Stagnation in GPU and CPU market (Come AMD, we need you!)

[ShaidarHaran]

Bah. I don't see a need to go on some quest to keep the same two around. For all we know they have been colluding for most of this decade. It would be interesting to see what would happen to the market if AMD died off. I highly doubt Intel would just reign supreme for the rest of time. Out of the ashes comes VIA-XXL!

Sure it would be "interesting" from a purely academic standpoint, but not from a consumer or enthusiast standpoint. x86 is a two-horse race. The only company that could ever stand any chance of competing would be IBM, and I don't think they have any interest in entering this cut-throat market. Pretty much the same story on the graphics front, with the minor exception of a 3rd horse (NV).

 

[John Reynolds]

The Geforce 3 while faster than the Voodoo 5 5500 was absolutely horrible in image quality compared to the V5. IE - 16 bit was faster than V5's dithered 16-bit (22-bit effective) mode but didn't look even remotely as good. But 32 bit color was a huge hit to performance making it roughly the same speed or slower for still worse Image quality (AA modes) while color quality was only marginally better for a huge hit in performance. Especially when you enabled AA on both cards. Add to that, AGAIN absolutely horrible 2D quality and yeah it didn't last very long.

I found the opposite to be true re the GF3/V5 comments. To this day I'd still take a V5 over a GF2, but the GF3 was hands-down a far better gaming card than the V5: features, performance, IQ, etc. And the GF4 really wasn't that much of a leap over a GF3 TI. And I certainly don't understand how pushing 1600x1200 w/32bpp, anisotropic filtering, and 2x AA could possibly be horrible IQ compared to what the V5 would've been limited to in a 2001/2002 game: 10x7, 22bpp, and maybe 2-4x AA with bilinear filtering. No thanks.

IQ is subjective and everyone's entitled to their own opinion, but when that opinion is so removed from reality I'd suggest glasses or meds.

 

[Geeforcer]

Yeah, I am not sure what exactly what went wrong for SB as far as GF3 was concerned (I still have one of those suckers in one of the 6 or so computers around here). Unless he was running some poorly-ported Glide games, to get 5500 to outperform a GF3 would be a hell of a trick.

GF2 Ultra, maybe?

 

[Moloch]

Yeah, I am not sure what exactly what went wrong for SB as far as GF3 was concerned (I still have one of those suckers in one of the 6 or so computers around here). Unless he was running some poorly-ported Glide games, to get 5500 to outperform a GF3 would be a hell of a trick.

GF2 Ultra, maybe?

Even the voodoo 5 6000 couldnt out perform the GF2 ultra, so I think he's altogether full of shit :smile:

 

[no-X]

Voodoo 1 - Oct 96
Voodoo 2 - Jan 98
Voodoo 3 - Mar/Apr 99
Voodoo 4/5 - Jun/Oct 00

About a year and a half between releases.

And where is Voodoo Rush and Banshee?

VG - Q4/96
Rush - Q2/97
V2 - Q4/97
Banshee - Q2/98
V3 - Q1/99
V5 - Q2/00 (+ Velocity)
V4 - Q3/00

 

[ShaidarHaran]

And where is Voodoo Rush and Banshee?

VG - Q4/96
Rush - Q2/97
V2 - Q4/97
Banshee - Q2/98
V3 - Q1/99
V5 - Q2/00 (+ Velocity)
V4 - Q3/00

We've already been over this..... Neither were 3d refreshes that served to increase 3d speed. They were simply attempts at combining 2d and 3d functions.

 

[no-X]

ShaidarHaran: Banshee increased IQ (better post-filter) and speed in many games, especially D3D ones. The fact, that almost all reviews benchmarked using Q2, which used heavy multitextuirng, doesn't mean, that Banshee is slower at average. If you don't believe, just try it

Anyway, single Banshee was capable of 1024*768, V2 wasn't.

 

[ShaidarHaran]

ShaidarHaran: Banshee increased IQ (better post-filter) and speed in many games, especially D3D ones. The fact, that almost all reviews benchmarked using Q2, which used heavy multitextuirng, doesn't mean, that Banshee is slower at average. If you don't believe, just try it

Anyway, single Banshee was capable of 1024*768, V2 wasn't.

Banshee was not a 3d upgrade. Sorry, it just wasn't. It was a niche product. 3dfx's bread and butter was the Voodoo line. V3 was their first true 2d/3d card. The previous attempts were too imperfect with too few customers to be considered upgrades of the regular Voodoo line.

 

[Geeforcer]

There is some big-time confusion regarding the release dates of certain products.

Voodoo 2 was released summer 1998. I think the earliest you could buy one was July of that year.
Voodoo 3 came around April 1999.

So the gap between V2 and V3 is less then a year. It's also worth noting that V2 was being challenged by TNT and Rage Fury cards for supremacy in the mean time.

