4850 vs 9600GSO SLI

Insomniator

Diamond Member
Oct 23, 2002
6,294
171
106
Hey guys I finally got my system up and running. I originally bought two 9600gso 768's at frys just to try SLI. Well, my board fried and I ended up replacing pretty much everything...

So I went from:

X2 5000+ @ 2.9
Abit KN9 SLI
Radeon 4850
4gb DDR2 1000
Ultra 500w

to

e5200 (currently at 3.5, 280x12.5.. 12 hrs prime stable...won't boot at 290 >: .....)
Evga 122 750i SLI
Radeon 4850/9600GSO 768 SLI (ass for overclocking... driver stops responding on anything higher than 600/1500/825)
4gb DDR2 1000
Corsair TX650


I'm still tinkering and everything, and using windows 7 so right there are some problems but first impressions are wooow quite a difference. Crysis on the 9600's plays silky smooth (first time I've ever seen the damn game play smooth) at 1900x1200/High. 34fps average vs 21 on the last system. I'm guessing the 4850 on the new system will fall somewhere inbetween but we'll see. Vantage is 6496 on the 4850 vs 8970 on the 9600's (6526 vs 7443 GPU).


Hopefully someone else here is curious about this kind of comparison. I haven't been able to find any GSO 768 SLI benchmarks online at all.

STOCK 9600's vs STOCK 4850

Numbers are min/max/avg

Warhead Gamer 1900x1200 0x AA

4850: 13.87/27.51/21.59
9600: 16.82/38.64/29.54

Warhead Gamer 1900x1200 8x AA

4850: 0/12.69/8.52
9600: 8.89/21.94/16.61

Warhead Enthusiast 1900x1200 0x AA

4850: 9.4/18.10/14.5
9600: 12.68/22.22/18.53

Warhead Enthusiast 1900x1200 8x AA

4850: 6.37/10.02/7.92
9600: Would not complete (froze on loop 2)

Far Cry 2 1900x1200 Very High 0x AA

4850: 32.26/59.13/43.53
9600: 32.68/67.86/44.72

Far Cry 2 1900x1200 Very High 4x AA

4850: 17.15/44.02/26.87
9600: 29.41/50.26/36.78

Far Cry 2 1900x1200 Ultra High 0x AA

4850: 22.37/44.75/31.77
9600: 29.54/54.21/39.16

Far Cry 2 1900x1200 Ultra High 4x AA

4850: 9.13/14.76/11.80 (Driver issue?)
9600: 25.50/43.06/31.13


I have a bunch of steam games (L4D, TF2, Lost Planet) but am not sure how to bench them effectively...

so far 9600's for 100 bucks >>>> 4850 for 100 bucks

 

AzN

Banned
Nov 26, 2001
4,112
2
0
Do some more benches.

GSO should overclock great. Maybe you need to modify the fan speeds and increase the voltage in bios.

I'm running a single g92 8800gts @ 1080p and I get 29fps in crysis ingame gpu benchmark @ high settings.
 

AzN

Banned
Nov 26, 2001
4,112
2
0
I wouldn't say it's substantial but it's still nice to see hefty gains. It would have been sweeter to compare it to 4870 512 and 1gb.

nicely done.

Try some overclocking. Edit the bios to raise your vcore to 1.1 and let it rip.
 

WT

Diamond Member
Sep 21, 2000
4,816
59
91
I'm running a pair of 384mb 9600GSO's OC'd to 685 and I just loaded Far Cry 2 onto the PC to test them out. Built in benchmark utils are a pleasure to use as well ;-) I see no reason why your cards can't OC better than that, unless your PSU is holding things back.

I've always been curious how well those cards in SLI compare to more modern single cards, but since its essentially a spare gamer rig, I rarely use it. In fact, I'm loaning it to a friend just so he can play some new games on it.
 

MustangSVT

Lifer
Oct 7, 2000
11,554
12
81
hmm interesting. i wonder if its worth anything to voltmod (soft) my 8800gs and o/c core higher.
 

Leyawiin

Diamond Member
Nov 11, 2008
3,204
52
91
That's a great SLI choice. Cheap - (what are they, $100 for the pair after rebate?) and the 768 MB frame buffer (per GPU) is better than the HD 4850's 512MB. The pair SLI'd are probably performing more like an HD 4870 512MB.
 

vj8usa

Senior member
Dec 19, 2005
975
0
0
Originally posted by: Leyawiin
what are they, $100 for the pair after rebate?

I remember seeing them for $36 after rebate on Amazon for a short while back in January. That was a hell of a bargain, and I actually found myself wishing I had an SLI mobo. They're harder to find nowadays, though.
 

WT

Diamond Member
Sep 21, 2000
4,816
59
91
IIRC, the $36 AR cards were the 384mb versions (eVGA brand dual slot cooler) that I purchased from the 'Egg. I hadn't seen the 768 cards go for that price, altho I may have just missed the sale.

I'm using mine on an MSI P6N Platinum board with an e6600 OC'd to 3.4, and it runs very well for the games that I play. My first foray into an SLI setup, and its definitely a good budget build.
 

Insomniator

Diamond Member
Oct 23, 2002
6,294
171
106
Originally posted by: Leyawiin
http://shop3.frys.com/product/...sr:SEARCH:MAIN_RSLT_PG

Only one rebate per household. Kind of ruins the deal (unless you have a willing friend or relative buy it and go through the rebate process for the second card).

haha exactly what I did with my gf. no pics.

Anyway does anyone know how to benchmark Left 4 Dead, TF2 or HL2 E2? I mean, I guess its not like any source based game would have a problem on even 1 9600.. I'm just curious.


Also it looks like windows 7 lags behind vista when it comes to Crysis. this kind of makes me want to go back to vista x64 because Crysis looks soooo much better on Very high/Enthusiast compared to high/gamer imo and is just a little slow on the current set up..
 

WT

Diamond Member
Sep 21, 2000
4,816
59
91
Ahh, the PNY brand ! My opinion .. I wouldn't buy a PNY card based on their ridiculous excuse for a 'lifetime' warranty. You can always roll the dice and hope for the best, but its not for me. Insom, there is no built-in benchmark for any of those games, which is why I mentioned it concerning Far Cry 2 ... its convenient for testing purposes.
 

cusideabelincoln

Diamond Member
Aug 3, 2008
3,275
46
91
You need to do a timedemo for source games. Just hop into a match online and type "record [filename]" (w/o quotes) into the console and the game will record your gameplay. When you're done, type "timedemo [filename]" in the console. At least this is probably the "best" way I know, although if you're getting really high frames you could be limited by hard drive speed.

An alternative method would be to run fraps while you play games online. For instance, try to find a server that's full in TF2 and plays the same map. Make multiple runs, trying to find the position on the map that gives you the best and worst framerates. Note these values and post them. The average framerate per session can't really be used because of the many variables associated with online gameplay, but if you do enough runs the absolute maximum and minimum could be a metric.