GeForce 2 Pro: admission of deceit by nVidia?

[Marty]

The GF2 pro, as you can read in Anand's review, features the same core clock as the vanilla GF2. It claims the same 1600Mtps fillrate, and thus the same performance. Now the question, does the fact that they are releasing a more expensive product with faster memory and unchanged performanced claims count as an admission that the GTS did not perform up to nVidia's claims, and that they, in effect, lied to the consumer. Opinions?

Marty

 

[sd]

I think its a loser! More attempts to hide delay of NV20.

 

[Zucchini]

It is faster if you read the review all the way through. Thats about all that matters anyways. Think of it as a budget ultra.

 

[Viper GTS]

The 1600 mtps fill rate is under ideal circumstances, assuming an infinite memory bandwidth. Everyone knows that the GFX series of GPU's generally hit their memory bandwidth limit before their theoretical fillrate limitations.

The GF2 Pro significantly increases the memory bandwidth, thus allowing the identically clocked GPU to really stretch it's legs.

It IS a faster product, one that better utilizes the sheer power of the GF2 GPU.

Viper GTS

 

[Pocatello]

Well, the NV20 will cost around $500 to $400, while the GTS Pro cost around $330-350 (probably cheaper when the NV20 arrives). I don't think the NV20 will be that great of a leap over the Pro in performance, while it cost more.

 

[NFS4]

<< Well, the NV20 will cost around $500 to $400, while the GTS Pro cost around $330-350 (probably cheaper when the NV20 arrives). I don't think the NV20 will be that great of a leap over the Pro in performance, while it cost more. >>


The Herc Pro can be had for under $300 and the Inno3D Pro can be had for $277. And by the time the NV20 gets here, it should be under $250. And there's also the fact that the card can be overclocked to be as fast as the Ultra. When overclocked, it tied the Ultra in performance:


<< Regardless of any inadequate cooling methods, the GeForce2 Pro was very willing and able to overclock. Reaching a maximum speed of 230 MHz in the core and 465 MHz in the memory, the GeForce2 Pro actually surpassed GeForce2 Ultra speeds when overclocked. The performance increase as a result of this overclocking is nothing to be taken lightly: although the CPU bottleneck at 640x480x32 results in no performance difference between the standard card and the overclocked card, at 1024x768x32 the overclocked GeForce2 Pro gains a 6% speed boost and at 1600x1200x32 is able to boast a 21% performance increase. Running at 57 FPS when overclocked, the GeForce2 Pro was actually able to tie the GeForce2 Ultra. >>


I think that a lot of you people are missing the beauty of this card. You get an Ultra for a LOT less money.

 

[FordLorider]

It also has a DDR CORE!

haha, nobody cought that.

I was just kidding

 

[Marty]

I'm not disputing that it is faster. All I am saying is that nVidia released one card which was purported to deliver 1600Mtps, and is now releasing another one, newer and more expensive, with the exact same performance claims. Does this represent an acknowledgement that the GTS never really delivered what nVidia claimed it did? I know that many people are well informed enough to figure out that the GTS was bandwidth limited, but I'm sure a lot more people bought the hype.

Marty

 

[BurnItDwn]

its no defeat ... its a good move by nvidia

as faster ram is becomming more available ... they figure ... hey .. since we can get this better ram now ... and its not much more expensive ... lets stick it on our boards and call em PRO ...

the original gts was great .. but the memmory bandwidth was a bottleneck

the core didnt really need a boost ... but increasing the memmory speed makes a big difference at higher resolutions.

Nvidia also understands that not everybody has $400 to $500 to shell out on a vid card, and so they add another item to their list of choices.

you know how ram has progressed from 12 and 10ns pc66 modules to 10 and 8ns pc100 modules ... so on and so forth ..

the ram marked is everchanging ...

when nvidia first started making the gts .. there just wasnt any ram out there capable of the speeds the gf2 pro is capable of.

since then ... ram technology has been developing further.

Rather then using only the old ram ... nvidia has decided to utilize the additional technology with the GF2 Ultra ... they might as well make a GF2 pro too ...

NV20 is delayed, yes ... but think about it ... is the competition threatening them at all.

ATI radeon is good ... but its not as fast a gf2 pro or ultra
3dfx ... well .. v5 5500 is good for some things ... but its not as fast as gf2
matrox ... their g400 was a great card for its day .. and the g450 is no faster then the g400 .. just smaller core and a few changes here and there ..

so .. as of now ... nvidia can fart around all they want and still be in a position of leadership for development of retail video devices


Nvidia is a buisness ... and they are certainly running like a buisness. They are gonna make as much profits as they can while constantly improving upon their current products ... as well as researching new products.

most people that i know dont buy PC hardware out of hype. they do research ... look at benchmarks ... and all that stuff ... and they ask me

those who do buy it out of hype ... end up with a MAC or one of those silly colored OEM computers .. or something like that.



so in response to

&quot;I'm not disputing that it is faster. All I am saying is that nVidia released one card which was purported to deliver 1600Mtps, and is now releasing another one, newer and more expensive, with the exact same performance claims. Does this represent an acknowledgement that the GTS never really delivered what nVidia claimed it did? I know that many people are well informed enough to figure out that the GTS was bandwidth limited, but I'm sure a lot more people bought the hype.&quot;

I would say no.

