What was the worst and most unsuccessful video graphic chip(s) ever built!

Hm, eight? You are relying on far outdated information -- that should be 4 fragment pipes with 8 textures total.

 

Ye, but that's what I was pointing out. The NV30 was meant to be a killer card(before and well after ATI released the R300) that could whoop ass even with SM2.0. It didn't.

US

 

Those exceeded PS2.0 spec's were actually driven by NV to Microsoft at one time, resulting in emergence of the PS2.a set, if I remember correctly, being for marketing reasons or else.
Because of some significant architectural limitations (shortage of register space and strict Vec4 format processing), NV30 had to rely on mixed FP16/FP32 code compilation just not to slow down enough to death in DX9. That finally make all of those questionable tactics by NV, like shader replacement and etc.

 

The problem with the poll is that the 'worst' chips aren't necessarily the most unsuccessful ones - Virge's are utterly awful 3D accelerators but just think of the number of laptops and cheap desktops that shipped with them. The same also applies to the likes of the FX 5200 (and other spawn of the NV30). So to select the worst and the most unsuccessful one cannot really include the likes of the FX 5800 or 2900 XT as they sold fairly well - in fact, there is only one "winner" and that's the Voodoo5 6000. It didn't make it to retail and thus has no degree of success against the others.

With regards to the NV30 vs. the R300, it's easy to forget that the former was simply down on performance figures compared to the latter anyway, regardless of its architecture and marketing (5800 Ultra vs 9700 Pro) - e.g. lower pixel processing rate, lower pixel fill rate, and less bandwidth. Even if it was on a level footing with regards to its capabilities, the NV30 would likely still fall behind the R300 in standardised testing (OpenGL driver issues at that time, notwithstanding).

Perhaps the poll should just be "What graphics chip/board do you hate the most?" - for me, it's the Parhelia followed by its 'offspring', the P750 and P650. The latter was so shite that a 3GHz P4 was faster in vertex shading and a Riva TNT2 outperforms it in general gaming. At least Matrox had the common sense to drop any mention of the word 'game' from their products a long time ago.

 

Savage 2000, it never even worked properly after years of development. Also that broke S3's neck for good.

Funny to see all the hate for NV30 If there was not a better competition out there, it would have been regarded as quite good actually. There you see what bad PR and a bunch of haters can do to a product.

 

I think people could have accepted a NV30 that was simply out-performed by a competing product. At worst people would have shrugged their shoulders and waited for the NV35 that soon followed it.

What they couldn't accept was all the months of lies, driver and benchmark cheats, Futuremark attacks, etc, only to end up with a late, overclocked, overhot, overloud product that was still outperformed by R300, and that even Nvidia didn't believe in anymore, and was soon cancelled.

Nvidia could have simply priced NV30 down and moved on, but instead they decided to go on a PR war of spin and lies that (as you can see from this thread) has tainted their reputation even years later.

 

no, the parhelia did bring some expectations, they were living of their G400 Max in worsened versions (G450, G550). G400 Max was at the top competing with TNT2 Ultra and V3 3500, the parhelia was thought as Matrox's return to gaming cards, with inspiring features : first 256bit card, four texture units per pipe, displacement mapping and 16x AA (a tech similar to nvidia's CSAA but lacking the multisampling part). but it was a failure, notably running at around 200MHz rather than 300MHz.

 

Yeah I think the only real failure in the poll is the NV1, IMO. It truly failed in every way, from sales to technology. NV2 never existed as anything anyone could buy. SiS Xabre and XGI Volari could be equal failures though.

 

The Bang! just wasn't that loud enough to leave a significant trail in the people's minds.

This product was actually an attempt for complete gaming platform-on-a-stick (much like Voodoo 1/2), but with a sound and totally incompatible rendering pipeline with the rest of the world.

 

Permedia and NV1

Interesting to see all the criticism of the Permedia. I had a totally different experience, primarily because the drivers were so rock solid. It definitely wasn't the most full featured or fast board, but man that card worked well with everything I ever threw at it, from games to 3d apps.

I would nominate the NV1, perhaps just timing, but didn't it actually run SLOWER than running 3d on the main processor? I think that automatically qualifies a chip for this particular award.

 

Hands down this has to go to Nvidia's NV1.

A potentially groundbreaking card that received little to no support. Card was rendered useless virtually as soon as it was released.

I almost bought a NV1 based card due to some of the SEGA games that were included. I'm so very very very glad I took a pass on it.

NV30 by comparison was an absolutely brilliant and stunning success. Good quality AF. Competitive speed in DX8 but total failure at DX9. With a record breaking loud graphics cooler fan. Although if anyone remembers back to that time. Many overclockers used Delta fans that were even LOUDER for cooling their CPUs. No wonder some people said they didn't notice the sound of the FX 5800.

Honorable mentions go to

Voodoo Rush - Sometimes as fast as a single Voodoo Graphics, but just as often slower. And paired up with horrible 2D chip.

Regards,
SB

 

Stuff like that is why I voted for the NV30 as the worst modern graphics chip. For the older stuff, NV1 is pretty much leader of the pack... A dubious honor, I'd say!

 

I know b3d should present an award to nvidia for the nv1 and demand Jen-Hsun give us an acceptance speech

 

Voodoo Rush was all the disadvantages of split 2D/3D and all the disadvantages of unified 2D/3D rolled into one.

 

The difference is that the 2900 XT at least managed playable performance in DirectX 10 titles, where as the FX proved completely incapable of running DirectX 9 games at acceptable framerates (even Far Cry in early 2004 brought the FX cards to their knees).

I don't actually think the 2900 XT was that bad in terms of performance - on release it performed on par with Nvidia's 8800 GTS 640, and was basically let down by size/power consumption, and the fact it wasn't as consistent as the 8800 GTS.

All it took was a die shrink for the product to be successful (in the form of the 3870), where as the NV30 had more fundamental architectural problems.

 

How would this pan out if the measure was most expensive to design/produce compared to how much they made off that product?

 

And was very playable on a 9800pro

 

I actually picked up a FX 5950 Ultra for $30 for my amazing, retro Abit BF6 440BX mobo + P3 1400 Tualatin. I wanted to play around with the top of the FX line. It and the 9800 Pro (maybe XT) are the fastest cards that will work in a AGP 2x slot. It runs Doom3 very well, and Guild Wars fairly well (with 4X AA too). This is in the GW Eye of the North areas that use normal mapping and other DX9 effects. I have a feeling though that a 9800 Pro would smack it around handily.

At one point, I had it in a Athlon 64 system and tried C&C3. That was ugly and very slow. Serious quality problems. Maybe the game knew the card sucked too much to really play it and dropped some effects completely, or the drivers were going to town on those shaders to make them run "fast". Maybe it simply couldn't do some effects. Who knows. lol.

And yes, you don't want to be stuck with one of these for FarCry. I'm glad that I didn't pay $400 (or whatever it was) for it.

 

Wow... I've expected the NV30 to win here, but certainly not by such a huge margin. Not with contenders like the Rush, Parhelia, Volari, Trident, etc. (note that I'm not including the Virge here, which was remarkably successful).

 

NV30.

As some people have said in this topic, the GPU itself was terribly designed, but what really makes NV30 stand out as the worst GPU ever was the way Nvidia acted. Fooling its customers and the press is something that won't be forgotten by enthusiasts.

I was tempted to say R600, but at least AMD/ATI had the decency to stick by the enthusiast community and be honest to the press. Charging $399 for R600 may have hurt AMD financially, but at least they priced R600 right for the performance it gave.