The five worst Nvidia GPUs of all time: Infamous and envious graphics cards

[Admin]

Although Nvidia has been the leader in gaming graphics for most of the 21st century, it has still released plenty of terrible GPUs. Here's our take on the worst five Nvidia GPUs of all time.

The five worst Nvidia GPUs of all time: Infamous and envious graphics cards : Read more

 

[waltc3]

I'll never forget nv30...after a weeks-long Internet "countdown," and more bogus benchmark numbers rumored than you have hair on your head, the "leaf blower" (so named because of the very loud cooler that sounded like a hair dryer) was released, and promptly flopped. Six months after initial shipping to panned reviews everywhere, nVidia pulled the plug and canceled the GPU entirely. 3dfx bankrupted itself on the STB factory in Mexico, which it purchased, apparently without doing due diligence, only to discover that the factory had secured more than $500M in debt that STB did not report to 3dfx. Boom, it was over just that quick. nVidia picked up what was left of 3dfx at a bankruptcy fire sale, and then promptly blamed the failure of nv30 on 3dfx, of course...😉

 

[Makaveli]

Good list.

 

[COLGeek]

With mostly recent GPUs (3 of 5), this seems to leave out a whole bunch of potential stinkers. Sill, great job starting a heated discussion that is sure to follow.

 

[CelicaGT]

I had a GeForce 256 as my first GPU, back when they were the new hot. Having owned many ATi/AMD and Nvidia GPUs over the years I managed to miss all the stinkers (maaaybe the GTX 970 qualifies with the partitioned memory but it was overall a good card) and get most of the best cards. I just picked up a 4070ti under MSRP on a fire sale, time will tell if this was a mistake. So far, so good.

 

[The Beav]

Although Nvidia has been the leader in gaming graphics for most of the 21st century, it has still released plenty of terrible GPUs. Here's our take on the worst five Nvidia GPUs of all time.

The five worst Nvidia GPUs of all time: Infamous and envious graphics cards : Read more

DLSS is proprietary because it works specifically with CUDA cores, it can't be done without them. No other cards have anything as advanced.

Also the 3080 was a great card, the only issue was that miners bought too many and there was a pandemic, otherwise it would have been like any other GPU release, and neither of those issues were caused by Nvidia...

 

[stonecarver]

Around 2005 a dud of a card was the Nvidia 5500. Were the cards in the article being judged on performance of the card or price and availability.

 

[PEnns]

What a strange article.

So, the 3080 was bad because......................people couldn't buy it???

 

[-Fran-]

I'll have to pick on the 3080, as it's not a bad GPU per se. I cringe at the 10GB for that price, but they (kind of) made it less horrible with the 12GB version (even less obtainable XD), but they do perform well. I do remember the power issue at launch though, so yeah, not quite perfect. Now, why do I think it's not a good pick for the list? Simply because the 3050 6GB will exist shortly. I know we can't predict the future and all, but... Come on. The other strong contender to that list is any DDR version of a GDDR GPU they've put out. And, if you wan a specific model, the 1630. That thing is just trash.

Other than that nitpick, the list is pretty spot on, yes. Another addition to the honorable mention could be the 8000 series that just fell off their PCBs. I think those were G92 dies?

Regards.

 

[atomicWAR]

What a strange article.

So, the 3080 was bad because......................people couldn't buy it???

Yeah that one rubbed me a little wrong too. I think I would have put the a either a 4070 or Ti version in its place as they are starved for ram considering the resolutions they aim for. Heck even the 4080 with it high price and its huge gap trailing the 4090 in performance would have worked for 5th.

But for sure I agree with the 5800 FX ultra being supreme trash earning the top spot (I got a 9700 Pro that gen after four gens of Nvidia on my primary rig). But hey they remembered the bring up the Geforce 256 this time even if in passing...It should have made the list of top 5 best Nvidia GPUs if you asked me.

 

[thestryker]

A pretty deserving list outside of the 3080. The 400 series was a really weird stop as the 460 was a really good card for its segment.

I don't think the 3080 should be in this list because the only real knock against it is the 10GB VRAM. The market was godawful, but none of that can really be blamed on nvidia so much as miners and then consumers for keeping the costs up when availability got better. I certainly bought a 6800 XT for a second machine this year over anything nvidia because the $/perf was so much better.

