame 65
Average Tri/Sec: -19736752, Tri/Frame: -933848
Recorded/Played Tris ratio: -0.98
!TimeDemo Run 3 Finished.
Play Time: 94.58s, Average FPS: 21.15
Min FPS: 8.31 at frame 145, Max FPS: 26.07 at frame 65
Average Tri/Sec: -19765344, Tri/Frame: -934730
Recorded/Played Tris ratio: -0.98
TimeDemo Play Ended, (4 Runs Performed)
==============================================================
==============================================================
SLI NO OC Driver 174.74
TimeDemo Play Started , (Total Frames: 2000, Recorded Time: 111.86s)
!TimeDemo Run 0 Finished.
Play Time: 64.04s, Average FPS: 31.23
Min FPS: 7.74 at frame 138, Max FPS: 41.59 at frame 980
Average Tri/Sec: -21431494, Tri/Frame: -686215
Recorded/Played Tris ratio: -1.34
!TimeDemo Run 1 Finished.
Play Time: 55.64s, Average FPS: 35.94
Min FPS: 7.74 at frame 138, Max FPS: 47.84 at frame 66
Average Tri/Sec: -24623870, Tri/Frame: -685097
Recorded/Played Tris ratio: -1.34
!TimeDemo Run 2 Finished.
Play Time: 55.21s, Average FPS: 36.22
Min FPS: 7.74 at frame 138, Max FPS: 47.84 at frame 66
Average Tri/Sec: -24807960, Tri/Frame: -684879
Recorded/Played Tris ratio: -1.34
!TimeDemo Run 3 Finished.
Play Time: 55.30s, Average FPS: 36.17
Min FPS: 7.74 at frame 138, Max FPS: 47.84 at frame 66
Average Tri/Sec: -24790002, Tri/Frame: -685430
Recorded/Played Tris ratio: -1.34
TimeDemo Play Ended, (4 Runs Performed)
==============================================================
==============================================================
Single Card. OC (core 600, Memory 950, Shader 1500) Driver 174.74
TimeDemo Play Started , (Total Frames: 2000, Recorded Time: 111.86s)
!TimeDemo Run 0 Finished.
Play Time: 86.71s, Average FPS: 23.07
Min FPS: 16.92 at frame 155, Max FPS: 28.64 at frame 996
Average Tri/Sec: -21814536, Tri/Frame: -945776
Recorded/Played Tris ratio: -0.97
!TimeDemo Run 1 Finished.
Play Time: 80.47s, Average FPS: 24.86
Min FPS: 16.92 at frame 155, Max FPS: 29.89 at frame 101
Average Tri/Sec: -23244668, Tri/Frame: -935202
Recorded/Played Tris ratio: -0.98
!TimeDemo Run 2 Finished.
Play Time: 80.25s, Average FPS: 24.92
Min FPS: 16.92 at frame 155, Max FPS: 30.93 at frame 71
Average Tri/Sec: -23284408, Tri/Frame: -934328
Recorded/Played Tris ratio: -0.98
!TimeDemo Run 3 Finished.
Play Time: 80.56s, Average FPS: 24.83
Min FPS: 14.54 at frame 1312, Max FPS: 30.93 at frame 71
Average Tri/Sec: -23221272, Tri/Frame: -935333
Recorded/Played Tris ratio: -0.98
TimeDemo Play Ended, (4 Runs Performed)
==============================================================
==============================================================
SLI OC (core 600, Memory 950, Shader 1500) Driver 174.74
TimeDemo Play Started , (Total Frames: 2000, Recorded Time: 111.86s)
!TimeDemo Run 0 Finished.
Play Time: 60.73s, Average FPS: 32.93
Min FPS: 9.54 at frame 144, Max FPS: 49.49 at frame 976
Average Tri/Sec: -22604230, Tri/Frame: -686353
Recorded/Played Tris ratio: -1.34
!TimeDemo Run 1 Finished.
Play Time: 53.25s, Average FPS: 37.56
Min FPS: 9.54 at frame 144, Max FPS: 55.77 at frame 101
Average Tri/Sec: -25706714, Tri/Frame: -684444
Recorded/Played Tris ratio: -1.34
!TimeDemo Run 2 Finished.
Play Time: 52.43s, Average FPS: 38.15
Min FPS: 9.54 at frame 144, Max FPS: 56.72 at frame 67
Average Tri/Sec: -26121552, Tri/Frame: -684746
Recorded/Played Tris ratio: -1.34
!TimeDemo Run 3 Finished.
Play Time: 52.59s, Average FPS: 38.03
Min FPS: 9.54 at frame 144, Max FPS: 56.72 at frame 67
Average Tri/Sec: -26063496, Tri/Frame: -685328
Recorded/Played Tris ratio: -1.34
TimeDemo Play Ended, (4 Runs Performed)
==============================================================

It seems odd that you get such a small performance boost with such a huge overclock, but these are real game results.