I think Nvidia has released the GF2 pro for the &quot;well informed&quot; type of user

those who are not the &quot;well informed&quot; type of user probably got stuck with some crappy integrated 8mb 2x agp card with no acceration features.

 

[Weyoun]

what marty is getting at, is all the people out there who bought the original GF2 are now seeing this card with, in their eyes, exactly the same performance. How many people out there do you honestly believe know the GF2 is memory bandwidth limited? Not many is my guess in relation to just how many ppl bought the card....

This now strikes a bogus match in their eyes, why release a card that's exactly the same? If they see, and realise this, nVidia can be viewed as CMOT dibbler from their point of view. To us: deciet by nVidia? Defniately not. To the main population who buys the 'bigger=better' card? It smells more than fishy.

This analysis really relies on being able to see from the more average consumer's point of view, and this (i think ) is what marty's getting at, hope this helps

 

[hans007]

its not a defeat, they couldn't get 400mhz ddr ram when the gts2 first came out

 

[BenSkywalker]

&quot;I'm not disputing that it is faster. All I am saying is that nVidia released one card which was purported to deliver 1600Mtps, and is now releasing another one, newer and more expensive, with the exact same performance claims.&quot;

1600x1200 @300FPS is less then the V5's rated MTexel speed, and yet it can't come close to pushing 300FPS running 1600x1200 32bit in Quake3, it has trouble coming within one tenth. What exactly does MTexel have to do with actual performance? Theoretical peak is something entirely different then actual perfromance, and it always has been. If someone is too stupid to figure that out for themselves, then that is there problem. Do you only look at the HP of a car and deem one to be the fastest? Anyone who does is pretty ignorant.

 

[Marty]

The word is deceit, not defeat. Not sure if that is the point some of you were addressing.

The Mtexel rating has been a better indicator of performance previously than it is now. Look at the original GF, it claimed 480Mpps. The TNT-2U claimed 300Mpps. Both pf those claims were much more grounded than the 1600Mpps that NVidia claimed for the GTS.

Whose job is it to find out the facts? NVidia was out there, hyping the whole &quot;giga-texel shader&quot; thing. The majority of people who bought GF2's were not technically adept enough to find out their card's limitations on their own.

What NVidia basically did was release a card that didn't live up to claims, purely for marketing reasons. They are only now getting around to putting &quot;their money where their mouth is.&quot; This is what I am pointing out.

Regarding your car analogy, here's a better one: You are out looking for a car to use on the Bonneville flats, so top speed is all that matters. One manufacturor claims more than double the speed, so you go with that. Only later do you find out that the claimed top speed is only reached when you are going downhill with the wind to your back.

Marty

 

[Sunner]

Marty, I disagree.
The GTS can do 1600 MTexels(Which is not the same as mpps, but thats nitpicking).
This however is in a very hypothetical situation, but its not lying.
There are several factors that add up to the final performance, the two most important today are mem bandwidth and fillrate.
The GTS has ~5.2 GB of bandwidth and 800/1600 MPix/MTex sec fillrate, these are facts and they are stated clearly on nVidia's website, as well as on the boxes of every GTS I've ever seen.

Now of course marketing is gonna focus on the fillrate since that number is extremely impressive, doing anything less would make them a pretty stupid marketing department.

Its no different than if Compaq or Dell had a P3-1 GHz with 64 MB of RAM.
We all know that a P3-750 with 256 MB would run circles around the 1 giger in most cases, but they'd never say that, they'd just be saying how great their comp is cause it has a 1 GHz CPU.
Its called marketing, live with it.

 

[Marty]

My point is that it is far too hypothetical a situation to warrant that claim. See my car analogy.

Marty

 

[Modus]

Marty,

<<Regarding your car analogy, here's a better one: You are out looking for a car to use on the Bonneville flats, so top speed is all that matters.>>

Yeah, too bad there's no such place where the GF2 can go to reach its top speed. Therefore, raw fillrate specs are meaningless and your argument is reduced to inept trolling.

Bye.

Modus

 

[SlimHarpo]

<< It claims the same 1600Mtps fillrate, and thus the same performance >>



It's pretty well known that Nvidia video cards are memory bandwidth limited. Even my lowly MX core will OC to 230 or more (from 166 stock), and it doesn't make one bit of difference. Any increase in memory clock, though, gives a nearly linear performance gain. I think this is more a testament to Nvidia's gpus being ahead of the available memory tech than anything else.

