Geforce 6800 vs. Radeon 9800 Pro

[JBT]

Originally posted by: AMDHardcoreFan
well just to pee off all you folks Im an ATI Fanatic (for Video Cards) and will stick to my 9800 Pro

good choice the 9800 pro is an excellent card

 

[Matthias99]

Originally posted by: JBT
The GT is ussully about 110 dollars more not quite "a few bucks" to most people. $290 at newegg for the 6800 vs 399.99 or more for a GT at any store.

You can say the same thing about the 6800NU versus the 9800Pro -- $190-200 versus $290. And the difference between $300 and $200 is a 50% increase, while from $300 to $400 is only a 33% increase -- proportionally, the GT is a better incremental value, as the performance gap between it and the 6800NU is larger than the gap between the 6800NU and 9800Pro. That is, the 6800GT costs 33% more than the 6800NU and is 20-30% faster than it, while the 6800NU costs 50% more than the 9800Pro and is only 15-30% faster (see benchmark results above).

Basically, IMO, if you're willing to spend $300 already, you should save/wait until you can get a 6800GT (or see if ATI releases an underclocked 16-pipe X800XT) -- it's only a matter of time before they're at the $300-350 pricepoint. If you absolutely can't wait (say, you've got a GeForce2 right now and you can't play any recent games well), I'd suggest stepping up to the 9800Pro. Keep the $100, wait for prices to fall, and six months from now you'll still get a decent resale value on the 9800Pro if you want to step up to a 6800- or X800-based card. But that's just my opinion.

 

[Oreo]

Originally posted by: apoppin

Originally posted by: HeroOfPellinor
Wow, talk about video card insecurity. Apoppin, you're so defensive about your card you still haven't figured out that we're discussing the 9800 Pro and the 6800...not the 9800XT. The extra memory on the 9800XT may help it maintain a trivial lead over the 6800 in some benchmarks, but that memory and speed won't be there on the 9800 Pro.

You're the one with ADD. i AREADY covered this :roll:



The 9800XT vs 6800 is a Known". We SEE benchs. BY EXTRAPOLATION (we KNOW how much faster the Xt is over the 9800pro) we can judge the performance of the 6800 vs the 9800p.

I could care less about my old card. I just want people to realize the 6800 is not that huge a leap over OLD technology because it is (really) crippled (compared to the "GT").

don't read extra into my posts.

BTW, What card do YOU have?

Are you fighting for "your" card or what is your deal anyway? Well if you are no wonder you are acting like an insecure kid. We are discussing VIDEOCARDS here, so don't take it so personal. You aren't a better person because you have a fast videocard. Do you agree? Personally I own a 9700 Pro and I don't really have any face to loose by claiming that one card is faster than the other, I'm just stating the cold facts.

That review you posted is comparing the 9800XT vs the 6800 wich isn't as good a comparison as the 9700 Pro vs the 6800 since they both have 128MB ram. Just extrapolate the 9700 Pro scores by 15% and you'll get a good estimate of the 9800Pro performance.

So to end this pointless discussion the 6800 is potentially more than 50% faster than the 9800 Pro depending on the circumstances (being the game, settings and rest of the system). The card itself actually is more than 50% faster (this is the cold truth) just that it doesn't get the opportunity to stretch it's legs in all situations. Alot of the time it is only 10-40% faster but that doesn't change the fact that the card itself is more than 50% as fast. If you still don't understand what I'm saying you need to learn some more basics about benchmarking before getting in to discussions you don't understand.

 

[Matthias99]

Originally posted by: Oreo
That review you posted is comparing the 9800XT vs the 6800 wich isn't as good a comparison as the 9700 Pro vs the 6800 since they both have 128MB ram. Just extrapolate the 9700 Pro scores by 15% and you'll get a good estimate of the 9800Pro performance.

I'm not sure that it's the RAM that's making the difference. The 9800Pro is just about smack in between the 9700Pro and 9800XT (around 10% faster than the former and 10% slower than the latter) in just about every benchmark I've ever seen. If you're gonna claim the RAM makes that much of a difference, provide some benches to prove it.