As such, that did not even come close to paralleling the 12 months (and counting) of G80 dominance.

 

[no-X]

Banshee was not a 3d upgrade. Sorry, it just wasn't. It was a niche product.

V3 was good upgrade for V2-SLI owners? I'm not sure, but Radeon 9800, GeForce 3 Ti500 etc. wasn't good upgrade from previous solutions and no one considered them as inferior solutions.

3dfx's bread and butter was the Voodoo line. V3 was their first true 2d/3d card.

Voodoo 2 was planned to be the last Voodoo and all newer 2D/3D solutions were planned to be named as Banshee. V3 should be named Banshee 2, but Banshee wasn't too successfull, so 3Dfx decided to use old proven name.

Banshee wasn't successfull, but it wasn't generally slower than V2. If all reviews would use only Quake IV for benchmarking, Radeon X1900 wouldn't end better than GF7900.

I used both V2 and Banshee and Banshee really wasn't worse solution.

 

[swaaye]

V3 was good upgrade for V2-SLI owners? I'm not sure, but Radeon 9800, GeForce 3 Ti500 etc. wasn't good upgrade from previous solutions and no one considered them as inferior solutions.


Voodoo 2 was planned to be the last Voodoo and all newer 2D/3D solutions were planned to be named as Banshee. V3 should be named Banshee 2, but Banshee wasn't too successfull, so 3Dfx decided to use old proven name.

Banshee wasn't successfull, but it wasn't generally slower than V2. If all reviews would use only Quake IV for benchmarking, Radeon X1900 wouldn't end better than GF7900.

I used both V2 and Banshee and Banshee really wasn't worse solution.

Banshee would perform about the same or even faster if a game didn't use multiple texture layers. Voodoo2 was clocked at 90 MHz, Banshee 100 MHz. The second TMU on Voodoo2 would just sit idle in single-textured games. I remember Unreal didn't use it until a patch. Multitexturing wasn't even close to being everywhere during Banshee's time. Banshee was quite the value at the time.

TNT was at a major advantage here too. It had twin pixel pipelines instead of a 1 pipe + 2 TMU design. This was an advantage of being single-chip, I'm sure. A Voodoo2 would've needed more chips to go 2x1.

I'm sure V3 has better image quality than V2. Especially for your 2D needs because you don't have to deal with the crappy pass-thru. V3 was an excellent Matrox-challenging 2D card too.

 

[no-X]

One of the biggest difference was the bandwidth between the chip and operational buffers: 0,72GB/s for V2 and up-to 2GB/s for Banshee. Many vendors clocked Banshee a bit higher - 110-115MHz for core and up-to 125MHz for mem, so single-texturing fillrate was 25% higher and F/Z-buffer bandwidth 170% better compared to V2.

As for 2D - Banshee was the only solution, which was comparable to G200 in terms of speed. Even it's 2D output was quite clear for that time - possibly good reference signal filtering circuits. 3Dfx probably learned from mistake with Rush, which had just horrible signal quality - independently on vendor and used 2D akccelerator (AT3D / AT25 / MX86251) - all of them were blurry, many of them produced ghosting artifacts and many MX86251 cards produced nice rainbow artifacts near text and other high-contrast objects).

 

[ShaidarHaran]

There is some big-time confusion regarding the release dates of certain products.

Voodoo 2 was released summer 1998. I think the earliest you could buy one was July of that year.
Voodoo 3 came around April 1999.

So the gap between V2 and V3 is less then a year. It's also worth noting that V2 was being challenged by TNT and Rage Fury cards for supremacy in the mean time.

As such, that did not even come close to paralleling the 12 months (and counting) of G80 dominance.

V2 was released at the very beginning of '98. Google for proof. Probably would be better than just going off one's recollection... There are company PRs from individual AIB partners dating back to Feb/Mar '98 announcing their V2 products, with Creative dating back to Comdex '97 link .

V3 was good upgrade for V2-SLI owners? I'm not sure, but Radeon 9800, GeForce 3 Ti500 etc. wasn't good upgrade from previous solutions and no one considered them as inferior solutions.


Voodoo 2 was planned to be the last Voodoo and all newer 2D/3D solutions were planned to be named as Banshee. V3 should be named Banshee 2, but Banshee wasn't too successfull, so 3Dfx decided to use old proven name.

Banshee wasn't successfull, but it wasn't generally slower than V2. If all reviews would use only Quake IV for benchmarking, Radeon X1900 wouldn't end better than GF7900.

I used both V2 and Banshee and Banshee really wasn't worse solution.

If later cards were to be named Banshee, why was the Banshee itself named Voodoo Banshee?

 

[AlNom]

Voodoo 2 was planned to be the last Voodoo and all newer 2D/3D solutions were planned to be named as Banshee. V3 should be named Banshee 2, but Banshee wasn't too successful, so 3Dfx decided to use old proven name.

