Test System:
E6300 @ 3.05 ghz on a Scythe Mine with custom 120mm fan for 800/1200/1600rpm
Asus P5B vanilla
2 gigs @ 870 mhz 5-5-5-15 timing
Onboard Soundmax audio
Samsung 250gig
Hitachi 160 gig
19” Viewsonic VA1912WB
Antec Sonata 2 with SP 2.0 450watt
Vista x86 ultimate
Nvidia forceware 174.74
All tests are done at 1440x900 otherwise noted.
Tested:
3dmark2006 default
Crysis 1.2
World in Conflict
Need 4 Speed Prostreet
As you can see, pixel fillrate does hold back by memory bandwidth to a certain extent but core clock speed increases does more than memory bandwidth alone.
Texture fillrate doesn’t get affected by memory bandwidth. Notice 8600gts texture fillrate is rather low for a card with a theoretical fillrate of 11200 MTexels/s @ clock speed of 700mhz. G92 has improved its texture throughput over G84 by 20+ percent.
Message edited by marvelous211 on 04-03-2008 at 06:18:22 AM
Benchmark was done with GPU portion of in game crysis benchmark. The new 174.74 drivers gave me substantial increase in FPS over the drivers CD, nearly 3fps in high settings. CD was using older 169 set of drivers. I would say it performs very close to 8800gt at those frame rates especially dx9 high settings.
I was able to play Dx9 high settings near smoothness. It does dip to high teens in Ice Level and the Deck however it was much more enjoyable. I was able to turn up shader to very high with minimal impact in fps. Post processing at very high in the other hand dropped my fps by 60+%. It might have something to do with 8800gs 384mb vram.
Dx10 high settings were using 2xAA and Dx10 highest uses 4xAA by default. Biggest problem with 8800gs with WIC with Anti Aliasing is the 384mb vram that hampers performance. Not to mention WIC favors pixel performance and memory bandwidth. Biggest gains from the overclock were @ dx10 highest without AA.
Message edited by marvelous211 on 04-02-2008 at 11:36:43 PM
The game was tested with fraps using Autobahn GP gripclass track. It's a 3 lap course. This game seems to love shader. Notice the substantial gain from the overclock once again. At 8xAA I was able to get nearly 20 more fps over the 8600gts without AA.
Conclusions:
What compelled me to buy this card over the 9600gt was that it has 96SP and more texture fillrate and it was cheaper. I'm on a 19" LCD monitor where memory bandwidth or vram isn't really an issue but longevity of the card. I do think that as more game use more shader 8800gs will surpass 9600gt in raw performance. Currently these cards are about equal in terms of raw performance at stock clocks. 8800gs wins some and 9600gt win some. At those overclocks however I'm convinced 8800gs does take the cake especially without AA where memory capacity or memory bandwidth isn't an issue. 9600gt however will always do better with AA in higher resolutions because of bigger memory module and memory bandwidth. At 1044mhz/2088DDR it's not exactly slow either. That's 50.2GB/s where stock 9600gt is 57.6GB/s.
The card was very quiet. At 700rpm in windows it wasn't even audible to my ear. My case doesn't exactly have lot of airflow with only 1 800rpm case fan and the size of this card made my cpu temperature rise by 5C. The temperature reached 80C with the overclock which I edited the bios to make the card fans to be more aggressive to 55% when gaming. It sits at 76C on load. At 55% fan speed it's only a slight hum to my ear.
With a huge overclock it gained as much as 30% more fps. Hopefully you gained some important information from this review.
Pros: Cheap price, Excellent performance per $$$, highly overclockable, quiet heatsink fan, 96 Stream processors, 1ns memory, life time warranty
Cons: Limited quantity, 384mb vram, huge size which can block airflow if you have a medium size case
Message edited by marvelous211 on 04-03-2008 at 06:08:19 AM
Decent review, however, how exactly did you test WIC and NFS? Running them exactly each time is rather difficult, so rather a lot of variables are introduced. Needless to say, Ocing really helps and the 8600GTS sucks.
