FX5900 or 9700pro? 1

[TheGoatMan]

The answer may seem obvious to some of you but I was wondering if the two are even comparible. I run a BFG asylum 5900 non-ultra and my friend has a 9700 pro. He has said nothing but garbage about my 5900 but has yet to out benchmark it. I know benchmarks aren't everything but I'm curious as if he's just blowing steam for the system I just built that put his to shame. Now I ran tests on my computer for both cards. After the scores he started on about image quality and how the ati is far better. Did I buy the wrong card or is he just mad at the fact that my card was faster and half the price?

 

[TheGoatMan]

Oh yeah. I ran aquamark,final fantasy XI,3Dmark01 and 03.

 

[SYAR2003]

Run 3dmark03 with the latest patch that removes
driver optimalizations for it .
Then the test will be more real .

 

[cdogg]

The FX 5900 non-ultra will have a slight edge in GPU clock speed (giving a better fill rate). To tell you the truth, the cards perform similarly in many benchmarks. But regardless of which one edges ahead overall, I doubt you'll be able to really "see" the difference.

Now, it is true that the design of Nvidia cards only allows it to render 16-bit and 32-bit graphic algorithms, even when set at the highest setting. However, ATI cards will stay at 24-bit, allowing for better precision. Keep in mind that at some instances when 32-bit code is being rendered, the Nvidia card may give a better image. At other times when the Nvidia card has to fall back to 16-bit, an ATI card will have the better view. To better understand why this happens and what kind of effects it has on benchmarks, see this article (particularly at John Carmack's comments):

http://www.tomshardware.com/column/20030219/3dmark2003-01.html

If you had the Ultra version, then your card would easily outdo the 9700 Pro in almost every benchmark that really counts. Heck, the FX 5900 Ultra has 1.5 times the bandwidth of the 9700 Pro. But the regular FX 5900 makes it a closer contest, and I haven't seen any real benchmarks on it yet.


~cdogg
[tab]"The secret to creativity is knowing how to hide your sources"
[tab][tab]- A. Einstein

 

[TheGoatMan]

From what I understand 3dmark doesn't really test your cards true abilities. But with the latest drivers I was able to bench my highest. The latest drivers from what I understand are the best quality. Apparantly aquamark is the best one for image benchmarking. I scored 46,640 on aqua and 6148 on 3Dmark03. I also ran 15625 on 2001 and 5222 on final fantasy XI. With the new drivers not the older "cheat" drivers. I scored high with them at first then found out how to run faster with the new drivers.

 

[SYAR2003]

U need to test 3dmark03 with build 340 .
Patch:

http://www.futuremark.com/pressroom/pressreleases/?111103

And the following drivers.

http://www.futuremark.com/community/drivers/?approved

Only then tha cards can be compaired.

Comment:
Parts of the program code have been changed so that possible 3DMark03 specific optimizations in current drivers will not work. 3DMark03 specific optimizations in drivers are against run rules of 3DMark03, because they invalidate the performance measurement results and thus make it impossible to compare performances of different hardware. A list of drivers that have been tested - and confirmed to produce valid 3DMark03 scores - has been published on Futuremark’s website.

 

[TheGoatMan]

Yeah thats what I heard. I'll try that thanks. You know in other forums this is a real touchy subject. Thanks for help.

 

[dakota81]

Hold on, synthetic benchmarks are not a good source of comparison because both ATi & nVidia take the popular benchmarking software & highly optimize their drivers for them. Out in the real world, the 5900U is slightly better that the 9700 & 9800 pro in DirectX8.1 or previous games.

However, when it comes to DirectX 9 performance, the GeForceFX series cards are failing across the boards to perform even close to the comparibly priced Radeon cards. In many of the soon to be released games (like Half Life 2, Tome Radier AOD, etc.) the 5900U is showing performance about on par with a 9600 Pro, and even the latest 5950U cards don't even come close to the 9800's.

Check around, the GeForceFX line is not a good DirectX 9 gpu.

 

[TheGoatMan]

Check out cdoggs post and it will explain why. So as not to jump to conclusion about nvidias credibility.

 

[SYAR2003]

Dakota ! Read my post will you .Buildt 340 no longer
allowes driver optimations .

 

[cdogg]

Dakota81 makes a valid point about optimization. Although steps have been taken to correct it in order to get more accurate benchmark scores, many experts question whether optimizations are necessarily bad.

As with any benchmark, Nvidia and ATI also strive to customize their drivers for games. This has been happening for as long as I can remember. Such optimizations aren't necessarily to find a way to cheat, but instead to improve code and increase FPS.

Although many still wonder about the motives behind the Nvidia controversy over MadOnion's 3dMark2003 benchmark, both ATI and Nvidia have been playing this game for a long time. With this being one exception, usually optimizations make an improvement without sacrificing quality.