So to end this pointless discussion the 6800 is potentially more than 50% faster than the 9800 Pro depending on the circumstances (being the game, settings and rest of the system). The card itself actually is more than 50% faster (this is the cold truth) just that it doesn't get the opportunity to stretch it's legs in all situations. Alot of the time it is only 10-40% faster but that doesn't change the fact that the card itself is more than 50% as fast. If you still don't understand what I'm saying you need to learn some more basics about benchmarking before getting in to discussions you don't understand.

You're going to need to back this statement up. If the card was really "more than 50% faster", it would be 50% faster in all GPU-limited situations, which does not appear to be the case. You're gonna have to make a stronger case than waving your hands at some bechmarks and declaring it "more than 50% as fast".

 

[Oreo]

Originally posted by: Matthias99
You're going to need to back this statement up. If the card was really "more than 50% faster", it would be 50% faster in all GPU-limited situations, which does not appear to be the case. You're gonna have to make a stronger case than waving your hands at some bechmarks and declaring it "more than 50% as fast".

Haven't I been backing up this statement the entire time?! What don't you understand? You saw the benchmarks that were played at 1600*1200, didn't you? I'm sure that if you look up the technical specs of each card they confirm that the 6800 is more than 50% as fast as the 9800Pro (just in raw numbers).

 

[HeroOfPellinor]

Originally posted by: Matthias99

Originally posted by: Oreo
That review you posted is comparing the 9800XT vs the 6800 wich isn't as good a comparison as the 9700 Pro vs the 6800 since they both have 128MB ram. Just extrapolate the 9700 Pro scores by 15% and you'll get a good estimate of the 9800Pro performance.

I'm not sure that it's the RAM that's making the difference. The 9800Pro is just about smack in between the 9700Pro and 9800XT (around 10% faster than the former and 10% slower than the latter) in just about every benchmark I've ever seen. If you're gonna claim the RAM makes that much of a difference, provide some benches to prove it.

Um, the fact that the 9800Pro with 256MB ram can be flashed to a 9800XT ought to tell you something.

 

[Megatomic]

You know, this is what I think. To me it doesn't really matter if card A is 300% faster than card B as long as card B is fast enough to be 100% playable to me in the games I play in.

i.e. If the 9800 Pro gets 50fps avg. at 1280x1024x32 with 2xAA/4xAF and looks great then I don't care in the least if the 6800 gets 150fps avg. at the same settings. Well, and as long as the minimum fps doesn't dip below 30...

But I know this thread is how guys with e-pen0s problems deal with their issues. Continue as you were.

 

[sharq]

Both cards suck. Buy a SIS or a Matrox.

 

[HeroOfPellinor]

Originally posted by: Megatomic
You know, this is what I think. To me it doesn't really matter if card A is 300% faster than card B as long as card B is fast enough to be 100% playable to me in the games I play in.

i.e. If the 9800 Pro gets 50fps avg. at 1280x1024x32 with 2xAA/4xAF and looks great then I don't care in the least if the 6800 gets 150fps avg. at the same settings. Well, and as long as the minimum fps doesn't dip below 30...

But I know this thread is how guys with e-pen0s problems deal with their issues. Continue as you were.

The argument is over value. Getting a faster and more powerful card today means not needing to upgrade for a year or two or more.

 

[Wahsapa]

-wonders if you guys notice hes getting a regular 6800, not a GT- that being said, save some money and get a 9800pro, or spend the extra money and get the 6800GT, but i think you should NOT get the regular 12pipe 6800.

 

[Oreo]

Originally posted by: Wahsapa
-wonders if you guys notice hes getting a regular 6800, not a GT- that being said, save some money and get a 9800pro, or spend the extra money and get the 6800GT, but i think you should NOT get the regular 12pipe 6800.

I agree. If you're going to spend $300+ for a card you should get a 6800GT.

 

[ToBeMe]

Buy the 6800GT.........it's worth the extra $$$ and will allow you a further time until the need for another upgrade as you will be in the latest generation cards now instead of a generation behind.

 

[Megatomic]

Originally posted by: HeroOfPellinor
The argument is over value. Getting a faster and more powerful card today means not needing to upgrade for a year or two or more.

