BFGTech & XFX GeForce 7800 GT

I forgot to include this picture in the review..... give you an idea of lengths of each card:

7800 GTX, 7800 GT, 6800 GT:

http://www.bit-tech.net/content_images/bfgtech_xfx_7800gt/xfx-lengths.jpg

;)

i hope quake4 to come to benches soon :)

if/when it comes out, we'll definitely be looking at it ;)

Originally Posted by bigz
if/when it comes out, we'll definitely be looking at it ;)

if? haven't you see the qcon videos :D

nope, not seen any yet :o

I've been too busy writing this week's two articles

Only thing I didn't like was that the SC:CT graph seemed a bit cramped so I couldn't quite tell what went with what for visual effects on the far right. Great review overall... I've already spotted them for $419, including the overclocked XFX version. Cooler definately seems to have the same layout as the 6800-series, but no heatsink on the voltage regulators from what I can tell. A pic of the guts would have been nice, but seeing as you were holding $900 of unreleased graphics cards, you're somewhat forgiven.

Hopefully someone figures out how to hack open the other four pipes (assuming is IS using a plain G70 core and not some variant, I missed that if it was stated)... I wonder what it can do on watercooling :D

Firehed,

It's a plain G70, but the bios pipes are "disabled"...not able to be done by normal means...will require a bit more research, apparently.

Let me fire up rivatuner... I'll grab a shot of NVStrap

http://www.suxeh.net/bigz/nvstrap.jpg

Notice that they say disabled in the state and there is no hardware masking.

Will require a genuine bios flash, as opposed to just 'enabling' them. Not a switch on/switch off operation.

Like the 9800SE... :) Only problem is, with that you turned it into a $150 paperweight if it failed. Now you'd be turning this into a $450 paperweight.

Originally Posted by Hamish

i really wanna see what an OC'd 7800gtx + 2gig + 2.8gig FX can do in FEAR :D

*cough* in SLi mode, obviously :D

Originally Posted by mclean007
*cough* in SLi mode, obviously :D

If bigz tells me how best to go about it, i'll give it a go.........i'd have to rob a gig of memory out of another machine but that's easy fettled :)

The NV-68 blocks with fit the GT's without the memory on the reverse side but looks like the GTX guys are stuck with air until the end of the month when the NV-78's hit the streets, unless someone is prepared to getto one!

Originally Posted by coolmiester
unless someone is prepared to getto one!

/raises hand

although seeing what the GT is capable of, I may just save a few quid and go for that

but either way, ill definitely be waiting for my states trip in november, where ill pick one up for the equivalent of £20 no doubt :D

i would love to see an 'apples to apples' F.E.A.R. run since apples to apples is all that matters to me

furthermore i would like to see a standard run through of 3dmark05. i fail to see why its so horrible. ive never taken it to be "you can run games that look like this!" ive taken it to be "card X gets this number and card Y gets this" or like a competition, its just fun to watch.

F.E.A.R. is very hard to compare apples to apples as the AI do things so differently during each manual run through that I do. I will try and see how consistent I can get it to be, but advanced AI systems make the game that little bit harder to benchmark - in much the same way that BF2 is not easy to benchmark 'real world' game play because the situations (on the same map) can be so varied, even with AI that should be relatively unintelligent.

There's no point me running a "fly by" without any shooting action, because how many people play first person shooters just to look at the pretty graphics? :)

I will have a look in to this though, as it would make a really good apples to apples comparison with it being the most advanced graphics engine around at the moment IMHO.

Moving on to your other point:

The problem with 3DMark is that many OEMs and system builders take a 3DMark score to be the be-all-and-end-all of gaming performance - the whole logic of "if it performs well in 3DMark it must be good!"

I am not willing to go down that route in all honesty, and I will continue with not using 3DMark as a benchmark for that reason alone. I'm sure that won't please you, but I don't think a benchmark should be the determining factor in your next video card purchase. :)

How can a pre-determined benchmark run represent any kind of true performance when ATI and NVIDIA (the latter got caught with their pants down back in GeForceFX days) spend their time optimising the benchmark to buggery. I'd rather see ATI and NVIDIA spend that 'wasted' time optimising the top 100 game titles to make my gaming experience as good as possible.

