6800GT oc'ing.... better

[Illissius]

I'm not basing this purely on fillrate, that was the entire point of the "other performance related things" part of it. What you suggest would indeed be the best way to evaluate it, but seeing as I have access to neither card (nor a website), this is the best I could do. And fillrate is actually pretty relevant, seeing as they're close to even at stock speeds, which happens to be at close to even fillrates as well.

 

[GeneralGrievous]

You keep coming up with that number off the top of your head. It's more like 15-20%, which is substantial.

Show me a review where the GT beats the Pro by 15-20% in a set of benchmarks. Call of Duty is not a "set". As for my 5% number, take a look at firingsquads 1600x1200 4x/8x benchmarks not counting CoD.

If you look at a more recent review using new drivers, you will see that the 6800U even soundly beats the XT in Far Cry,

Several sites have gotten different results than anandtech on this. Even on anandtech's benches the 6800s only performed at that level in Nvidia's 4 handpicked demos.

(last two added for those who decide based on the "what if I'm unreasonably lucky and achieve the highest overclocks I've heard of anywhere" factor)

I think you are a bit conserative on that last Pro clockspeed. If I was able to get an XT to 585, surely a Pro can get that high too.

 

[SickBeast]

Originally posted by: GeneralGrievous

You keep coming up with that number off the top of your head. It's more like 15-20%, which is substantial.

Show me a review where the GT beats the Pro by 15-20% in a set of benchmarks. Call of Duty is not a "set". As for my 5% number, take a look at firingsquads 1600x1200 4x/8x benchmarks not counting CoD.

If you look at a more recent review using new drivers, you will see that the 6800U even soundly beats the XT in Far Cry,

Several sites have gotten different results than anandtech on this. Even on anandtech's benches the 6800s only performed at that level in Nvidia's 4 handpicked demos.

I love how you hand-pick my quotes to skew things in your favor. In any event, what I said makes sense, you just refuse to acknowledge truth and logic. By your own numbers, the X800PRO is at an 11% disadvantage to the GT when both are overclocked (and that's a best-case scenario for the X800PRO). If I had benchmarks to show you I would have done it a long time ago, but I don't. I'm going to bookmark this thread and post benches when they become available, and I'm 99% positive they will make you look very silly.

After re-reading your last post, it's obvious to me that we're not on the same page. You're comparing firingsquad's numbers which are based on NON-OVERCLOCKED cards. I thought that the whole point of this discussion was to decide the faster card when overclocked. I guess I was wrong.

 

[GeneralGrievous]

By your own numbers, the X800PRO is at an 11% disadvantage to the GT when both are overclocked

Where exactly did I give a 11% number? Do you want me to show you the calculations?

CoD: 43% lead
IL2: 12% lead
LOMAC: 5% lead (in all fairness AF was not used which benefits ATI)
SC: -2% lead
TR: 2% lead
UT: even

I'll leave out farcry until they do some sm3 benches. With Cod its a 10% lead. Without, its a 3.4% lead. Substantial, isn't it?

I thought that the whole point of this discussion was to decide the faster card when overclocked. I guess I was wrong.

If cards perform similarly at stock, and overclock similarly (percentagewise, relative to the original clockspeed), its common sense to reason they will be similar overclocked as well.

If as you claim fill rate is not an end all number, why do you say that the Pro needs 25% extra clockspeed to beat an XT? It certainly does look like the XT has some problems utilizing its raw pixel fillrate power.

Now, you say that the XT is faster than the Ultra, which is not the general consensus AFAIK.

Depends on where you look. Best case scenario I've seen for the Ultra is that its right with the XT. But I've only seen that on anand's benches, everywhere else the Ultra gets beat.

 

[Illissius]

I think you are a bit conserative on that last Pro clockspeed. If I was able to get an XT to 585, surely a Pro can get that high too.

460 * 16 = 7360
585 * 12 = 7020

 

[SickBeast]

Originally posted by: GeneralGrievous

By your own numbers, the X800PRO is at an 11% disadvantage to the GT when both are overclocked

Where exactly did I give a 11% number? Do you want me to show you the calculations?

CoD: 43% lead
IL2: 12% lead
LOMAC: 5% lead (in all fairness AF was not used which benefits ATI)
SC: -2% lead
TR: 2% lead
UT: even

I'll leave out farcry until they do some sm3 benches. With Cod its a 10% lead. Without, its a 3.4% lead. Substantial, isn't it?

I thought that the whole point of this discussion was to decide the faster card when overclocked. I guess I was wrong.

If cards perform similarly at stock, and overclock similarly (percentagewise, relative to the original clockspeed), its common sense to reason they will be similar overclocked as well.

If as you claim fill rate is not an end all number, why do you say that the Pro needs 25% extra clockspeed to beat an XT? It certainly does look like the XT has some problems utilizing its raw pixel fillrate power.

Now, you say that the XT is faster than the Ultra, which is not the general consensus AFAIK.

Depends on where you look. Best case scenario I've seen for the Ultra is that its right with the XT. But I've only seen that on anand's benches, everywhere else the Ultra gets beat.

Wow. I'm getting tired of discussing this with you because you make little sense when you post and you're not even understanding what I'm trying to say.

The 11% number is the shortfall of the X800PRO even when overclocked to an outrageous 570mhz in comparison to the X800XT. The X800PRO IS 25% slower than the XT at the same clock speed; give me a break, it has 25% fewer pipelines! On top of this, you replied to my 11% overclocking figure with benchmarks of cards at stock speeds!! I have never once disputed the fact that the GT and the X800PRO are almost identical at stock speeds! Once you overclock, the GT is obviously substantially faster, in my estimation by at least 15-20%.