Maybe. In rare cases you may be right. But I'd be willing to bet $$$ that most gamers that frequent AT forums upgrade more frequently than that. In the last 12 months I have had maybe 7 different video cards in my computer. And I plan to get the best of this current generation of cards when the NV50/R500 come out. Here's real value, getting a card that premiered at $350 or more for half that.

Aside from the OP's question and a few other responses, the majority of this thread has been an e-pen0s contest.

 

[HeroOfPellinor]

Originally posted by: Oreo

Originally posted by: Wahsapa
-wonders if you guys notice hes getting a regular 6800, not a GT- that being said, save some money and get a 9800pro, or spend the extra money and get the 6800GT, but i think you should NOT get the regular 12pipe 6800.

I agree. If you're going to spend $300+ for a card you should get a 6800GT.

Sure, but why spend $200 on a 9800 Pro when you can get a 9800 for $150? Or a 9700 Pro for $130? It never ends. The AT value tables all show the 6800 gives you the greatest value for the money right now and I totally concur.

 

[Oreo]

Originally posted by: HeroOfPellinor

Originally posted by: Oreo

Originally posted by: Wahsapa
-wonders if you guys notice hes getting a regular 6800, not a GT- that being said, save some money and get a 9800pro, or spend the extra money and get the 6800GT, but i think you should NOT get the regular 12pipe 6800.

I agree. If you're going to spend $300+ for a card you should get a 6800GT.

Sure, but why spend $200 on a 9800 Pro when you can get a 9800 for $150? Or a 9700 Pro for $130? It never ends. The AT value tables all show the 6800 gives you the greatest value for the money right now and I totally concur.

Ok, but MY CHOICE would be the GT inspite of the value table. Just my personal preference.

 

[HeroOfPellinor]

Originally posted by: Megatomic

Originally posted by: HeroOfPellinor
The argument is over value. Getting a faster and more powerful card today means not needing to upgrade for a year or two or more.

Maybe. In rare cases you may be right. But I'd be willing to bet $$$ that most gamers that frequent AT forums upgrade more frequently than that. In the last 12 months I have had maybe 7 different video cards in my computer. And I plan to get the best of this current generation of cards when the NV50/R500 come out. Here's real value, getting a card that premiered at $350 or more for half that.

Aside from the OP's question and a few other responses, the majority of this thread has been an e-pen0s contest.

We're not discussing a hardcore gamer here. There'd be no question...go buy an 6800 Ultra off Ebay. Hell, a hardcore gamer wouldn't have even asked. This guy, since he's looking at the 9800 Pro, wants to make a good investment and we're trying to help him do so.

 

[VirtualLarry]

Originally posted by: HeroOfPellinor
Sure, but why spend $200 on a 9800 Pro when you can get a 9800 for $150? Or a 9700 Pro for $130? It never ends. The AT value tables all show the 6800 gives you the greatest value for the money right now and I totally concur.

What about a 6800 LE for ~$200? 8 pipes (at least, unconfirmed for now), *and* support for SM3.0.

NV is obviously trying to hit ATI "below the belt", in their high-end mid-range market with this one.

That might actually make the 9800Pro an even better value, if the prices on those have to drop to accomodate those that might otherwise simply choose the 6800LE instead. Or they may simply start to fade off of the market, if they can't be produced/sold at a much lower cost, because of the memory costs.

 

[Megatomic]

Wow, when overclocked the 6800LE is amazing. I'd buy one of those if I could.

 

[Matthias99]

Originally posted by: Oreo

Originally posted by: Matthias99
You're going to need to back this statement up. If the card was really "more than 50% faster", it would be 50% faster in all GPU-limited situations, which does not appear to be the case. You're gonna have to make a stronger case than waving your hands at some bechmarks and declaring it "more than 50% as fast".

Haven't I been backing up this statement the entire time?! What don't you understand? You saw the benchmarks that were played at 1600*1200, didn't you?

I saw a handful of benchmarks where the 6800 scored about 50% faster than the 9800Pro (or, at least, 50% higher than extrapolated numbers that are close to the 9800Pro). That doesn't mean it's "more than 50% faster" across the board, which is what you're implying. There were benches where the 6800 was *slower* than the 9800Pro, too, if you'll recall.

