7900GT worse than the 7800gt?

[redbox]

Here are some benches with 1600X1200 on BF2 with 4xaa high quality

7800 GT

and here are the recent posts of the 7900gt with the same settings:

7900 GT

The test set up with the 7900 GT had a better motherboard, ram , and cpu.
Why are the number so far off then?

 

[JBT]

who knows whats wrong with the numbers but either way looking at the specs the 7900 GT is a little bit faster than the old 7800 GTX's.

 

[redbox]

Yes but the numbers don't show us that either the 7800gt was worse than Anandtech led us to believe or these 7900gts are better than Anandtech is telling us.

 

[Topweasel]

The 7900GT should be showing numbers very similar to the 7800GTX as they have the same memory Faster core and same number of piplines. Very little should seperate these two. The 7900GT numbers are the ones most likely to be screwed up.

 

[broly8877]

Or maybe they used a different time demo/benchmark?

 

[Lonyo]

Possibly a different version of the game and demo used, ie: the demo is of a different map.

 

[Golgatha]

Originally posted by: Lonyo
Possibly a different version of the game and demo used, ie: the demo is of a different map.

Possibly different demo ran, different video drivers, and different version of the game. Apples to oranges.

 

[Golgatha]

Here's a better comparison.

http://www.firingsquad.com/hardware/nvi...geforce_7600_gt_performance/page10.asp

http://www.firingsquad.com/hardware/nvi...rce_7900_gt_gtx_performance/page10.asp

63.8 vs. 56.6 FPS at 1600x1200 with all the goodies on.

The X1800XT holds its own at those types of resolutions in most of the games tested in the 7900GT review too. I do wonder if 512MB of RAM is coming into play at those resolutions. I'd like to see a 256MB vs. 512MB X1800XT article with all those same tests ran.

Of particular interest to me are the Quake 4 benches. The X1800XT and 7900Gt switch places substantially after 1280x1024.

http://www.firingsquad.com/hardware/nvi...rce_7900_gt_gtx_performance/page11.asp

 

[redbox]

Originally posted by: Golgatha

Originally posted by: Lonyo
Possibly a different version of the game and demo used, ie: the demo is of a different map.

Possibly different demo ran, different video drivers, and different version of the game. Apples to oranges.

even if that is all true they ran the 7800gt with only 1gb of ram a fx55,and old video drivers. They ran the 7900gt with 2gb something proven to help in BF2, a fx57, and new video drivers. Are you telling me depending on the map and drivers, which i assume to be better on the 7900gt run, they will cause that much difference given the platform the 7900gt was ran on is so much better.

 

[Matthias99]

Originally posted by: redbox
even if that is all true they ran the 7800gt with only 1gb of ram a fx55,and old video drivers. They ran the 7900gt with 2gb something proven to help in BF2, a fx57, and new video drivers. Are you telling me depending on the map and drivers, which i assume to be better on the 7900gt run, they will cause that much difference given the platform the 7900gt was ran on is so much better.

Possibly; depends on what they changed. Different map? Different demo? Different number of players? Different game/driver versions? Did they use FRAPS or built-in benchmarking? They also show the 7800GTX 512MB performing worse than a stock 7800GTX in the earlier article, so clearly something is different.

You really can't compare results between different sets of benchmarks like that, unless the site tells you they ran exactly the same benches. As was said earlier, it's an apples-to-oranges sort of thing.

 

[redbox]

I would like an explantion as to why they are so off, I believe Anandtech doesn't bench BF2 online so players would be at 16 all the time. I looked at the benches of the 7800gt earlier and the new ones, there is a 15fps difference between the two. I just can't see different demo's causing that much of a difference given the new one is ran on a better rig (i.e. 2gb of ram and a faster cpu).

 

[DefRef]

I ran 3DMark06, FEAR, Doom 3 and CS:S VST on Tuesday night with my 7800GT @ 505/1165 and replaced it with a 7900GT CO @ 554/1726 (stock is 500/1500) on Weds. night and got about a 20% boost in the scores. Actual gameplay feels a lot smoother, too.

 

[josh6079]

The issue is not whether the 7900 GT is actually better than the 7800 GT (that's a no-brainer). The fact is that newer drivers, hardware, or even fps markers should not give a worse result as compared to the older ones. The fact that the older technology has shown better results is not an issue with the technology, but has to deal with how anandtech presents its data. Where, or rather, how do they get those numbers?

 

[outbreak721]

I am at a decision between the eVGA Geforce 7800 gt and eVGA Geforce 7900 gt. The 7900 gt had better specs, but what halted me was the high-end performance test it gave on some popular fps games on anandtech's review. With what you guys say, does that mean anandtech's numbers are wrong? So should I just go with the 7900 gt?

