The Official 9600GT Reviews Thread

are you confusing the 9600gt with the 9600gt-sli mito?

 

Yes, that's where I got the impression that 9600GT is faster than old 8800GTS - I didn't realize they just toss in old results for the other cards tested with older drivers (though given the amount of cards they compare to it's not really surprising...).

 

I believe I did confuse.

 

9600GT SLI outperforms 3870X2, but is about $90 cheaper according to this: http://en.expreview.com/?p=273

 

Too few sampling points(only 3DM and COD4 AFAIR). OTOH, considering the 9600GT is pretty much on par with/better than the 3870, it's hardly surprising that in SLi it is on par/better than the X2. The caveat is that you need an SLi mobo to go with that(and nV's chipsets are ugh-worthy...at least the Intel ones), and that you'd have two cards filling two PCIE slots and that's that(the 9600GT doesn't do TRI-SLi), whilst with the X2 you'd have the possibility of going with another one and getting a QuadCF configuration.

More options is a good thing...heck, nV even slightly getting off of their arses has gotten AMD to lower prices and bump clocks on its parts...the X2 will probably get cheaper as well once the 9800GTX and the GX2 are released, if not sooner.

 

Scott's review is up.

 

http://www.techreport.com/articles.x/14168

Look at the Texture filtering: Bilinear texels/clock and Peak bilinear
texel filtering rate (Gtexels/s) .. the G94 is no slouch.

Since B3D don't review anymore, i've come to really like TechReports reviews and have come to hold them in high regard. I also like the fact that they review at 1680x1050. Thx Scott

US

 

Yes but those numbers are a bit misleading - if you'd only look at filtering the G80 numbers would be twice as high, but it's limited by texture address. So if you'd look at trilinear texels/s the G80 numbers would stay the same but G92/G94 would only be half. Anyway neither chip has a deficit in texture sampling performance...

 

I doubt many people will go quad GPU, even with the $$$ and power budget it's pretty ugly. you're doubling the troubles of dual GPU.

 

Because they dont have one.

 

Well, the $$$ is equivalent to what you had to spend for dual 2900s or GTSs, and so is the power budget. WRT to scaling I wholly agree though, from what I've seen it's going to be a fairly long road. I simply mentioned it as an alternative that exists in one scenario and not in the other.

 

With different drivers, all these reviews are quite worthless, since the 174-FW does seriously improve performance in not only a single case. (This is true for G80 and G92 also).

http://www.pcgameshardware.de/aid,6...4_sorgt_fuer_bis_zu_65_Prozent_mehr_Leistung/

 

Interesting tidbit from the techreport review

http://www.techreport.com/articles.x/14168/4

Could explain why Nvidia hasn't gone GDDR4 - maybe it's architecture will suffer more from this phenomenon.

 

I don't buy that. The 3870 is more than twice lower. Even assuming you'd have to throw away half the data read (burst length is indeed twice as high as for gddr3) it should still be significantly faster than that.
And, even if I'd were to believe that, suddenly in a Crossfire configuration this is no longer an issue and the performance gets _5_ times higher compared to a single card? Talk about good scaling here...

I'd think for a good memory controller efficiency shouldn't suffer that much. You try to read much more data at once anyway ideally.

 

This has seemingly caused some confusion for a few reviewers:

http://www.techpowerup.com/reviews/NVIDIA/Shady_9600_GT

But I presume that most are careful to run a system at stock PCI Express clocks so it mostly wouldn't have made any difference to published reviews.

Jawed

 

MSI preps a 2 GiB 9600GT:
http://translate.google.com/transla...d_Display-Port/&langpair=de|en&hl=de&ie=UTF-8

 

300mhz ddr1?