I think the instances of releasing the "same" card with meaningfully difference specs would be far more deserving. The 30 series example of that would be the 3060 8GB (and likely forthcoming 3050 6GB), because not only did it have less VRAM but the bus width drop caused the performance to plummet. There's also the multiple types of VRAM on the 1030, and the 1060 3GB which had fewer cores on top of less VRAM.

 

[ThomasKinsley]

After reading this article I don't know whether to conclude that the author is very young or I'm very old, but since this is the worst Nvidia cards of "all time" there are many cards that truly deserve to be on this list. The GeForce 4 MX series comes to mind (particularly the MX440). That card was so bad that many games of its day included warnings in the recommended and minimum requirements; "Compatible with GeForce 3 and GeForce 4 cards except the GeForce 4 MX440." Of course this meant OEMs received most of those cards and foisted them onto unsuspecting consumers who expected them to run all of the latest and greatest titles. I briefly had the unfortunate experience of owning a PC with an MX440 in '03 or '04 but sent it back upon finding out the video card slot was PCI (not to be confused with PCI-E) instead of AGP. I would have tolerated the card except it generated artifacts and screen tearing in such a dramatic fashion that it was virtually unplayable.

 

[thestryker]

After reading this article I don't know whether to conclude that the author is very young or I'm very old, but since this is the worst Nvidia cards of "all time" there are many cards that truly deserve to be on this list.

Neither! In one of the other articles they mentioned they're purposely keeping things to more recent generations and highlighting the biggest market impact type cards.

The GeForce 4 MX series comes to mind (particularly the MX440). That card was so bad that many games of its day included warnings in the recommended and minimum requirements; "Compatible with GeForce 3 and GeForce 4 cards except the GeForce 4 MX440." Of course this meant OEMs received most of those cards and foisted them onto unsuspecting consumers who expected them to run all of the latest and greatest titles. I briefly had the unfortunate experience of owning a PC with an MX440 in '03 or '04 but sent it back upon finding out the video card slot was PCI (not to be confused with PCI-E) instead of AGP. I would have tolerated the card except it generated artifacts and screen tearing in such a dramatic fashion that it was virtually unplayable.

You're certainly right about the GF4 MX440 or should we call it the GF2 MX with DDR as that's all it really was.

 

[Futility]

Strange not to see a mention of the 7900 GT fiasco.

Nvdia blamed OEMs for aggressive overclock and early artifacting failures of many units. Card manufacturers quietly sent RMA replacements too those affected. I personally received a eVGA 7950 GX2 in lieu of a bad 7900GT.

7900GT models quickly disappeared from stores to be replaced by newly invented GTO and GS lineups.

 

[bit_user]

I'm going to stake out a rather controversial position and claim the only real problem with the RTX 4060 Ti is its price. If the same exact card had been sold as the RTX 4050 Ti and at $100 cheaper, I think it'd be quite popular. The bus width and memory capacity would also make much more sense, since the x50 tier is usually 128-bit and the RTX 3050 also had 8 GB.

Also, I appreciated the dishonorable mention of their proclivity towards proprietary technologies. Good call.

 

[Jagar123]

I'm going to stake out a rather controversial position and claim the only real problem with the RTX 4060 Ti is its price. If the same exact card had been sold as the RTX 4050 Ti and at $100 cheaper, I think it'd be quite popular. The bus width and memory capacity would also make much more sense, since the x50 tier is usually 128-bit and the RTX 3050 had 8 GB.

It's always about price. The price of these cards are terrible for what they offer hardware wise. Price to performance is always king and the 4000 series mostly failed. AMD is only slightly less so pulling the same shenanigans via rasterization performance.

 

[bit_user]

It's always about price.

No, some of the cards were bad due to high failure rates, noise, inconsistent performance, etc. The GTX 970 even brought on a class action lawsuit, on that last point.

Price to performance is always king and the 4000 series mostly failed. AMD is only slightly less so

I did make a similar comment on the "Worst AMD cards" article, BTW.

 

[thestryker]

I'm going to stake out a rather controversial position and claim the only real problem with the RTX 4060 Ti is its price. If the same exact card had been sold as the RTX 4050 Ti and at $100 cheaper, I think it'd be quite popular. The bus width and memory capacity would also make much more sense, since the x50 tier is usually 128-bit and the RTX 3050 had 8 GB.

