NVIDIA GeForce 7800 GTX 512

Printable View

14-11-2005, 03:04 PM
Steve

NVIDIA GeForce 7800 GTX 512

Quote:

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
14-11-2005, 03:21 PM
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
14-11-2005, 04:50 PM
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..
14-11-2005, 05:21 PM
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. . . ;)
14-11-2005, 05:26 PM
buzz_lips

Quote:

Originally Posted by 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..

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.
14-11-2005, 08:17 PM
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)

Quote:

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/
14-11-2005, 09:21 PM
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 :o
14-11-2005, 10:13 PM
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...
14-11-2005, 10:54 PM
Tobeman

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

I know I can't afford either anyway (:D), 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 :o
14-11-2005, 11:25 PM
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
15-11-2005, 12:09 AM
Hans Voralberg

Havent u notice that Tom's hardware has been a bit dubious since 2003 ??
15-11-2005, 12:17 AM
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.
15-11-2005, 12:25 AM
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...
http://graphics.tomshardware.com/gra...s/image013.gif
15-11-2005, 01:10 AM
neonplanet40

This card overclocks well too :) Hmmm wonder what its like on water......
15-11-2005, 01:36 AM
CompGeek

Quote:

Originally Posted by neonplanet40

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

Wet
15-11-2005, 02:54 AM
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...
15-11-2005, 09:30 AM
kempez

The GTX 512 looks to stomp over a card that has taken ATI a LONG time to develop. Nvidia see that they got beat with the x1800xt and then just *boom* Nvidia come back with this: a 512mb card almost identical to the card that they brought out in JUNE, but that goes like hell.

Bit unsure of what was up with the Farcry benchies mean but hey I don't play that all that much ;)

ATI's next gen best turn out to be better :)
15-11-2005, 12:49 PM
kalniel

Nice review. Especially like the graphs and the 'playable settings' report.

Wonder what they got for sticking this in though?

Quote:

Originally Posted by hexus

Neo is the man, and that bastardised line from the Matrix is awesome.

:p

Agreed with the far cry bench.. seems a bit odd.

Nice card though. There's an interesting, if slightly rumour-based and probably inaccurate editorial on the creation of the 7800 gtx 512 here:
http://www.penstarsys.com/previews/g...gtx/index.html

Sounds like the real champs are the foundaries who managed to get better than expected parts out.
15-11-2005, 04:39 PM
Rys

It certainly is a driver bug in Far Cry, which is why I said so. I try not to make this stuff up too much :D

When I get back home (out of the office right now), I'll show the performance on a driver that works. I had to test the NVIDIA boards with 81.89, at least to begin with, for obvious reasons.
15-11-2005, 05:26 PM
ssnzh

Any one know how to eable resolution 1920X1200 at Need for speed Most wanted???

http://www.hexus.net/content/item.ph...key=1131950738

this test artile show us it is possible running 1920X1200 with 7800GTX, if any one know how, please tell me.;)

http://img.hexus.net/v2/graphics_car.../bench-nfs.png
16-11-2005, 08:50 PM
moshpit

Quote:

Originally Posted by Rys

It certainly is a driver bug in Far Cry, which is why I said so. I try not to make this stuff up too much :D

When I get back home (out of the office right now), I'll show the performance on a driver that works. I had to test the NVIDIA boards with 81.89, at least to begin with, for obvious reasons.

I never claimed you made anything up, I'm saying it's blatent that you have something not setup right. The driver excuse is only a partially acceptable answer as Guru3D used the same EXACT driver, same chipset based motherboard, same OS, same vid cards, and got MUCH better results then you for the GTX512 in Farcry. I think this bug is MORE then a driver issue, it's a configuration issue. Sorry bro, but driver bug isn't floating my boat ;)

Why are you suffering for it and not them? Because something they configured right and you didn't. Well, let me rephrase that tho, As I said before, it may NOT be something you configured, but instead something you FORGOT to check, memory timings. That's one possibility, another could be something like PCI latency or any number of configurable settings that may be misbehaving for you. I JUST ordered a BFG 512GTX also, and as soon as it arrives (god only knows when, it's on backorder :( ), I'll test it myself on the same mobo your using. It may take a week or 2 till i get it since I cheaped out and choose ground shipping after the freaking cost of the card couldn't afford 2nd day ;)

I think the review that actually sold me on this card had to be the Guru3D review, thier conclusion tickled me SOOO much I just had to have : "What about Quake 4 then ? A 1600x1200 resolution with 8xAA and 16xAF .. yes, 78 FPS ! Craaazeey ! It's the fellowship of cards, Mordor in your box !" Mordor in your box. Damn, that amused me so damn much I just had to see for myself...
18-11-2005, 12:15 AM
Rys

Quote:

Originally Posted by ssnzh

this test artile show us it is possible running 1920X1200 with 7800GTX, if any one know how, please tell me.;)

I used UniWS. The Widescreen Gaming forums are invaluable if you game on displays like the 2405.
18-11-2005, 12:18 AM
Rys

Quote:

Originally Posted by moshpit

I never claimed you made anything up, I'm saying it's blatent that you have something not setup right. The driver excuse is only a partially acceptable answer as Guru3D used the same EXACT driver, same chipset based motherboard, same OS, same vid cards, and got MUCH better results then you for the GTX512 in Farcry. I think this bug is MORE then a driver issue, it's a configuration issue. Sorry bro, but driver bug isn't floating my boat ;)

There was no system misconfiguration. There's a driver bug when benchmarking using the commandline settings that I use, for the farcry.exe. Simple as that! Guru3D likely don't use the same commandline or recorded demo that I use.
18-11-2005, 02:06 AM
kempez

Quote:

Originally Posted by Rys

I used UniWS. The Widescreen Gaming forums are invaluable if you game on displays like the 2405.

Also Tweakguides are great for it too. Amazing game and driver guides :)
18-11-2005, 10:01 PM
moshpit

Quote:

Originally Posted by Rys

There was no system misconfiguration. There's a driver bug when benchmarking using the commandline settings that I use, for the farcry.exe. Simple as that! Guru3D likely don't use the same commandline or recorded demo that I use.

AH HA then. Okay, NOW it's clearer. If it's simply that then I would think then that it might benefit both of your sites to contact them and compare your results and testing methods since cooperation of that sort would work to improve both sites. I don't know what kinda relations Hexas has with Guru3D, but I would hope that kind of collaberation could be doable as it would certainly make things easier for us end-users to not have to question one site because it's suffering a bug. It could be your method is superior to thiers, or vice-versa, but I would think this big a difference with the same game, same drivers, almost exact same hardware all together, would push you guys to find out or at least explain WHY your methods are bugged and thier's arn't, yet you still insist upon using the bugged method over the working one. This could REALLY help you avoid this in the future I think.

As sombody who does reviews like you should know, it's better to NOT publish questionable results then to pubish them and catch flak for something that is unique to your methods of testing. I'm not coming down on you guys as I ENJOYED your review and found it informative thruout, but I HAVE to say something about those Farcry benches as they're just to far off to justify your using that custom method your using unless you can show that method to be superior and bug free.
20-11-2005, 02:43 PM
geoffcollins

Thanks Hexus

I love the fact that you put graphs up for 1920*1200.

Keep it up!