Hardware XFX GeForce 9800 GT 512MB XXX Edition

http://www.bit-tech.net/hardware/graphics/2009/02/05/xfx-geforce-9800-gt-xxx-edition-review/1

We take another look at affordable graphics, this time with XFX's GeForce 9800 GT 512MB XXX edition. Fitted with a custom cooler and packing increased clocks on the core, shader and memory, we find out how it compares with the competiton.

 

bloody proxy I am behind filters the article because of the XXX in the URL !

 

This is one of the types of cards for which some folding scores would have been nice to see

 

why only two games?

 

3 surely?

And it was felt that as an older architecture, our full gamut of benchmarks would have been a little over the top - the three we've included are more than enough to get an idea of performance in a variety of different game engines - would three pages of Crysis results really make it clearer that it's a few frames per second faster?

 

I have to say, as an owner of a 8800GT myself, and an overclocked golden sample at that, these really still remain as competent cards for most if not all games. At resolutions of 1680x1050 and below, its hard to justify spending twice as much for what amounts to merely more FPS.

For smaller monitors, this clearly has to be the Nvidia card of choice for anyone wanting to play modern games.

I have a question though... especially given that 9800GTs have all but replaced the 8800GT...

if you have an 8800GT, can you add a 9800GT and do SLI, or do you need another 8800GT, because as your article says, they are the same cards, just relabelled (the 65nm ones anyway)???

 

From Fallout3:
"[...] The engine unfortunately has a 60fps cap that we’ve been unable to remove thus far, but even despite this, we believe the test is very relevant. [...]"

So how come HD4850 averages over 60fps? Did u manage remove the cap? Saw it in other reviews also.

And overclocking "like a BEAST" for the extra 3fps is useless, in my opinion. Won't notice the difference.

Otherwise, yeah, crap value for this card. For 10$ (here) extra, the Sapphire HD4850 512MB would be a better choice.

 

Once again some people will get ripped off by a pre-overclocked card. $160? Come on, a 9800 GTX+ can easily be found for that, and even a few GTX 260's have gone for around $160 on ebay. Good review though.

 

Apparently you can force the cap off by setting ‘iPresentInterval=0” in both “Fallout.ini” and “FalloutPrefs.ini” located in “User\Documents\My Games\Fallout 3\”.

 

It's an error that has been replicated across a number of articles - we're fixing it now with what it should say. Thanks.

 

It really depends on the Game you choose to stress test with. If we'd chosen STALKER clear sky rather than Far Cry , the improvements would have been better, as that game is able to take full advantage of any increase in core, shader or memory clock.

As it stands, an extra 10% performance ontop of an already overclocked card isn't too shabby - that 3fps boost is ontop of a 35FPS average. Considering it's FPS for nothing, we'd say that's pretty good going.

 

fixed now

 

750mhz core?!!! is this due to the power management additions? if not you've got one hell of a G92 chip there

the 8800GT has proved to be a sterling buy for those who got one last year. i see no reason to upgrade tbh, especially with the games coming out atm, theres just nothing to justify spending more really (unless you have a 30" monitor or a crysis addiction)

 

SI there anyway we could get DB noise ratings when you review kit? I don't know if this is possible for you guys, but would be very useful for me and my machine is not only a gaming rig.

 

i have a evga standard 8800GT which i have flashed the bios to the akimbo SSC version which i then overclocked for a 740mhz core, 1800mhz shader and 1020 mhz memory overclock on the stock heatsink and fan. 8800GT was deff a good buy last year when i picked it up.

 

what is diffrent in 8800GT xxx editon?
They only test 3 games have played the most. So I can't find the table of crysis testing.

 

I'm pretty sure the back to front cooling setup is because the card is single height, and the designers were worried about cases without an additional venting below. Such a situation would likely lead to a greater bottling of heat by the back of the case if dumping the hot air from the card off the back side, than dumping towards the front where it might more easily be pushed up and out of the graphics card's section of air.

But for those of us deliberately ducting their cases, pushing air from front to back, with a rear exhaust by the graphics card, this design would be less effective than a front to back based cooler.

 

".... benefits from both a third party cooling solution and a hefty factory overclock. With the core clock increased from 600MHz to 670MHz, the shader clock bumped up from 1,500MHz to 1,600MHz and the memory upped from 900MHz (1,800MHz effective) to 975MHz (1,950MHz effective)."

That's not even 10percent.. is that supposed to be a lot?

 

correct, an overclock of 70Mhz to the core, which usually has the most direct impact on performance is in fact 11.6 percent, with the shader increased by 6.6 percent and the memory by 8.3 percent - that's pretty decent to be honest, and roughly inline with what you'll see from even the top end of overclocked cards,

 

Ermm.. okay. I was just comparing it to the percentages that you would get from a CPU.