 

The quads seem to hold you back but that`s messed up. Magnus72 has a 2.4 CPU and he gets better results at the same OCing values...

 

Is that DX10 High? In my opinion is 38 fps average really good if it is in DX10. DX9 on the other hand 38fps average is done without a sweat. Oh just noticed that it is in 1680x1050 res. Hmm should be better than that.

 

1680x1050 res if my LCD res, so thats what I can work with. And yes, DX10 HIGH.
I do notice however that Crysis does not detect MMX on my CPU. Anyone know how to force it on?

Log Started at Friday, April 04, 2008 11:55:31
Running 64 bit version
Executable: C:\Games\Crysis\Bin64\crysis64.exe
FileVersion: 1.1.1.6156
ProductVersion: 1.1.1.6156
Using STLport C++ Standard Library implementation

--- CPU detection ---
Total number of logical processors: 4
Number of available logical processors: 4

Processor 0:
CPU: Intel Intel(R) Core(TM)2 Quad CPU Q6700 @ 2.66GHz
Family: 6, Model: 15, Stepping: 11
FPU: On-Chip
CPU Speed (estimated): 2659.961548 MHz
MMX: not present
SSE: present
3DNow!: not present
Serial number not present or disabled

Processor 1:
CPU: Intel Intel(R) Core(TM)2 Quad CPU Q6700 @ 2.66GHz
Family: 6, Model: 15, Stepping: 11
FPU: On-Chip
CPU Speed (estimated): 2659.949178 MHz
MMX: not present
SSE: present
3DNow!: not present
Serial number not present or disabled

Processor 2:
CPU: Intel Intel(R) Core(TM)2 Quad CPU Q6700 @ 2.66GHz
Family: 6, Model: 15, Stepping: 11
FPU: On-Chip
CPU Speed (estimated): 2659.888910 MHz
MMX: not present
SSE: present
3DNow!: not present
Serial number not present or disabled

Processor 3:
CPU: Intel Intel(R) Core(TM)2 Quad CPU Q6700 @ 2.66GHz
Family: 6, Model: 15, Stepping: 11
FPU: On-Chip
CPU Speed (estimated): 2659.951094 MHz
MMX: not present
SSE: present
3DNow!: not present
Serial number not present or disabled