Your statement which I bolded is particularly telling. For every "megahurt" that the GT is overclocked, it gains 25% over the X800PRO because it has 16-pipelines instead of 12. That's why a GT overclocked 14% to 400mhz is faster than an X800PRO overclocked 14% to 570mhz. The X800PRO can't beat the XT when it's overclocked that high, but the GT CAN. It's the same reason why a 9600XT at 500mhz can't touch a 9700PRO at 325mhz.

As for the 6800U vs. the X800XT, the reviews that you have seen are probably nowhere near current. With the latest drivers, the 6800U seems to have the edge, but that is irrelevant. Most people have come to the conclusion that the two cards are essentially equal. Only the fanboys have been saying that one is vastly better than the other.

 

[SickBeast]

Here's a link you might find interesting.

You're going against 75% of the AT readers here if you think the X800PRO is equal or better than the GT.

 

[GeneralGrievous]

I have never once disputed the fact that the GT and the X800PRO are almost identical at stock speeds! Once you overclock, the GT is obviously substantially faster, in my estimation by at least 15-20%.

Do you realize how ridiculous these statements are? Assuming you can even overclock a GT by 15% (420 core), a Pro would have to LOSE performance with overclocking in order for a GT to beat it by 20%.

The X800PRO IS 25% slower than the XT at the same clock speed; give me a break, it has 25% fewer pipelines!

No it's not.

http://www.firingsquad.com/hardware/ati_radeon_x800/page26.asp

In neither bench does the XT beat the 525/500 Pro by 25%. And that is even with the pro's memory at a lower speed. By your logic the XT should be 33% faster with 33% more pipelines, but its more like 20%.

You're going against 75% of the AT readers here if you think the X800PRO is equal or better than the GT.

I don't disagree with that. At more or less equal performance I'd take the card with the larger feature set (the GT). But if the Pro becomes significantly changer I might change that decision.

It's the same reason why a 9600XT at 500mhz can't touch a 9700PRO at 325mhz.

No, thats because it's still has half the pipelines.

4x500=2000
8*325=2600

There's your raw fillrate numbers. It's not the asme with the Pro and GT, 12 vs 16 pipelines is not comparable to 4 vs 8 pipelines.

 

[SickBeast]

Originally posted by: GeneralGrievous

The X800PRO IS 25% slower than the XT at the same clock speed; give me a break, it has 25% fewer pipelines!

No it's not.

http://www.firingsquad.com/hardware/ati_radeon_x800/page26.asp

In neither bench does the XT beat the 525/500 Pro by 25%. And that is even with the pro's memory at a lower speed. By your logic the XT should be 33% faster with 33% more pipelines, but its more like 20%.

Umm...OK to properly compare those two cards, you need to bench at 1600x1200 w/ 4XAA/8XAF. In those benches, the XT was 36% faster than the PRO in Far Cry and 34% faster in Tomb Raider. In lower resolutions, these cards are CPU limited, so if one is only 10% faster than the other it's due to the CPU and not the graphics card.

You should also note that they only hit 525mhz on the core of their X800PRO which goes directly against your comment that 570mhz on the core is a "conservative estimate", or that it's even possible for that matter.

Also, based on your own link, the X800PRO is 21% slower than the XT, even when it's overclocked. That means it's more than 20% slower than an overclocked GT, which puts it directly in-line with what I said, disproving what you just said here:

Originally posted by: GeneralGrievous

I have never once disputed the fact that the GT and the X800PRO are almost identical at stock speeds! Once you overclock, the GT is obviously substantially faster, in my estimation by at least 15-20%.

Do you realize how ridiculous these statements are? Assuming you can even overclock a GT by 15% (420 core), a Pro would have to LOSE performance with overclocking in order for a GT to beat it by 20%.

You simply don't know how graphics cards work with regards to pipelines. Your own benchmark link proves this to me.

 

[GeneralGrievous]

You should also note that they only hit 525mhz on the core of their X800PRO which goes directly against your comment that 570mhz on the core is a "conservative estimate", or that it's even possible for that matter.

??????????????????????????? If I got 585 core its clearly possible. Search futuremark's ORB and you'll see Pros with 58x core. Clearly, if you are looking for a max possible overclock, only the max number matters?

You simply don't know how graphics cards work with regards to pipelines. Your own benchmark link proves this to me.

Actually I do. you simply don't have any grasp on simple math and logic. I cannot believe you are unable to see the serious logical fallacy in front of you. Anyway, I'm done here. Just go ahead and buy your GT and be astonished when its not 20% faster than a Pro.

 

[SickBeast]

Originally posted by: GeneralGrievous
Just go ahead and buy your GT and be astonished when its not 20% faster than a Pro.

But your own benchmarks confirm that it is!

 

[DKlein]

Not to take sides here, but...

Do you realize how ridiculous these statements are? Assuming you can even overclock a GT by 15% (420 core), a Pro would have to LOSE performance with overclocking in order for a GT to beat it by 20%.

First off a 350MHz core overclocked to 420MHz is a 20% overclock, and it is a speed I have reached FYI

Now, with my limited knowledge of the workings of a graphics card, I would think that a card with 33% (16/12=1.3) more pipelines (the GT) overclocked by 15% would be exactly 20% faster than the other card (4/3 * .15 = .2). So all things being equal the GT would be 20% faster than the Pro at such speeds, were the Pro set at stock speeds. Now if I've missed something please correct me.

 

[stnicralisk]

Think of it this way. Each pipeline is almost its own processor. When you overclock the GT you are overclocking 16 processors. When you overclock an x800 pro you are overclocking 12.