Also, without a 9600GT, how exactly can you conclude that the 8800GS does as well as the 9600GT in some settings? What you have done is guessed or brought in outside reviews in, not conclude. You're also forgetting that the 9600GTs are real good OCers, making them a real good deal.
Finally, you're missing the price at which you bought it at or the MSRP price. Without this info, how am I supposed to know if the 8800GS has a better price/performance than the 8600GTS?
You have tests with 2 badly programed games such as NFS: prostreet and Crysis. Any chance of seeing some mainstream games? Source engine, Unreal 3, Quake 4 or doom3 ?
Also for the 3dmark06 did you run default? No aa ? or wat?
Decent review, however, how exactly did you test WIC and NFS? Running them exactly each time is rather difficult, so rather a lot of variables are introduced. Needless to say, Ocing really helps and the 8600GTS sucks.
Also, without a 9600GT, how exactly can you conclude that the 8800GS does as well as the 9600GT in some settings? What you have done is guessed or brought in outside reviews in, not conclude. You're also forgetting that the 9600GTs are real good OCers, making them a real good deal.
Finally, you're missing the price at which you bought it at or the MSRP price. Without this info, how am I supposed to know if the 8800GS has a better price/performance than the 8600GTS?
WIC has ingame benchmark. In prostreet the track is rather long. I ran fraps multiple of times and each time the average frame rate was within 1 fps of each other. I used average of multiple runs from the ingame testing.
There are dozens of reviews out for 9600gt which I made a judgment using multiple sources.
8800gs is currently @ newegg for $140 shipped. $110 after rebate. I think newegg or EVGA should give me a 9600gt to play with for advertising them in my review.
I would love to do a in depth review of both of these cards.
Message edited by marvelous211 on 03-29-2008 at 11:47:15 PM
You have tests with 2 badly programed games such as NFS: prostreet and Crysis. Any chance of seeing some mainstream games? Source engine, Unreal 3, Quake 4 or doom3 ?
Also for the 3dmark06 did you run default? No aa ? or wat?
good job though
Thanks.
3dmark was run on default settings 1280x1024. NO AA was used. I do have Unreal 3 and COD4 but this was just a mini review for you guys to compare.
Message edited by marvelous211 on 03-29-2008 at 11:21:34 PM
Wow, You don't waste time. Nice job and thanks for sharing this with us. You got a real nice OC out of that card and temps aren't bad considering clocks. For $110, you must be thrilled how this does vs the 8600GTS.
I also found it interesting the SP 2.0 450W is enough for your OC'ed setup. Good to know as I have one pulled from my sonata II I can use with a GS now. It's supposedly 22 amps combine max 12v according to this link anyway. http://forums.guru3d.com/showthread.php?t=205763
Wow, You don't waste time. Nice job and thanks for sharing this with us. You got a real nice OC out of that card and temps aren't bad considering clocks. For $110, you must be thrilled how this does vs the 8600GTS.
I also found it interesting the SP 2.0 450W is enough for your OC'ed setup. Good to know as I have one pulled from my sonata II I can use with a GS now. It's supposedly 22 amps combine max 12v according to this link anyway. http://forums.guru3d.com/showthread.php?t=205763
Yeah tell me about it. I was benchmarking all last night and stuck all the info in excel and wrote the article when I woke up. I had 8600gts data available before hand.
Mine is version 2. It has 32 amps on 2 12volt rail.
My card looks exactly like the picture I posted. It's green. It's so much longer than my 8600gts.
Organize the reviews and then send them to my email: theo.dm@gmail.com and I'll post it up on my site. We need all the reviews we can take, since we're trying to please everyone.
Organize the reviews and then send them to my email: theo.dm@gmail.com and I'll post it up on my site. We need all the reviews we can take, since we're trying to please everyone.