Total number of system cores: 4
Number of cores available to process: 4
---------------------
Windows Vista 64 bit SP 1 (build 6.0.6001)
System language: English
Windows Directory: "C:\Windows"
Prerequisites...
* Installation of KB940105 hotfix required: no! (either not needed or already installed)
Local time is 11:55:31 04/04/08, system running for 3 minutes
4094MB physical memory installed, 3346MB available, 8388607MB virtual memory installed, 18 percent of memory in use
PageFile usage: 28MB, Working Set: 6MB, Peak PageFile usage: 29MB,
Current display mode is 1680x1050x32, (Unknown graphics card)
IBM enhanced (101/102-key) keyboard and 16+ button mouse installed
--------------------------------------------------------------------------------
Stream Engine Initialization
Network initialization
[net] using iocp socket io management
Physics initialization
MovieSystem initialization
Renderer initialization
Console initialization
Time initialization
Input initialization
Sound initialization
Sound - initializing AudioDevice now!
Sound - starting to initialize DirectSound output!
Sound - 2 drivers found:
Sound - available drivers: 0 Speakers (High Definition Audio Device) !
Sound - available drivers: 1 Digital Output Device (SPDIF) (High Definition Audio Device) !
Sound - initializing FMOD-EX now!
Sound - initialized FMOD-EX
Sound - available record drivers: 0 Microphone (High Definition Audio Device) !
Sound - available record drivers: 1 Line In (High Definition Audio Device) !
Sound - using FMOD version: 00040723 and internal 00040723!
Sound - initializing SoundSystem now!
Font initialization
AI initialization
Initializing Animation System
Initializing 3D Engine
Script System Initialization
Entity system initialization
Initializing AI System
[PlayerProfiles] Login of user 'Swiftnc' successful.
[PlayerProfiles] Found 2 profiles.
Profile 0 : 'default'
Profile 1 : 'Swiftnc'
[GameProfiles]: Successfully activated profile 'default' for user 'Swiftnc'
$3r_FSAA = $60 $5[DUMPTODISK, REQUIRE_APP_RESTART]$4
$3r_FSAA_samples = $60 $5[DUMPTODISK, REQUIRE_APP_RESTART]$4
$3r_FSAA_quality = $60 $5[DUMPTODISK, REQUIRE_APP_RESTART]$4
[net 15:55:52.016] connection requested to: <local>:64087
[net 15:55:52.018] resolved as: <local>:64087
[net 15:55:34.453] network hostname: SHODAN
[net 15:55:34.468] ip:192.168.1.136
[Error] <Flash> Error: CallMethod - 'gotoAndPlay' on invalid object. [Libs/UI/HUD_GrenadeDetect_Friendly.gfx]
[Error] <Flash> Error: CallMethod - 'gotoAndPlay' on invalid object. [Libs/UI/HUD_GrenadeDetect.gfx]
-----------------------------------------------------
*LOADING: Level island loading time: 48.89 seconds
-----------------------------------------------------
$3demo_restart_level = $60 $5[]$4
$3g_godMode = $61 $5[]$4
$3demo_file = $6benchmark_gpu $5[]$4
$3demo_ai = $60 $5[]$4
$3demo_num_runs = $64 $5[]$4
$3demo_quit = $61 $5[]$4
$3hud_startPaused = $60 $5[]$4
$3sys_flash = $60 $5[]$4
$3r_VSync = $60 $5[DUMPTODISK, RESTRICTEDMODE]$4
$3r_NoDrawNear = $61 $5[]$4
$3demo_time_of_day = $68 $5[]$4
$3s_DialogVolume = $60 $5[DUMPTODISK]$4

 

The most striking thing I see is this:

In the non-overclocked case SLI increases the maximum FPS by about 84%.
But, the average FPS only increases by 70%. Why? Perhaps it's because SLI, rather than increasing minimum FPS, actually REDUCED it by 7%.

The overclocked case is even more extreme. In this case, SLI increases the maximum FPS by 84% (as it did in the non-overclocked case), but the minimum FPS is REDUCED by 35%! As a consequence, the average FPS only increased by 53%.

Strange stuff (to me, anyway).

[edit] This reminds me, a little, of the non-SLI/SLI 3DMark06 scores Justin just posted for dual and quad machines. The dual got an increase of 51% in the sm2/sm3 scores (the increase would have been higher at higher resolution certainly), but the quad got an increase of only 37%.

 

Swiftnc <<<<<<<< Thanks, again for the homebrew SSD results you have shared! (see the 'DIY SSD' thread)

Dont give up on the idea yet !!!!

Are your drives allowing 'optimize for performance' to be enabled, you may have to use the Hitachi MicroDrive Filter, it sits in the device chain and filters the 'removable device' flag from Windows allowing windows to see it as a 'true' fixed disk. See this guys experience driver is available here you need to follow this post to modify the driver for your device.

Also try Enhanced Write Filter - it prevents windows from doing lots of little writes to the card and occasionally flushes all the writes in one go. It can increase performance and card lifetimes. This guys site shows where and how.

Also you might want to disable access times (registry key) (prevents drive being written to everytime a file is read)

Maybe you will get some results with some of these ideas, at least you have the XP install 'spare' allowing you to experiment!

 

Actually I posted earlier...

I really think this is a BIOS issue.

 

Have you checked in the BIOS to see if it has an event log, and if so whether the event log flagged anything related to the CF cards?

 

I swear, I`m gonna sugest opening a separate Crysis forum

 

Good, when you do please make sure you first post tells me how to fix Crysis from using MMX on my processor!

 

No game uses MMX in the first place. That is an old feature that came along the old Pentium MMX CPU´s back in the old days. You don´t even have MMX on that CPU in the first place

 

They won`t make a crysis subforum,it`s not a gaming forum . Double drats
What I really want to see is somebody testing the fastest RAID (0) with SSDs....