 

[josh6079]

In my opinion the numbers that anandtech has presented are questionalble in thier accuracy. Honestly, the 7800GT is going to still give you great performance for your money. Buy one, or two, ride out with it until you have some other paychecks and maybe step up or wait with satisfaction until DX10--that is a different matter though. I would like anandtech to present some information that would really help solve the nonsense numbers. What program were they using to measure fps? If it is any possibility that drivers are the cause, what were they for which tests? Something is just wrong here. I just wonder how far anandtech goes in its research rather than in its advertising and by showing "statistics" at one time only to counter them at a later date is a form of marketing.

 

[GTaudiophile]

This thread has me concerned. Is stepping-up to the 7900GT OC from a 7800GT a mistake???

 

[pibrahim]

Originally posted by: josh6079
The issue is not whether the 7900 GT is actually better than the 7800 GT (that's a no-brainer).

QFT.

The 7900GT is better than the 7800GT, there's no debate. Does it really matter why some of the numbers are scrambled - I doubt you're going to get a definitive answer as t the exact cause of it anyway. Simple fact is that the 7900GT is better, as the vast majority of benchmarks prove.

 

[Hauk]

Have no worries. It's a no brainer as someone said...

 

[Lonyo]

Originally posted by: redbox
I would like an explantion as to why they are so off, I believe Anandtech doesn't bench BF2 online so players would be at 16 all the time. I looked at the benches of the 7800gt earlier and the new ones, there is a 15fps difference between the two. I just can't see different demo's causing that much of a difference given the new one is ran on a better rig (i.e. 2gb of ram and a faster cpu).

Then you are an idiot.
Link to UT2003 demo's, same hardware and everything, different demo gives 25% difference in results

Different game and cards to the 7800/7900GT's, but the principle is exactly the same.

 

[josh6079]

That is some interesting information...but for this topic it is irrelevant. The information you displayed is only true if the hardware is unchanged. In the data between the cards, the 7900 GT had better hardware. I'm not going to argue which card gives better performance statistically, per dollar, or overall; however, if anandtech is going to present data one month and present a set of complete opposite data several months later, then how much of thier research is actually correct? This isn't to worry anyone about steping up...if you want to go ahead, you probably will see improvement in the long run. (Personally, if you already have a 7800GT why not just get another one instead of stepping up? You'll be better off in terms of performance for the dollar.) I just think people should be aware that anandtech has not provided accurate information between the 7800 GT and 7900 GT.

 

[DefRef]

You know, you're sounding like a troll who's looking to bash on the 7900GT on the basis of one set of numbers. (That you signed up a new ID doesn't bode well for your veracity either.) The fact that many other sites show a marked improvement over the 7800GT and other user's posts don't matter - just that AT's numbers mean, SCREW THE 7900GT!!! Hmmmm....

The 7900GT CO is ~$300, same as the 7800GT. For those within the window to step up, it's a no brainer to drop the $35 to get 20% more speed, that can really be felt in games.

 

[DefRef]

dp

 

[josh6079]

I'm not looking to bash the 7900 GT. I think I've even previously stated that if one went with the 7900 GT they would probably be better in the long run (considering DX9 isn't going to run much longer). Yes, it is a better card, and its overclockablity is as promising as it is juicy. Sure, "redbox" started the thread "7900GT worse than 7800gt?", but that was probably to grab attention to people so that they would know about the weird numbers. He even added a question mark in the title as if agreeing that that must be wrong. So looking back at your "Troll" statement, I don't think you understand what I'm saying. THE NUMBERS ARE WRONG. WHY? Yes I'm a junior member as far as forums go in anandtech, but sorry that doesn't meet your extreme credential status. I like your Bush technique in bringing up an entirely different issue and not addressing the topic.

 

[sadffffff]

the possibility of differing demos explains this just fine. especially seeing that the other cards in the bench went down from previous review too.

what youre saying is something to the effect of, "well the demo may be different and tougher on the hardware... but since the system specs and drivers went up a bit, that should exactly cancel out that difference in demo. and thus we should see results as if the demo hadn't changed"

dont you see the obvious flaw in that reasoning? just becuase it didnt work out like you thought... sure it could had worked out so that the increased system specs made up for the demo difference, but, not knowing how much the demo changed, its just as likely that the sys specs made little or no difference. its all apples and oranges man

that said, if you dont know anthing about the demo, its also technically possible that there is a problem with the benches.

a nice lesson in reasoning

 

[redbox]

Thank you all for your comments it has been a learning experience I had never thought that demo's could be that drastically different. I have one more question does anyone know if the numbers Anandtech put up are average, max or min framerates? Also how many runs of the bench do they run.

Again thank you for the comments.