[edit]
I just read your second post, where you acknowledge the point made above, and emphasize their claims to a 1600 MTex fill rate. As for this, I don't know... I suppose it is somewhat deceptive, but isn't this the norm in the industry? Does any card out there actually deliver on its &quot;theoretical&quot; specs?
[/edit]

 

[Noriaki]

nVidia always overhypes everything, what's your point?

Look at the &quot;all new NSR functions!&quot; NSR consists of 7 per-pixel shading effects, 3 or 4 of which a Voodoo2 could do (I mean come on...one of the effects is applying a second texture...wow that's sure new) the only thing all new about them is the name..nVidia hypes the hell out of everything

The card runs at a 200Mhz core clock, and runs with 4 dual texturing pipelings, 200Mhz*4*2 = 1600Mtexels/s

But the whole texel thing is a hype..what the hell is a texel? You only get 1600 if every pixel has a multiple of two textures.
Single texturing games can't make use of the 2nd texture unit (and you can't apply them to different pixels because it's on the same pipeline) so you get 800Mtexel/s if you only use single textures.
If you have 3 textures/pixel then each pixel needs 2 pipelines, and the 2nd unit on the 2nd pipe is unused so you get 1200Mtexels/second
I could continue...but I'm sure you get the idea.

Is it decit, sure if you want to call it that go ahead.
Is it decit becuase they are using faster memory with a core clocked the same...not really no...
ask anyone what sounds faster:
GF2 GTS: 1600Mtexel/s fillrate with 366Mhz DDR memory
or
GF2 Pro: 1600Mtexel/s fillrate with 400Mhz DDR memory

I'm sure they'd pick the pro it's got a bigger &quot;second number&quot; and I'm usre nVidia stamps 400MHZ DDR MEMORY on their box in huge letters.

 

[NFS4]

<< GF2 GTS: 1600Mtexel/s fillrate with 366Mhz DDR memory >>


Actually, it's 333MHz DDR

And my Asus V7700 GeForce 2 GTS does 200/400 so I'm not complaining

 

[Marty]

I'm not trying to troll, I just want to flush this out.

nVidia's argument for buying the GTS was that it was the fastest out there - 1600Mtps. nVidia's argument for buying the GTS is that it is one of the fastest - also 1600Mtps - except this time, it actually does what they say it will. I don't think that this is a defensible position, and that it constitutes an admission of the misrepresentation of the GTS's level of performance.

Marty

 

[SlimHarpo]

You have a point, but I don't think it's really damning. As people have pointed out in this thread, no vid card delivers its theoretical fillrate. None of them really come close, except under the most contrived situations.

The salient point in selling the GTS was that it was &quot;the fastest thing out there&quot;. Well.....it was. Yes, it did not deliver on it's theoretical maximum fill rate, but neither did any of the competition's card. Nvidia's card was still the fastest thing out there (not to say that it was the &quot;best&quot;, of course...)

Now Nvidia puts out a new card that has, as you point out, the same core speed, but a faster memory clock, and claims that it is faster than the GTS. Well.....it is.

I don't really find this &quot;deceptive&quot;, as I don't think they ever made any claim that either card would ever deliver the theoretical fill rate arrived at by multiplying clock speed by pipelines by texturing units. Is ATI deceptive because they advertise a fill rate based on pipelines with three texturing units, one of which is never utilized in today's games? Not really, at least in my book. Buying a vid card based solely on it's theoretical fill rate is just dumb. Maybe there are some people who do this, but they are dumb, and are probably used to getting suckered by now

 

[ICyourNipple]

this whole argument is stupid and ridiculous. i doubt a single card has EVER lived up to the claim specs in real world situations. if you want deceit, look at 3dfx's claim to have 64MB of memory. does it really work like 64MB on a single chip card? of course not. that, IMO, is much more of a lie.

 

[sandorski]

But it does have 64mb of ram! Bad analogy.

This thread really addresses a common problem. Every Nvidia chip has been overhyped, with the first generation not meeting the hype, but then the next does finally live up to it. Deceptive? Yes. Nvidia the only one? No.

Anyone that has owned a S3/Diamond Viper2 knows about deception! Many features never worked correctly, while others were later dropped because S3 couldn't figure out how to get them to work!

ATI has also overhyped their cards in the past, but I don't remember any particular examples, I just remember some ads that gave the impression of gaming ownage when, in fact, they pretty much sucked!

Matrox tried and failed to make gaming cards, but I don't know if they overhyped/misled anyone.

3dfx, to my recollection, has never not delivered what they claim, unless vidcard models are to be included. They are more disliked by many for their refusal to implement certain features. Most of the time they seemed to have made the right choice(IMO), but they have been a little harsh when talking of their competitors. I could be wrong here, due to the fact that I've never had a bad experience with a 3dfx card, but have had with others. So, I maybe a little biased and not see where 3dfx has misled.