I am not totally against 3DMark though, there are some nice tools inside the benchmark - the fillrate/pixel shader/vertex shader tests, along with the batch tests are all really useful and I use them to establish what I'm looking at in terms of GPU performance capabilities.

You can't just do a FEAR custom timedemo? Or would that make it so the AI is all prerecorded and you become that much less CPU liminted?

1.) I don't know whether F.E.A.R. currently has a timedemo feature.

2.) I don't particularly want to resort to time demos as they do not represent real-world game play.

:)

Originally Posted by bigz
1.) I don't know whether F.E.A.R. currently has a timedemo feature.

2.) I don't particularly want to resort to time demos as they do not represent real-world game play.

:)

Now that's a quote of a man with balls.



As for 7800 GT, I'm just happy that 6800GT prices dropped as a result. Now I might be able to buy it =).

Originally Posted by bigz
I'm sure that won't please you, but I don't think a benchmark should be the determining factor in your next video card purchase. :)

believe me its not. its really just a thing of entertainment for me. like sitting back and watching a horse race or something. i dont take it to mean anything. i know it doesnt since in 3dmark 05 a 9800pro can almost double a geforcefx 5900 and in real life a 9800pro isnt double the video card. but i still find it entertaining. and i see why you leave it out. TBH its a great idea if there are still so many people that think it means anything. its just like "oooh card Z gets 90000 points? wow! company A is in the lead now."

but anyways, if a F.E.A.R. A-to-A comparison is possible in any way it'd be great.

Originally Posted by sadffffff
but anyways, if a F.E.A.R. A-to-A comparison is possible in any way it'd be great.

I will try :)

You might of mentioned this in the review bigz but maybe i missed it!

....did you use the bundles drivers or download the latest??

I've just installed the 77.77 and the cards don't seem to want to overclock nearly as far as with the 77.72's using coolbits :?


Oh yeah and what have you found best to have the Sli rendering mode set to??

Sorry for all the questions!

http://www.coolercases.co.uk/images/sli/sli_rendering.jpg

Hi Paul - I used a mixture of drivers due to the time I had between receiving the cards and the drivers.

I used 77.76 on all cards except the 7800 GT's (and the HIS X850 XT PE, of course), but used 77.77 on the 7800 GT's - I am told that the performance is very similar between the two drivers. I mentioned the drivers/test setup at the bottom of page 5. ;)

I have also had issues with overclocking via coolbits, which is why I left the overclocking out. PowerStrip allowed me to hit 500MHz with the XFX pair in SLI, but I wasn't 100% sure that the settings were 'safe', as I couldn't use either RivaTuner or Coolbits to check the clocks were ok using NVIDIA's loop back artifact tester. I did play through some games on them at those clock speeds and I couldn't see artifacting, however you're not alone in finding that 77.77 aren't great for overclocking. :)

I've also confirmed that you will not be able to unlock these pipelines, much like NV41 and A04 silicon used for NV43.

Cheers bigz, it's kinda comforting to know i'm not the only one having issues with the 77.77's but wouldn't you think they would get this type of fundamental error ironed out before releasing for public consumption!

One day everything will work straight out the box.............what am i saying, no it wont

My bad for missing the driver setup on page 5........its been how shall i say, an interesting week!

Time permitting, you still get the vote for producing a condensed article for the various ideal configurations and reg edits which could be a GFX Bible for people like myself who have to trawl page after page of useless information.

Have a butchers here
http://www.xtremesystems.org/forums/showthread.php?t=71558

Makes for some interesting reading on the overclocking front, quite interesting to see, for the first time overclocking doesnt actually have any (or very very little) effect on performance at certain speeds.

I want one...(ehh maby two would be better... ) but their soooooo
expensiv :'( :'(