I'm sure that if you look up the technical specs of each card they confirm that the 6800 is more than 50% as fast as the 9800Pro (just in raw numbers).

You're "sure"? Maybe you should "check".

6800:
12 pipes, 325Mhz core, 350Mhz RAM
Fillrate: 3900MPixels/sec.
Bandwidth: 11.2GB/sec.

9800Pro:
8 pipes, 380Mhz core, 340Mhz RAM
Fillrate: 3040MPixels/sec.
Bandwidth: 10.8GB/sec.

That's, uh, let's see... a 28% edge in fillrate, and a 4% edge in memory bandwidth. Not exactly "more than 50% as fast (just in raw numbers)". In fact, the 6800NU should lose even more performance (percentage-wise) with AA and AF at higher resolutions, since it will be limited so badly by its (relatively) slow RAM (the GT and Ultra run GDDR3 at 500Mhz and 550Mhz, respectively). And this is best-case theoretical numbers.

Like I said, you're gonna have to do more than point at some benchmarks and say "Look! It's 50% faster!" if you want to call the 6800 50% faster than the 9800Pro. It's just not, except in a few limited cases.

 

[Oreo]

Originally posted by: Matthias99

Originally posted by: Oreo

Originally posted by: Matthias99
You're going to need to back this statement up. If the card was really "more than 50% faster", it would be 50% faster in all GPU-limited situations, which does not appear to be the case. You're gonna have to make a stronger case than waving your hands at some bechmarks and declaring it "more than 50% as fast".

Haven't I been backing up this statement the entire time?! What don't you understand? You saw the benchmarks that were played at 1600*1200, didn't you?

I saw a handful of benchmarks where the 6800 scored about 50% faster than the 9800Pro (or, at least, 50% higher than extrapolated numbers that are close to the 9800Pro). That doesn't mean it's "more than 50% faster" across the board, which is what you're implying. There were benches where the 6800 was *slower* than the 9800Pro, too, if you'll recall.

I'm sure that if you look up the technical specs of each card they confirm that the 6800 is more than 50% as fast as the 9800Pro (just in raw numbers).

You're "sure"? Maybe you should "check".

6800:
12 pipes, 325Mhz core, 350Mhz RAM
Fillrate: 3900MPixels/sec.
Bandwidth: 11.2GB/sec.

9800Pro:
8 pipes, 380Mhz core, 340Mhz RAM
Fillrate: 3040MPixels/sec.
Bandwidth: 10.8GB/sec.

That's, uh, let's see... a 28% edge in fillrate, and a 4% edge in memory bandwidth. Not exactly "more than 50% as fast (just in raw numbers)". In fact, the 6800NU should lose even more performance (percentage-wise) with AA and AF at higher resolutions, since it will be limited so badly by its (relatively) slow RAM (the GT and Ultra run GDDR3 at 500Mhz and 550Mhz, respectively). And this is best-case theoretical numbers.

Like I said, you're gonna have to do more than point at some benchmarks and say "Look! It's 50% faster!" if you want to call the 6800 50% faster than the 9800Pro. It's just not, except in a few limited cases.

Where have I implied that it is 50% faster across the board? That is exactly the opposite of what I have been saying but you don't seem to want to read what I write, but read what you want to see.

Yeah, you are right that those numbers aren't 50% higher on the 6800 than on the 9800Pro but some of them aren't correct. There's no way they only have 11.2 and 10.8GB/sec memory bandwith. Aren't they up in the 30's? Where did you get those numbers from by the way?

If my calculations are correct their bandwiths are:
9800Pro: 21.8MB/sec
6800NU: 22.4MB/sec

 

[Matthias99]

Originally posted by: Oreo
Where have I implied that it is 50% faster across the board? That is exactly the opposite of what I have been saying but you don't seem to want to read what I write, but read what you want to see.

