NVIDIA GeForce 7800 GTX 512

[Steve]

NVIDIA's efforts with the new G70 earn them the overall performance crown back, and by quite some margin. At 550MHz across the board and with 512MiB of memory running at 1.7GHz effective, it would have been embarassing had that not been the case. That didn't happen, though, and the reference-clocked GTX 512 has no trouble dispatching GeForce 7800 GTX and Radeon X1800 XT to the sidelines.

Quite simply a performance monster, and one that'll take ATI a big effort to beat. The GeForce 7800 GTX 512 impresses hugely, from quiet and efficient cooler, to manageable increase in power draw over the original GTX, with monsterous performance and day-zero availability in-between.

While certain facets of ATI's new Radeon hardware make it more appealing on paper, it's taking time for those technological advantages to come to fruition as part of games, making the latest GeForce hardware even more appealing in the interim. The current real-world performance of both GTX and GTX 512 cement that statement.

http://www.hexus.net/content/item.php?item=3904

[neonplanet40]

WOoohoo its awesome! Definetly worth the wait. Price is even more steep than i expected though! Im sure it will drop a little though in a few weeks!

ATI is really in the dogs now!

Neo

[sawyen]

Owned... Totally owned... Ahh! Whats that shiny hat behind ur back ATI? Give it back now.... You could just wait till R580 comes out..

But the £500 price tag is way way too much..

[geo]

Well, duh! No wonder the SLIFire setup didn't work --you didn't hook the cards together, ninny! This is the *real* purpose of the ATI interconnect model. . .

[buzz_lips]

Owned... Totally owned... Ahh! Whats that shiny hat behind ur back ATI? Give it back now.... You could just wait till R580 comes out..

But the £500 price tag is way way too much..

YUP, pleased with the price I paid for my x1800xt now. I still think after a few drivers these ATI cards will be up there.

[moshpit]

I have to call bunk on your Farcry benchmarks. I have no clue what went wrong there, but this site is the ONLY one showing Farcry giving X1800XT that kind of lead. You attribute it to a bug in the drivers, yet none of the other sites running this bench show this kind of difference. As much as I like this site, I have to say I think something was done wrong somewhere with your Farcry testing that caused even worse performance loss then is normally attributed to that bug as it's NOT consistant with the majority findings.

http://www.pcper.com/article.php?aid...e=expert&pid=5

http://www.guru3d.com/article/Videocards/285/17/

http://www.hothardware.com/hothardwa...ewarticle.aspx (this one won't let me link straight to the page, lameness.but interestingly enough tests with both Nforce 4 SLI board AND Crossfire motherboard handy)

The Radeon X18000 XT was able to outrun the GeForce 7800 GTX 512MB by a couple of frames per second at 1600x1200 when AA and aniso were used, but at the lower resolution the GeForce came out ahead by a similar margin.

If we factor in the performance of the SLI configurations though, NVIDIA's platform is untouchable. A pair of 512Mb GeForce 7800 GTX cards is over 15% faster than a pair of 256MB cards, and remains CPU bound even when FarCry's resolution is ratcheted way up and anti-aliasing and anisotropic filtering are used.

Is simply a couple of several EVERY bit as trustworthy sites showing something VERY different then what you got. I don't question your ethics AT ALL, instead I question your techniques used as they're inconsistant with too many others results. The CPU limitation of this game should NOT allow so high a margin as you saw. Granted, I too question SOMEWHAT where they show a single 7800 GTX 512 beating a X1800XT at all, but the X1800XT's lead should ONLY be marginal at best.

Edit: Even further looking into this, CPU scaling should make a much larger difference then the cards in question then you chart showed. These next results taken from the Guru3D article show CPU scaling is not done with this game even with the mighty FX-57 and beyond.

http://www.guru3d.com/article/Videocards/285/24/

[Tobeman]

Not quite sure what differences SP2 might make, but the hexus setup was using that and the PC Perspective one was not. The motherboard and CPU are the other differences, but I don't think that sort of margin is created by that. Neither state whether the 32bit or 64bit drivers are used either.

