Palit GeForce 9800 GTX

[W1zzard]

To read this review go to: http://www.techpowerup.com/reviews/P...orce_9800_GTX/

Please Digg the review to help us promote it.

[newtekie1]

I've noticed that a lot of the 9800GTX cards don't clock to the correct speed on the memory. Like you said, the 0.8ns RAM should be able to hit about 1250MHz. Perhaps the manufactures are putting less voltage through the RAM and running it at slower speeds to help with the heat? Or maybe they are running tighter than spec timings? My eVGA 9800GTX overclocks to similar numbers to yours.

[Temps_Riising]

Very nice review! I like the fact that your test setup is a realistic but upto date average system with average overclocks that are pertinent to the average TPU enthusiast.......in my experience a rarety with many reviews from other sites.

@ Newtekie.....I would be interested to see, how the 9800GTX at say 775mhz core would perform against a G92 GTS at say 825mhz on the core......do you think the GTS would match it?

[WarEagleAU]

Very strange indeed on that ram. Maybe its a bad batch?

[Castiel]

Well a friend had to return 3 of his Palit 9800GTX's cause some where DOA and one of them wasn't the same as the others.

[Jejking]

The load temps are VERY high, and I suspect that the OC-ability suffered horribly. Now you have reapplied the thermal paste. Can you check again the max frequencies? I think you might be surprised

[jbunch07]

very nice in depth review!

[Wester547]

Great review, though the Quake 4 frame rates still seem rather precariously low when it comes to every graphics card tested and shown.

Another thing in question - a $40 price for the 8600 GTS seems a tad low, even if it's not quite a midrange card anymore - shouldn't it be anyplace closer to around $80? Speaking of the 8600 GTS, I'm also rather shocked that the 512MB one seems to be slower than the 256MB one for the most part. That's a Biostar one if I'm not mistaken, but some of the reviews I've seen of other 512MB ones (the Gainward one on Fudzilla an the Vvikoo one on Hartware.de) have it getting significantly higher frame rates (a 20%-45% augmentation) when FSAA is enabled at higher resolutions. I myself have a Gainward one, and am getting 37FPS average in F.E.A.R. at 1600x1200 w/4x FSAA & 16x AF instead of 30FPS (running it at stock speeds, with no texture filtering optimizations enabled, though even with them on I still get 37FPS running the built-in test at the aforementioned settings). Not to vaunt or anything along that nature.

And just to add, I think the GeForce 7900 GTX's score in 3DMark03 is quite high; it is not significantly afar from the 30,000 mark, when if my recollection serves me correctly that card used to get around 24,000 at the most at stock speeds (not in SLI, of course). It must be that it's been paired with a good CPU and that nVidia had ameliorated upon their drivers quite a bit.

Also, SP3 isn't installed on the test machine? If I may ask.

And whilst the 9800 GTX may not be all that challenging to the 8800 GTX, consider how many times more expensive the 8800 GTX was, say, a year ago? So a $280 price tag is rather nice, keeping that thought at view. I wouldn't mind coming to grips with a board that's slightly slower at that price.

Nevertheless, good review.

[W1zzard]

- we use a custom quake 4 timedemo which is from one of the most demanding parts of the game
- fixed the 8600 gts price to 65$
- 7900 gtx 27000 is as far from the 30k mark as it is from the 24k mark (3k difference each)
- no the machine is still sp2, sp3 has not effect on framerates afaik

[xubidoo]

Originally Posted by W1zzard

- we use a custom quake 4 timedemo which is from one of the most demanding parts of the game
- fixed the 8600 gts price to 65$
- 7900 gtx 27000 is as far from the 30k mark as it is from the 24k mark (3k difference each)
- no the machine is still sp2, sp3 has not effect on framerates afaik

i used to be in a clan for Quake 4 ,and within the Q4 community its pretty well known that Antialiasing in Q4 is not implemented properly ,same goes for SMP ,if u enable either or both then performance goes downhill fast,as your benchmarks appear to confirm