What? What does this:

The card itself actually is more than 50% faster (this is the cold truth) just that it doesn't get the opportunity to stretch it's legs in all situations. Alot of the time it is only 10-40% faster but that doesn't change the fact that the card itself is more than 50% as fast.

mean if not "the 6800 is at least 50% faster (except where CPU-limited)"? It can't be both "10-40% faster" AND "more than 50% as fast", unless what you're trying to say is that there are only a handful of situations where it's 50% faster -- in which case you just stated it VERY badly.

Yeah, you are right that those numbers aren't 50% higher on the 6800 than on the 9800Pro but some of them aren't correct. There's no way they only have 11.2 and 10.8GB/sec memory bandwith. Aren't they up in the 30's? Where did you get those numbers from by the way?

You're right; they should be double that, since it's DDR.

I calculated it from the clocks on the cards:

350 Million cycles/sec. * 32 bytes (ie, 256 bits) per cycle * 2 (DDR) = 22,400 megabytes/sec. = 22.4GB/sec.

340 Million cycles/sec. * 32 bytes (ie, 256 bits) per cycle * 2 (DDR) = 21,760 megabytes/sec. = 21.8GB/sec.

Where do you *think* they get bandwidth numbers from?

You may be used to seeing numbers reporting Gigabits per second, which would give numbers eight times as high. Or else you've been seeing numbers for the 6800GT/Ultra and the X800 cards (which would be up in the 30s and 40s).

 

[Oreo]

The card itself actually is more than 50% faster (this is the cold truth) just that it doesn't get the opportunity to stretch it's legs in all situations. Alot of the time it is only 10-40% faster but that doesn't change the fact that the card itself is more than 50% as fast.

Alot of the time you will only see 10-40% higher performance. Clear now?

Ok, I'm not sure exactly how much faster the 6800 is than the 9800 Pro in theory (raw numbers, fillrate tests etc) but it still is 50% faster in situations when the only seeming bottleneck is the graphicscard so it seemed like a logical conclusion. The only way to calculate the theoretical performance of a card isn't only the bandwith och fillrate though, is it?

 

[Matthias99]

Originally posted by: Oreo

The card itself actually is more than 50% faster (this is the cold truth) just that it doesn't get the opportunity to stretch it's legs in all situations. Alot of the time it is only 10-40% faster but that doesn't change the fact that the card itself is more than 50% as fast.

Alot of the time you will only see 10-40% higher performance. Clear now?

I understand what you're saying, but you're misinterpreting the benchmarks. You're reasoning in reverse -- if card A is 50% faster than card B, you would expect it to be about 50% faster in the benchmarks. But you can't look at a benchmark (at least not a single benchmark), see that card A is scoring 50% better than card B, and then declare that card A is "50% faster" than card B without qualification (you could say that it performs up to 50% better, but that's it). It could be that you picked a benchmark where card B happens to perform much worse than it does (relative to card A) in other tests. If card A was faster in every benchmark (or at least in ones where it was clear there was not a CPU/system limitation) by ~50%, then yes, I would call it "50% faster" than card B.

Ok, I'm not sure exactly how much faster the 6800 is than the 9800 Pro in theory (raw numbers, fillrate tests etc) but it still is 50% faster in situations when the only seeming bottleneck is the graphicscard so it seemed like a logical conclusion.

But it's not faster in all such situations, which is what you would expect if the hardware was really 50% faster. It seems more likely to me that what you're seeing in these extreme results are cases where the 9800 is being constrained by other factors (such as driver problems, or just game code that runs badly on the 9800) and the 6800 is not.

On a theoretical basis, if both cards are running at their maximum performance, the 6800 should (everything else being equal) be about 30% faster (assuming it's not bandwidth-limited, as it probably is when it's running AA/AF at high resolutions). In *theory*, the only way the 6800 could run 50% faster than the 9800Pro is if the 9800Pro is running at a lower efficiency than the 6800 in a particular benchmark.

The only way to calculate the theoretical performance of a card isn't only the bandwith och fillrate though, is it?