Might be worth running the benchmarks on the newer v81.94 or v82.10 betas? Iono

[moshpit]

But the Guru3D review used SP2.

http://www.guru3d.com/article/Videocards/285/14/

Between the Guru3D and Pcper results, there's not alot of difference between em. Not like we're seeing with this review. The differences in the motherboard performance used are VERY slight too, and that's including the Hexus testbed's own Asus A8N-SLI Deluxe (a wonder board btw, I love mine ). There's a driver revision performance issue plagueing this review and this review only. It could be the combination of the vid driver used mixing poorly with some other system/sound/other driver being used perhaps would be my first guess.

32 bit only tests were done in all of the mentioned tests. Not one review stated they were using x64 windows.

Edit: Another point by looking at the Guru3D results would be that the CPU limitation didn't weaken until 1600x1200 in Farcry and then left the GTX512 firmly in the lead. While I still question how Nvidia caught ATI in this very traditionally ATI-centric game, enough sites are showing this same or similar results to know it's probably true. Mayhap deeper driver optimizations or cheats should be looked for, I don't know. But it's there, Nvidia wins Farcry along WITH all the other tests now. I have to say that until Steve has a chance to respond and explain this inconsistancy, I have to say the Farcry benchmarks as presented are deeply flawed and should have been left out with that much questionability about those results evident.

Edit #2: Further investigation leads me to believe this performance issue which only makes itself evident in Farcry for you, Steve, could be down to something as simple as having messed up and allowed the Asus A8N-SLI Deluxe auto detect your timings. I'm using the Corsair XMS3200LL Pro, VERY similar to the memory you used and a couple of points I've found is that the bios incorrectly detects my timings as 2.5-4-3-6 2T which is nuts. Manually setting 2-3-2-6 1T makes Farcry leap forward in performance on my rig. I get another SMALL but noticable increase by loosening up the last timing from 6 to 8 even tho the rams rated at 6. Something that has been noticed by several other hardware sites mods/reviewers has been that the A64 memory controller runs smoother with a native 7 or 8 setting to that last number instead of the default faster number. Not sure what's causing this, but it DOES seem evident in E6 stepping CPU's more then any of the others from my understanding and is true for sure on my X2 4400+ E6 rev. But how this would affect the Nvidia card and not the ATI card I cannot even begin to guess so this theory may be WAY off base...

[Tobeman]

True, and I know very little about benchmarking *slips away from this thread*

I know I can't afford either anyway (), but I've still got my Asus EN7800GTX TOP Edition 256mb sitting in its box waiting to be used while I save up for a processor

[moshpit]

And now for some complete BS for you guys, I don't know WHAT tom's is trying to feed the world with these benches, but they're so far off as to be CRAZY showing the GTX512 STOMPING all over the X1800XT. This is the opposite side of the spectrum and is showing a severe weakness in the setup for the ATI card's testing procedures.

http://graphics.tomshardware.com/gra...tx_512-09.html

[Hans Voralberg]

Havent u notice that Tom's hardware has been a bit dubious since 2003 ??

[moshpit]

Heck, they've been dubious longer then that...

I just wanted to show what UNETHICAL benchmarking looks like. VERY different then what's going on in this review where a simple technical mistake slanted the results unexpectedly.

[CompGeek]

Take a look at this though:
When they reviewed the x1800 XT they showed the x850 beating the GTX... which was total BS in ATi's favor...

[neonplanet40]

This card overclocks well too Hmmm wonder what its like on water......

[CompGeek]

This card overclocks well too Hmmm wonder what its like on water......

Wet

[moshpit]

Heh, I only really posted Tom's Hardware's results mostly as the joke they are. I haven't been a reader of Tom's in many years now. Van brought me back for a VERY short time to read his stuffs, but mostly for a little entertainment, not because he was particularly that good. I agreed with his AMD bias, but hated how he harped SOOO much on it that it ruined ANY chance he had of impartial reporting. I even sometimes got a little embarrassed he was on the pro-AMD'ers side. He wasn't best booster of our credability...