~cdogg
[tab]"The secret to creativity is knowing how to hide your sources"
[tab][tab]- A. Einstein

 

[dakota81]

I suppose I should have stated it in a much different manor; one single synthetic benchmark program should never be used to compare two pieces of software. To really know which card is better, you must compare an assortment of benchmarks, becuase like in the currect Radeon vs. GeForceFX, the FX wins out marginally in DirectX 8 & previous software, but the Radeon takes a dominant lead in DirectX software.

The driver optimizations go even further that just synthetic benchmarks too; for example, to try and keep up with the FX cards in Unreal, the Radeon drivers force the use of low resolution textures even if you have high textures selected in the game configuration. Unreal is a widely popular software used for benchmarking, but is this really an improvement for the game?


It's a complicated mess, but the bottom line is, never look at a single benchmark score to be the tell-all factor in any hardware comparison.

 

[TheGoatMan]

You know I have to hand it to you guys for being so nice about the subject. This question is what got everyone in the other forums so pissed off. I think my 5900 is doing a great job and shouldn't worry about benchmarks or whats better. After all I scored it for $200 bucks at best buy. After rebate of course. You see I have this friend who has done nothing but talk trash on it. He dished out $350 for his 9700 pro. It might be a better card but he doesn't bek=lieve my fx was a better value. Any more insight on the subject will be of great help. Thanks guys.

 

[cdogg]

Well, that's about all the insight I have!

"[blue]I have to hand it to you guys for being so nice...This question is what got everyone in the other forums so pissed off[/blue]"

We appreciate the compliment. Dakota, edfair, paparazi, wolluf and many others have been around here for quite some time. Discussions can get heated at times, but we always try our best to keep it professional and listen to other points of view!


If you get the chance, you might want to explore other forums at this site. Each one seems to have its own crowd and different set of experts. You'd be surprised to find out how many old-timers are still around!!


~cdogg
[tab]"The secret to creativity is knowing how to hide your sources"
[tab][tab]- A. Einstein

 

[dakota81]

Here's an interesting editorial I just caught, definately worth reading:

http://www.hardocp.com/article.html?art=NTQ5

 

[TheGoatMan]

I read the article and got more insight on the subject. I remember when it was easier. Way before I had any interest in computers but playing games my parents bought a new system which was a PII 450. They also went and bought a voodoo 3 video card. Now talk about being the envy of the block at the time. I still use that card in the same system and sometimes pop it in my new one to see how it runs. Whatever happened to 3dfx anyways?

 

[dakota81]

Well, couple things caused 3dfx's downfall. The first was they merged with another former graphics card maker STB, this was when the Voodoo3 cards first came out; and then only 3dfx/STB could make Voodoo cards. Well, all the other video card makers (Creative Labs, DiamondMM, etc.) had to switch to nVidia's TNT chip to make cards.

Then came the Voodoo5; a dual gpu behemoth that was expensive and couldn't keep up with the current nVidia products; I think this was about when the first GeForce's came out. People just abandonded 3dfx for the better product.

That was pretty much the end right there. nVidia later on ended up purchasing all the intellectual property of what was left of 3dfx/STB, but never released any Glide drivers for the GeForce cards (I do miss Glide, the effects were so much cooler than Direct3D at the time).

 

[paparazi]

I think comparisons are really only valid on price at the time of purchase only, the reason I say this is because you can quite often buy a higher performing card at a lower price, just because 4 weeks earlier it was replaced by a model with a revised number.
The ATI 9700pro might HAVE BEEN $350 (around £320 when sold in the UK) when current but only 4-5 weeks ago I saw old stock going for under £200 UK, almost identicle to the current price of a Geforce FX5900 non ultra.
So the original question does indeed become more relevant, comparing two cards at the same price point even if there respective model releases were 8months apart.
Martin


Replying helps further our knowledge, without comment leaves us wondering.

 

[TheGoatMan]

My friends is the pro all in wonder so it has held its value pretty well and came with alot of goodies. That he has never used I might add. His only concern was gaming performance so I guess the all in wonder was a bit of overkill considering all the accessories are still in the box after buying it last march. Hes one of those guys who has to have the best even though he won't use it. Like he spent $160 on a thermal take case and hooked up only like 3 fans out of ten.

 

[nlm9802]

exactly. I just checked prices, per pricewatch, an fx5900 is going for 198, 9700 pro, 229.

using the same arguement I could say that my 50$ radeon 9000 64MB was a better buy then a 300$ voodoo5

as far as the original question, there are a lot of other factors to benchmarking than just the video card. turning off hyperthreading can net better than 2000 points on 3dmark2001. obviously dual channel ddr pc3200 will outperform pc2700..
the question is, can you see or "feel" a difference playing a game on his machine vs yours? chances are, probably not, unless other hardware components are different.