Those are the theoretical numbers that show how fast the card should run -- it cannot possibly draw more pixels per second than the maximum theoretical fillrate, and it cannot transfer more than that much data per second in and out of its onboard memory. Maximum texels/sec. is also important, but both of these cards can do one texel/pipeline/clock, so the pixel and texel fillrates are the same (with the GeForceFX cards, the maximum texel fillrate was double the pixel fillrate because of their 4x2 architecture). There are other things that can impact it, such as efficiency of their AA/AF implementations and how good their pixel and vertex shaders are, but the fillrate and bandwidth are really what determine maximum theoretical performance.

 

[apoppin]

Originally posted by: JBT
The GT is ussully about 110 dollars more not quite "a few bucks" to most people. $290 at newegg for the 6800 vs 399.99 or more for a GT at any store.

These deal die after a few hours . . . but here's a few GT deals:

PNY GeForce 6800GT 323.50 @ Page Computers

PNY 6800 GT Deal, Take Two, @ emicrox.com ($365)

and there's a few more (1@$340) for a "few" bucks more than the 6800standard's $300.

---------------------------------------------------------------------------------

Originally posted by: Oreo

Originally posted by: apoppin

Originally posted by: HeroOfPellinor
Wow, talk about video card insecurity. Apoppin, you're so defensive about your card you still haven't figured out that we're discussing the 9800 Pro and the 6800...not the 9800XT. The extra memory on the 9800XT may help it maintain a trivial lead over the 6800 in some benchmarks, but that memory and speed won't be there on the 9800 Pro.

You're the one with ADD. i AREADY covered this :roll:



The 9800XT vs 6800 is a Known". We SEE benchs. BY EXTRAPOLATION (we KNOW how much faster the Xt is over the 9800pro) we can judge the performance of the 6800 vs the 9800p.

I could care less about my old card. I just want people to realize the 6800 is not that huge a leap over OLD technology because it is (really) crippled (compared to the "GT").

don't read extra into my posts.

BTW, What card do YOU have?

Are you fighting for "your" card or what is your deal anyway? Well if you are no wonder you are acting like an insecure kid. We are discussing VIDEOCARDS here, so don't take it so personal. You aren't a better person because you have a fast videocard. Do you agree? Personally I own a 9700 Pro and I don't really have any face to loose by claiming that one card is faster than the other, I'm just stating the cold facts.

That review you posted is comparing the 9800XT vs the 6800 wich isn't as good a comparison as the 9700 Pro vs the 6800 since they both have 128MB ram. Just extrapolate the 9700 Pro scores by 15% and you'll get a good estimate of the 9800Pro performance.

So to end this pointless discussion the 6800 is potentially more than 50% faster than the 9800 Pro depending on the circumstances (being the game, settings and rest of the system). The card itself actually is more than 50% faster (this is the cold truth) just that it doesn't get the opportunity to stretch it's legs in all situations. Alot of the time it is only 10-40% faster but that doesn't change the fact that the card itself is more than 50% as fast. If you still don't understand what I'm saying you need to learn some more basics about benchmarking before getting in to discussions you don't understand.

Did you have a point Oreo? If you take the 9700Pros vs. 6800 benchs by extrapolation and the 9800xt vs. 6800 you get an even better idea how the Pro will do then just one extrapolation. Just 'cause my way is different from yours' doesn't mean it is any less valid.

You're the one that should read and try to understand before you type.

:roll:

 

[apoppin]

Originally posted by: Oreo

The card itself actually is more than 50% faster (this is the cold truth) just that it doesn't get the opportunity to stretch it's legs in all situations. Alot of the time it is only 10-40% faster but that doesn't change the fact that the card itself is more than 50% as fast.

Alot of the time you will only see 10-40% higher performance. Clear now?

Ok, I'm not sure exactly how much faster the 6800 is than the 9800 Pro in theory (raw numbers, fillrate tests etc) but it still is 50% faster in situations when the only seeming bottleneck is the graphicscard so it seemed like a logical conclusion. The only way to calculate the theoretical performance of a card isn't only the bandwith och fillrate though, is it?

It is 50% faster in only a few situations. In a few situations the XT beats the 6800. So what.

The 6800 might be 25% faster than the 9800Pro across-the-board as it appears to be ~15% faster - overall -than the 9800XT.

and the 6800LE gets beat by the 9800XT across the board
(o/c'ing might be the 6800LE's only "saving grace")

:roll: