Goldie's GeForce 4 TI & Geforce FX graphics card thread

Hey all, here my GeForce4 Ti & GeForce FX thread, may as well make a thread of these cards, so I won't go too far off topic in this thread here:

So here goes!

The FX 5700 Ultra can be fine tunes to even beat the 5800 Ultra my friend Mahigan from x-3dfx manged this quite easily.
Mahigan & SWZSSR both passed on to me about why DX9 is so dreadful on even the highest end FX boards to left alone the designs flaws they also have within.

2D looks great. 3D is something new that Nvidia developed. It does 32bit precision for shaders, and can fit two 16bit precision shaders in one 32bit registry entry. It can't do 24bit precision, which is why it loses badly to the R300 series.

The Microsoft specific was 24bit precision. When running 16bit on the FX, you see that it is lower quality. When running 32bit, you get a large performance penalty, this is how the NV30 series looses hard from ATI's R300 Kahn series of graphics chips, ATi didn't have this problem also how and why the just dominated NVIDIA in the DX9 API, as for Ultra Shadow, this was just an acceleration of shadows with the FX series in Doom3, even with that in place, it loses to the R3x0 based GPUs by quite a fair margin and also by a lot more in other cases with FSAA & AF are added in the mix.

The FX 5700 Ultra was built on IBM's 0.13 process, the 5800 Ultra was built on TSMC's 0.13 process. The TSMC 0.13 process packed more transistors tightly. Transistor density: 0.628 Million per mm2 for TSMC. 0.616 Million per mm2 for IBM. More transistors packed per MM2 for TSMC than for IBM. Packing more transistors into a tiny space creates extreme heat. This thermal wall prevents clock speeds from rising as high as they otherwise could.
So you can clock a 5700 Ultra higher than you can a 5800 Ultra, without needing as large or noisy a cooler, which is also a sweet advantage of the 5700 Ultra.

All 5800 Ultra's die, Usually the GPU solder balls that crack and fail, It's also the fact there is a lid that is thermal glued onto the die.
It dries up and starts to cook, a better heatsink can surely help here.
That's why they're rare, because most of them died. Even the one that Pixelpipes got died.
He had to get another one.

But yea SWZSSR & Mahigan surely told me a lot of these FX based cards, there were a lot of things wrong with them when it came to build quality and rendering limitations in API's like DX 9.0a and OpenGL 2.0.
But in DX 8.1 the FX 5700 Ultra and up are great cards same for DX7 , OpenGL 1.3, but also older version of OpenGL and indeed Glide Emulation, latyer versions of OpenGL were quite troublesome and ran miles better on ATi cards durin the timeline of 2002 - 2003 and onward.
So hey at least they still have a good purpose for some things, which is what I want to explore.

If comparing to the GeForce4 Ti 4600, the GeForce FX 5700 Ultra only beats it with Texel Fill Rate;

Here the spec sheet of the GeForce4 Ti-4600:
https://videocardz.net/nvidia-GeForce4-ti-4600

Here the spec sheet of the GeForce FX 5700 Ultra:
https://videocardz.net/nvidia-geforce-fx-5700-ultra

Overall they are quite on par with each other.
---------------------------------------------------------
Hallo zusammen, hier ist mein Thread zu GeForce4 Ti und GeForce FX. Ich dachte mir, ich erstelle einfach mal einen eigenen Thread zu diesen Grafikkarten, damit ich in diesem hier nicht zu sehr vom Thema abweiche:

Also los geht’s!

Die FX 5700 Ultra lässt sich so feinabstimmen, dass sie sogar die 5800 Ultra übertrifft – mein Freund Mahigan von x-3dfx hat das ganz leicht geschafft.
Mahigan und SWZSSR haben mir beide erklärt, warum DX9 selbst auf den leistungsstärksten FX-Karten so furchtbar läuft, ganz abgesehen von den Konstruktionsfehlern, die diese Karten zudem aufweisen.

2D sieht großartig aus. 3D ist eine Neuentwicklung von Nvidia. Es bietet 32-Bit-Präzision für Shader und kann zwei Shader mit 16-Bit-Präzision in einem 32-Bit-Registereintrag unterbringen. Es unterstützt keine 24-Bit-Präzision, weshalb es deutlich gegen die R300-Serie den Kürzeren zieht.

Die Microsoft-spezifische Einstellung war 24-Bit-Präzision. Wenn man auf der FX-Serie mit 16-Bit läuft, sieht man, dass die Qualität schlechter ist. Bei 32-Bit entsteht ein erheblicher Leistungseinbruch – so verliert die NV30-Serie deutlich gegen ATIs R300-Kahn-Serie von Grafikchips. ATI hatte dieses Problem nicht; außerdem ist dies der Grund, warum sie NVIDIA bei der DX9-API einfach dominiert haben. Was Ultra Shadow betrifft, handelte es sich lediglich um eine Beschleunigung der Schatten bei der FX-Serie in „Doom 3“; selbst mit dieser Funktion verliert sie deutlich gegen die R3x0-basierten GPUs und in anderen Fällen, in denen FSAA und AF hinzukommen, sogar noch viel stärker.

Der FX 5700 Ultra wurde im 0,13-Prozess von IBM hergestellt, der 5800 Ultra im 0,13-Prozess von TSMC. Der 0,13-Prozess von TSMC ermöglichte eine höhere Transistordichte. Transistordichte: 0,628 Millionen pro mm² bei TSMC, 0,616 Millionen pro mm² bei IBM. Bei TSMC sind mehr Transistoren pro mm² untergebracht als bei IBM. Die Unterbringung von mehr Transistoren auf kleinstem Raum erzeugt extreme Hitze. Diese thermische Grenze verhindert, dass die Taktraten so hoch steigen, wie sie es sonst könnten.
Daher lässt sich ein 5700 Ultra höher takten als ein 5800 Ultra, ohne dass ein ebenso großer oder lauter Kühler benötigt wird – was ebenfalls ein großer Vorteil des 5700 Ultra ist.

Alle 5800 Ultra gehen kaputt. Meistens sind es die Lötkugeln der GPU, die reißen und ausfallen. Hinzu kommt, dass eine Abdeckung thermisch auf den Chip geklebt ist.
Diese trocknet aus und beginnt zu überhitzen; ein besserer Kühlkörper kann hier sicherlich Abhilfe schaffen.
Deshalb sind sie so selten, weil die meisten von ihnen den Geist aufgegeben haben. Sogar die Karte, die Pixelpipes hatte, ist kaputtgegangen.
Er musste sich eine neue besorgen.

Aber ja, SWZSSR & Mahigan haben mir sicherlich viel über diese FX-basierten Karten erzählt; es gab eine Menge Probleme bei ihnen, was die Verarbeitungsqualität und die Rendering-Einschränkungen bei APIs wie DX 9.0a und OpenGL 2.0 anging.
Aber unter DX 8.1 sind die FX 5700 Ultra und höher großartige Karten, ebenso wie unter DX 7, OpenGL 1.3, aber auch unter älteren OpenGL-Versionen und sogar der Glide-Emulation; spätere OpenGL-Versionen waren ziemlich problematisch und liefen in den Jahren 2002–2003 und danach auf ATi-Karten um Längen besser.

Also hey, zumindest haben manche Dinge noch einen guten Zweck, und genau das möchte ich näher untersuchen.

Im Vergleich zur GeForce4 Ti 4600 übertrifft die GeForce FX 5700 Ultra diese lediglich bei der Texel-Füllrate;

Hier das Datenblatt der GeForce4 Ti 4600:
https://videocardz.net/nvidia-GeForce4-ti-4600

Hier das Datenblatt der GeForce FX 5700 Ultra:
https://videocardz.net/nvidia-geforce-fx-5700-ultra

Insgesamt sind sie ziemlich gleichauf.

-------------------------------
The graphics cards:
----------------
Die Grafikkarten
-------------------------------

Here my Leadtek GeForce4 Ti 4600:
---------------------------------------
Hier ist meine Leadtek GeForce4 Ti 4600:

Leadtek WinFast A250 Ultra TDH AGP 128MB 128Bit DDR Rev.A3 0210
This card is A3 revision and was made in year 2002, during week 10




Here my XFX Pine Group GeForce FX 5700 Ultra:
---------------------------------------
Hier ist meine XFX Pine Group GeForce FX 5700 Ultra:

XFX Pine Group Geforce FX 5700 Ultra AGP 128MB 128Bit GDDR2 Rev.A1 0304
This card is A1 revision and was made in year 2003 during week 04

---------------------------------------
(Sobald diese Karte eintrifft, werde ich diese Fotos durch bessere ersetzen.)

Quote from Marlon

I really like the FX Series. I bought a Terratec Mystify FX 5800 brand new back in the days.

It was a heavy beast and remember the noisy Fan during 3D. But reach 500/1000Mhz without Problems. It was crazy.

I paired the Card with a Palomino 1900+ on a Budget K7S5A Pro.

The FX replaced my 64MB Radeon 9500 @ 8Pipes. Later the Radeon 9800SE replaced the FX because the Fan got on my nerves...

oh damn do you still have your FX 5800 nowadays or did you sell it back then?

As the 5800 Ultra goes I am still curious how they would perform if water cooled or given a better cooler in general, was the core/mem speed of 500Mhz / 1000Mhz DDR really the max speed?

Makes me wonder if people managed to do VRAM mods with 256MB 128Bit GDDR2 with tighter ram timings so higher clocks could be achieved, or even 5700 Ultra's for that matter wit ha similar mod, unless VRAM amount was snot much of an issue back then?

My main graphics cards were the Hercules 3D Prophet 9700 Pro for my Dual AthlonMP 2200+ and the Leadtek Ti 4600 for the Dual P3/EB 1000.
I loved both as well, never thought to go for the FX 5800 Ultra back then, parts of me wish I actually did go for one, just out of the purest of curiosity, ya know, the 3dfx influence them cards had, I miss that old style NVIDIA, they are really not the same company anymore these days.

Hmm quite some very interesting idea's are forming here, thanks for the cool replies!

That video though some cases the Ti-4600 just smashes the FX 5800 like in UT GOTYE 1999 but in Unreal II which uses DX 8.1 the FX 5800 had quite the lead, only when things got busy like combat or indoor area's the Ti 4600 got a lot closer, I think I can get my FX 5700 Ultra at the same performance level all thanks to the lesser transistors and die process by IBM, my friend Mahigan got his FX 57000 Ultra beating the 5800 Ultra so it can be done.

The system I have atm is my AMD AthlonXP 2200+ Thoroughbred-A build a 1.8Ghz AXP should be fine, my Ti 4600 ran nicely in it, I expect the 5700 Ultra doing just fine as well, I could test my 9800 pro in there as well this card will change the face of 3D graphics completely, it's the 9700 Pro done right as it adds the F-Buffer,

Mahigan added the following regarding the ATi Radeon 9800 Pro:

The ATI Radeon 9800 Pro (released in 2003) featured hardware F-Buffer (Fragment Stream FIFO) technology. This on-chip memory buffer allowed the GPU to execute pixel and fragment shaders of near-unlimited length without the severe performance loss traditionally caused by multi-pass rendering

Here my 9800 Pro which I do have for a while
---------------------------------------------------------
Hmm, hier kommen ganz schön viele interessante Ideen zusammen, danke für die tollen Antworten!

In diesem Video schlägt die Ti-4600 die FX 5800 in manchen Fällen regelrecht um Längen, wie zum Beispiel in UT GOTYE 1999, aber in Unreal II, das DX 8.1 nutzt, hatte die FX 5800 einen deutlichen Vorsprung. Erst wenn es hektisch wurde, wie zum Beispiel im Kampf oder in Innenräumen, kam die Ti 4600 deutlich näher heran. Ich glaube, ich kann meine FX 5700 Ultra dank der geringeren Transistoranzahl und des Fertigungsprozesses von IBM auf das gleiche Leistungsniveau bringen – mein Freund Mahigan hat es geschafft, dass seine FX 5700 Ultra die 5800 Ultra schlägt, es ist also möglich.

Das System, das ich derzeit habe, ist mein AMD AthlonXP 2200+ Thoroughbred-A – ein 1,8-GHz-AXP sollte ausreichen; meine Ti 4600 lief darin gut, ich gehe davon aus, dass die 5700 Ultra ebenfalls gut funktionieren wird. Ich könnte dort auch meine 9800 Pro testen – diese Karte wird die 3D-Grafik komplett revolutionieren, sie ist sozusagen die 9700 Pro, wie sie sein sollte, da sie den F-Buffer hinzufügt,

Mahigan fügte bezüglich der ATi Radeon 9800 Pro Folgendes hinzu:

Die ATI Radeon 9800 Pro (veröffentlicht im Jahr 2003) verfügte über die Hardware-F-Buffer-Technologie (Fragment Stream FIFO). Dieser integrierte Speicherpuffer ermöglichte es der GPU, Pixel- und Fragment-Shader von nahezu unbegrenzter Länge auszuführen, ohne die erheblichen Leistungseinbußen, die traditionell durch Multi-Pass-Rendering verursacht wurden.

Hier ist meine 9800 Pro, die ich schon eine Weile habe
Sapphire Atlantis Radeon 9800 Pro AGP 128MB 256Bit DDR Rev.C1 0411

As Pentium 4 Northwood-C goes, well I must admit that most people do forget that if you Overclock these beyond the 4.2Ghz mark they beat all Athlon 64s in gaming, A64s barely OC'ed beyond the 3.2Ghz mark, a P4 HT Northwood-C at 4.2Ghz+ was doable especially in combination with the Intel i875P Canterwood chipset with 2x 1GB PC_4000/4200 or 4400 memory sticks.
This makes these P4 Northwood-C's super fast CPU's, this was Northwood-C's leverage over all single core AMD Athlon 64 CPUs.

The Pentium4 HT Extreme Edition Gallatin CPU's were the best ones, these have unlocked multipliers and 512KB L2 and 2MB L3 cache, beyond the 4.2Ghz mark you can do some amazing stuff with these, sadly hard to find and if you do find one...well hah, just hope you can afford it

But I would personally try and find a P4 HT Northwood-C 3.4Ghz these have 512KB L2 and the 800Mhz FSB and a CPU multiplier of 17x, with DDR-500 memory aka PC-4000 you can reach a clock speed of 4250Mhz, with PC-4400 a speed of 4675Mhz could be reached, good luck keeping up with one of these with even an A64 3700+ @ 2.4Ghz won't come close.
An A64 would need to run 3675Mhz to keep up with these, the fastest default clocked A64 3700+ flies at 2.4Ghz, consider yourself lucky if you even surpass the 3Ghz Mark.

The 3.2Ghz P4 HT Northwood-C's have a Multiplier of 16x, my 3Ghz model has a multiplier of 15x, lower than 3Ghz I would not use.
A few members form x-3dfx recommended me the Pentium4 HT Northwood C due to the much higher OC advantage, but avoiding Prescott all together, Nobody likes Prescott, these use twice the power an create twice the heat, and are barely less than 3% faster, not really worth it.

There are Athlon64 X2's as well, the same for Core 2 Duo's of course you could go for something like those, but most games were single threaded, this is where 4.2Ghz+ P 4 HT Northwood-C's truly shined, you can also disable HT also making them run cooler as well, all the options were there.
---------------------------------------------------------
Was den Pentium 4 Northwood-C angeht, muss ich zugeben, dass die meisten Leute vergessen, dass diese Prozessoren, wenn man sie über die 4,2-GHz-Marke hinaus übertaktet, alle Athlon 64 bei Spielen schlagen. A64-Prozessoren, die kaum über die 3,2-GHz-Marke hinaus übertaktet wurden, war ein P4 HT Northwood-C mit 4,2 GHz+ machbar, insbesondere in Kombination mit dem Intel i875P Canterwood-Chipsatz und 2 x 1 GB PC_4000/4200- oder 4400-Speichermodulen.
Das macht diese P4 Northwood-C-Prozessoren zu superschnellen CPUs – dies war der Vorteil des Northwood-C gegenüber allen Single-Core-AMD-Athlon-64-Prozessoren.

Die Pentium 4 HT Extreme Edition Gallatin-CPUs waren die besten; diese verfügen über entsperrte Multiplikatoren sowie 512 KB L2- und 2 MB L3-Cache. Jenseits der 4,2-GHz-Marke kann man mit ihnen Erstaunliches erreichen – leider sind sie schwer zu finden, und wenn man doch eine findet … nun ja, hah, dann bleibt nur zu hoffen, dass man sie sich leisten kann

Ich persönlich würde aber versuchen, einen P4 HT Northwood-C mit 3,4 GHz zu finden – diese haben 512 KB L2-Cache, einen 800-MHz-FSB und einen CPU-Multiplikator von 17x. Mit DDR-500-Speicher (auch bekannt als PC-4000) lässt sich eine Taktrate von 4250 MHz erreichen, mit PC-4400 könnte man sogar 4.675 MHz erreichen – viel Glück dabei, mit so einem Schritt zu halten; selbst ein A64 3700+ mit 2,4 GHz kommt da nicht annähernd heran.
Ein A64 müsste mit 3.675 MHz laufen, um mit diesen mithalten zu können; der schnellste A64 3700+ mit Standardtakt läuft mit 2,4 GHz – man kann sich schon glücklich schätzen, wenn man überhaupt die 3-GHz-Marke überschreitet.

Die 3,2-GHz-P4-HT-Northwood-C-Prozessoren haben einen Multiplikator von 16x, mein 3-GHz-Modell hat einen Multiplikator von 15x; unter 3 GHz würde ich nichts verwenden.
Einige Mitglieder von x-3dfx haben mir den Pentium 4 HT Northwood C empfohlen, da er deutlich bessere Übertaktungsmöglichkeiten bietet, aber Prescott sollte man komplett meiden – niemand mag Prescott. Diese Prozessoren verbrauchen doppelt so viel Strom, erzeugen doppelt so viel Wärme und sind kaum mehr als 3 % schneller – das lohnt sich wirklich nicht.

Es gibt auch Athlon 64 X2-Prozessoren, ebenso wie Core 2 Duo-Prozessoren – natürlich könnte man sich für so etwas entscheiden, aber die meisten Spiele waren Single-Thread-Anwendungen. Hier glänzten die 4,2 GHz+ P4 HT Northwood-C-Prozessoren wirklich. Man kann HT auch deaktivieren, wodurch sie zudem kühler laufen – alle Optionen standen zur Verfügung.

There I just updated the photos of the XFX Pine Group GeForce FX 5700 Ultra, enjoy!
It's that I bought it last Sunday at like 2:35 PM and it arrived this afternoon at like 6PM! amazing how fast shipping can be even it's from Germany to the Netherlands, this was the fastest shipping I ever had with a package from Germany, my guess is that the German town of "Hanau" is very close to the Dutch border.
---------------------------------------------------------
Ich habe gerade die Fotos der XFX Pine Group GeForce FX 5700 Ultra aktualisiert, viel Spaß damit!
Ich habe sie nämlich letzten Sonntag gegen 14:35 Uhr gekauft und sie kam heute Nachmittag gegen 18:00 Uhr an! Erstaunlich, wie schnell der Versand sein kann, selbst von Deutschland in die Niederlande – das war der schnellste Versand, den ich je mit einem Paket aus Deutschland erlebt habe. Ich vermute, dass die deutsche Stadt „Hanau“ ganz in der Nähe der niederländischen Grenze liegt.

if I want faster I have a Sapphire Atlantis Radeon 9800 Pro and a Leadtek GeForce 6800 GT AGP 256MB GDDR3, those two are worlds in difference stand alone, there are always faster cards, but yea I still think the FX 5700 Ultra is capable for some good things. the Glide Emulation part is one I'd really want to explore with it.

Quote from Marlon

Have fun with your GeForce.

I still got my Pentium 4 Northwood 3.2, could be a nice build.

My Last FX build was a Shuttle XPC with Pentium 4 2.4 HT (yesss 2.4) and a passive OEM 256MB Leadtek 5200U.

Because i hate loud Computers in my age.

It run Doom3 fine but the GeForce needs faster TSOP Memory Modules for my preferred Resolution.

I know my passive Sapphire 9600XT Ultimate could do it much better, but hey its FX.

Display More

Yea this card is deffo a lot of fun, it's not like the might ATi R300KAHN< but it's still quite capable for other games that use DX 7, DX 81 as well as OpenGL 1.0-1.3, plenty of choices there.
Max Payne 1 &2 are on my mind the same for Soldier of Fortune 1, not sure of SoF II uses DX 8.1 or DX 9, Unreal II The Awakening is the ideal single player DX 8.1 game which the NV36-A1 should be able to handle nicely.

I did improve the image quality here by adjusting the Digital Vibrance & Saturation, now the image output is more like that of an ATi card, or well it's close.


Played some Q3A + FSAA x4 & AF x8 flawlessly, I still have a few more games in mind from Star Wars just as Jedi Knight II: Jedi Outcast and Jedi Knight; Jedi Academy for example.

Descent 2 + D2X Rebirth 0.58.1 with D2X retro 1.31 runs masterfully, this mod uses OpenGL ES 1.2 and the NV36-A1 does this fearlessly with FSAA x4 & AF x8 I get around 180fps at 1024x768x32 max settings.
Yea I think I'll keep playing around with this cool card, for now anyway my AMD AthlonXP 2200+ may be the bottleneck, which it is, it's still good enough t enjoy the games I play with it.

I ran 3D Mark 2000 and 2001 SE B330both at 1024x768x16 + Triple Buffering and also at 1024x768x32 with Triple Buffering enabled.

I found out that 3D Mark 2000 defaults to 1024x768x16 with Triple Buffering on but 2001 SE does 1024x768x32 but only Double Buffering enabled, so pan this out more evenly I tested with Double & Triple Buffering on both 3D Marks with16 bit and 32bit renders.

3D Mark 2000 on 5700 Ultra @ 1024x768x16 + Triple Buffering - 9829 pts (Default Test)

3D Mark 2000 on 5700 Ultra @ 1024x768x32 + Triple Buffering - 9606 pts


3D Mark 2001 SE B330 on 5700 Ultra @ 1024x768x16 + Triple Buffering - 9685 pts

3D Mark 2001 SE B330 on 5700 Ultra @ 1024x768x32 + Triple Buffering - 9547 pts


3D Mark 2001 SE B330 on 5700 Ultra @ 1024x768x16 + Double Buffering - 9428 pts

3D Mark 2001 SE B330 on 5700 Ultra @ 1024x768x32 + Double Buffering - 9441 pts (Default Test)

it also seems the results are never really consistent, games will give me the real numbers on what to expect, this is why I never liked 3D Mark as it lacked a consistent result.

Well I found the FX 5600 Ultra too weak compared to the FX 5700 Ultra, and purely for DX 8.1 a FX 5700 Ultra is on par with a TI 4600 but costs much less, 40 Eur for a 5600 Ultra, but nah man not in a billion years, those go for lots more but in the end not even worth my time.
https://www.ebay.ie/sch/i.html?_nkw=FX,+5600+Ultra
Left alone the countless confusing models as well, I paid less for my 5700 Ultra than the most so called "good" FX 5600 Ultra's are going for.

The FX series are weak in DX,9 especially ATi's Radeon 9800 Pro is what I have for full DX9 support these two pics show how ATis R350 demolished the NV30U based card...





FX is a no go for DX9 because FX series lack the needed features to make DX9 a pleasant and welcoming experience, that left alone to make DX9 even doable doable for them, light DX9 renders can work but the heavy stuff like what Far Cry, F.E.A.R. First Encounter Assault Recon, Soldier of Fortune II Double Helix all use, as some few examples the FX 5950 Ultra will do single digits as where the 9800 Pro will do 140+ fps at 1024x768x32, maybe even more all depending what system you use.

So if I want to emulate heavy stuff my 9800 Pro fulfills that jobs with ultimate precision, plus that the R350 has the F-Buffer as well something R300 not even has...

The ATI Radeon 9800 Pro (released in 2003) featured hardware F-Buffer (Fragment Stream FIFO) technology. This on-chip memory buffer allowed the GPU to execute pixel and fragment shaders of near-unlimited length without the severe performance loss traditionally caused by multi-pass rendering

All easy pickings there.

But yea as a collector I still would love to add a 5800 Ultra, 5900 Ultra and 5950 Ultra, all because my reasons to add them to my archive of cool graphics cards despite their negative background.

An other card that can be used for Glide Emulation is the ATi Radeon 8500 AGP 64MB/128MB these have full shader Moel 1.4 support so there is that, even an 8500 LE can do the job after some fine tuning same for Radeon 9100 AGP 128MB DDR, these are much cheaper cards and get the job one just as well.

And i know most of you here don't need emulation inrealtiy and no neither do I, but it's never about "need", because if it was; I would not even be here in the first place.

The word "need" is in existent when it comes to these hobbies, i do what my ind desires and all about finding out if it is all possible or not and only a die hard collector would understand what I am referring to.

I too had a Voodo 5 6000 myself, the FSAA quality is nice, but they lack Aniso-Tropic Filtering aka AF, so even a FX with FSAA x8S or an ATi Radeon with FSAA x6 and either with with AFx8 or even AF x16 enabled, as the Radeons allow , this will always look miles better than any 3dfx card as Direct3D or OpenGL goes, for Glide Emulation I am not sure of AF can also be applied, or will it even work, I wonder, this is something I want to find out myself with my FX 5700 Ultra and 9800 pro even being.

As drivers go, I always stuck with ForceWare 84.21 WHQL because this driver is rock solid all my GeForce cards from Gefrce2 Ultra - GeForce 6800 Ultra, they all ran fine with them and by using the same driver and the same system it is easier for me to compare.


 

As the 5700 Ultra goes I find it quiet compared to most stock coolers I heard out there, my XFX FX 5700 U uses the NVIDIA reference cooler as well.
But I guess you were referring to them OTES Blower like coolers the 5800U & 5950U use hmm yea those ARE loud lol.

Yea 5700 was a indeed a decent move in the right direction, ideal line of cards if you seek an alternative over the TI 4200.


The 5100 / 5200 / 5500 & 5600 Series has this shader layout:

Pixel Shaders: 4
Vertex Shaders: 2
Texture Mapping Units: 4
Render Output Units: 4
(Some 5200 series of cards came with 64Bit & 128Bit DDR)

The 5700 Series has this shader layout:

Pixel Shaders: 4
Vertex Shaders: 3
Texture Mapping Units: 4
Render Output Units: 4
(Some 5700 series of cards came with 64Bit & 128Bit DDR)

The FX 5800 Series / 5900 & 5950 Series has this shader layout:

Pixel Shaders: 4
Vertex Shaders: 3
Texture Mapping Units: 8
Render Output Units: 4
(FX 5800 Series came with 128Bit GDDR2, while 5900 & 5950 series came with 256it DDR)

Here in in the Dutch Retro clubs we nick name the LE the "Lame Edition" and SE the "Shitty Edition" :topmodel Haha WHY
---------------------------
Hier in den Niederländischen Retro-Clubs nennen wir die LE die "Lame Edition“ und die SE die "Shitty Edition“ :topmodel Haha, WARUM

More of the FX lineup specsheets can be found here:

TechPowerUp

www.techpowerup.com

Yea NV40 was a whole turnaround in the right directio they stil lost to ATis R420 once FSAA & AF were combined, but having SM 3.0 was more of a niche than a true advantage it was nice for reso's like 800x600 or even 1024x 768x32 but still I found the 6800 Ultra even too weak to make full use of SM 3.0.

I was usinga Dual AMD OpteronDP 244 CG Sledgehammer build back in 2004 with the ASUS X800XE PE which 666psycho bought from me a few years later i think it was in 2007, not really sure anymore and the XFX Pine Group Geforce 6800 GT AGP Dual DVI @ 6800 Ultra speeds with a slight OC was in my AMD Athlon64 3200+ Clawhammer build, it was nice having best of both worlds though.

Quote from Marlon

I never went this far.

I ended up with Athlon XP-M 2800+, KT600 and 6800 Back in the days. It was a quiet nice Bang for Bug Setup.

I switched to mobile Turion X2 and 8600GT-M after my Desktop hand no Chance for Upgrade.

Btw. I have some X850XT PE's AGP. R300 on steroids.

To be very honest, I preferred the X800 XT PE due to it s nice quiet fan and being a fastest single slot solution, R481 didn't offer much extra less than 5% fasteer than the X800 SXT PE, that extra performance could easily be gained with the X800 XT PE just the same.

I still seek an ASUS AX800 XT PE AGP as well as a Leadtek 6800 Ultra AGP, not easy but I am sure it's not impossible either.
At the end of the month I will make a search topic for these two specific cards