If later cards were to be named Banshee, why was the Banshee itself named Voodoo Banshee?

It is a new trademark and back in the days of originality and new naming, they came up with "Banshee" for their 2D/3D "all-in-one" solutions. The Voodoo signifies that it has got Voodoo power - it is brand recognition for 3dfx.

 

[SugarCoat]

And now your post is gone

yep i realized my oversight as soon as i posted and was too lazy to spin it off, sorry

 

[Silent_Buddha]

I found the opposite to be true re the GF3/V5 comments. To this day I'd still take a V5 over a GF2, but the GF3 was hands-down a far better gaming card than the V5: features, performance, IQ, etc. And the GF4 really wasn't that much of a leap over a GF3 TI. And I certainly don't understand how pushing 1600x1200 w/32bpp, anisotropic filtering, and 2x AA could possibly be horrible IQ compared to what the V5 would've been limited to in a 2001/2002 game: 10x7, 22bpp, and maybe 2-4x AA with bilinear filtering. No thanks.

IQ is subjective and everyone's entitled to their own opinion, but when that opinion is so removed from reality I'd suggest glasses or meds.

Oh, there wasn't a single game I could run satisfactory at 1600x1200x32pp with 4xAA on the Geforce 3.

At the time when I was running them. 2xAA on the V5 was far superior to 4xAA on the Geforce 3. Even if run at a lower resolution.

ESPECIALLY, when you took into account the shimmering of Ground Textures as you moved. With the fledgling state of AF in those days, it was still a far cry from the V5 at the time.

This was most evident in EQ1 with levitation on. The constant bobbing up and down meant that the ground was constanly shimmering on the GF3. Likewise edges on Mountains were continuously crawling with jaggies on the Mountains. The V5 5500 took care of all those problems. Even at only 2x AA it was superior to the GF3 with 4x AA.

And if that wasn't enough the 2D quality was a killer. Reading with fine print at 1600x1200 on a 21" CRT with the Geforce 3 was just an exercise in pain compared to even the V5 which wasn't quite as good as the Matrox cards at the time. And this was the major area the Geforce 4 impoved upon compared to the GF3 that convinced me to use it as the main card in a second machine I built for gaming. While using the existing machine with the V5 for existing games that could run acceptably on it.

However, even then the AA and AF quality of the GF4 left a lot to be desired. But at that point the V5 simply could not keep up in modern titles.

So I suppose just like you could call my opinion into question. I could also call your opinion in IQ into question, but what's the point? It's an opinion and different people's vision are affected by different things. Maybe you are the one that needed glasses? Pfffft.

Or maybe, you can accept the fact that we notice different things rather than attempting a mild flame.

Regards,
SB

 

[Geeforcer]

Of course, both GF3 and V5 had rotated grid 2xAA and 4x-8x AF provided a far better texture filtering than 2x super sampling + biliniar....

But, let's not let the facts get in the way of things. GF3 sux amirite? V5 FTW.

 

[Silent_Buddha]

Of course, both GF3 and V5 had rotated grid 2xAA and 4x-8x AF provided a far better texture filtering than 2x super sampling + biliniar....

But, let's not let the facts get in the way of things. GF3 sux amirite? V5 FTW.

Except that neither the Geforce 3 nor even the Geforce 4 had RGSS AA. They did however have Quincunx, but that isn't even worth mention. And they did introduce MULTI-sampling with the Geforce 3 series to replace the Supersampling of the Geforce 2 series.

However their MSAA wasn't (IMO) nearly as good as the quality received from 3dfx's RGSS AA nor was it even better than their own SSAA. However, the speed hit was less so it was a good compromise that ended up being the future of AA.

You did however, make me question my memory so I looked it up just in case my memory was going bonkers on me. This was the first article I ran across.

http://www.anandtech.com/showdoc.aspx?i=1470&p=7

Or as you put it...

But, let's not let the facts get in the way of things.

Regards,
SB

 

[Geeforcer]

I thought I was pretty clear. GF3 and GF4 had 2x rotated grid AA (RGMS) which provided largely identical edge aliasing reduction to 2x RGSS of V5 while AF filtering did a better job of clearing up the textures then 2s SS + bilinear filtering ever could. I realize that this is almost word-for-word repeat of my previous post, but hopefully it will sink in this time.

As far as the "fledgling state of AF" on GF3 that "left a lot to be desired"... you do realize that overall AF image quality provided by GF3 and 4 series would not be matches until R520 came along half a decade later.

 

[Zaphod]

Define 'rotated'. Relative to what?

GF3 sample pattern (2X and QX):

GF4 sample pattern (2X and QX):

R9700 sample pattern (2X):

FX5800 sample pattern (2X):

FX5900 sample pattern (2X):

Source for all images: B3D