I don't think this is necessarily controversial as it felt like everything below the 4080 should have been around $100 cheaper and dropped a tier. I agree completely that most people would have been happy with this card at $300.

The main thing I wonder about is how much of an impact the poor pricing this generation is going to have. While the sub-$300 market isn't my normal purchase place the fact that we don't really have new good options in the $100-250 range feels really bad.

 

[bloodroses]

There's also the defective 8600m fiasco that led to manufacturers such as Apple and Dell suing Nvidia.
Dont forget the 970 issue either where performance would plummet when more than 3.5GB was used out of its 4GB.

 

[Thunder64]

After reading this article I don't know whether to conclude that the author is very young or I'm very old, but since this is the worst Nvidia cards of "all time" there are many cards that truly deserve to be on this list. The GeForce 4 MX series comes to mind (particularly the MX440). That card was so bad that many games of its day included warnings in the recommended and minimum requirements; "Compatible with GeForce 3 and GeForce 4 cards except the GeForce 4 MX440." Of course this meant OEMs received most of those cards and foisted them onto unsuspecting consumers who expected them to run all of the latest and greatest titles. I briefly had the unfortunate experience of owning a PC with an MX440 in '03 or '04 but sent it back upon finding out the video card slot was PCI (not to be confused with PCI-E) instead of AGP. I would have tolerated the card except it generated artifacts and screen tearing in such a dramatic fashion that it was virtually unplayable.

I cam here just to comment about the Geforce 4 MX series. Easily deserves a spot or at least an honorable mention. There is a difference between not being about to buy a card near MSRP and a card that is a complete turd and basically a Geforce 2 rebrand.

 

[Co BIY]

Seems weird to bash the last three generations of Nvidia cards when they have established a near monopoly in gaming GPUs and one of their biggest problems is that employees are getting lazy because they are paid so much and their stock is valued "to the moon" without irony.

No Nvidia fan boy here - this post written using an Intel iGPU.

 

[jlake3]

There's also the defective 8600m fiasco that led to manufacturers such as Apple and Dell suing Nvidia.
Dont forget the 970 issue either where performance would plummet when more than 3.5GB was used out of its 4GB.

As a GTX 970 owner I can't say I was thrilled about that, but in practice I don't recall it being something I bumped up against on a regular basis. I'm sure it burned some people, but in my experience I ran into other bottlenecks before I ran into VRAM issues the driver couldn't manage.

As for the picks in the article...

The 3080 I was going to defend because the whole market was scalped to heck and back during that time, and most of the 30-series seemed like decent cards if you managed to get one, BUT THEN I remembered that the 10gb and 12gb versions had different core configurations and clockspeeds, in addition to the 30-series problem of not having enough VRAM. Selling two cards that have the same name but different VRAM amounts, but are actually different in more ways than just the VRAM, that definitely deserves a spot in the hall of shame.

4060Ti I think deserves it. You can say it's just priced wrong and/or named wrong or whatever, but Nvidia did what they did there, and it really seems like a 60-class contender in some ways saddled with a 50-class memory system.

For the 2080/Turing I probably would have done 2060(/Turing) for not being powerful enough to take advantage of it's namesake new feature. I bought my RX 5700 at that time and I remember the extra cost for RT performance that was unimpressive even at less than Ultra settings not making any sense to me personally.

Can't argue with the Thermi, and the FX 5800 is before my time. Dishonorable mention is good. I've got a bone to pick with a few other Nvidia cards, but if there's a focus on cards that are both recent and had significant impact on the market, I totally understand the omission. The Titan Z was dumb, but older and of little impact because it was so out to lunch. The GTX 1630 is overpriced and should not be a GTX, but no one cares and no one is buying it. GT 210 I dislike even as a non-gaming video adapter because they kept selling it well after iGPUs passed it in performance and standards, and I think they wanted good money for that junk up to and maybe even beyond the architecture getting dropped the main drivers.

 

[Thunder64]

As a GTX 970 owner I can't say I was thrilled about that, but in practice I don't recall it being something I bumped up against on a regular basis. I'm sure it burned some people, but in my experience I ran into other bottlenecks before I ran into VRAM issues the driver couldn't manage.

As for the picks in the article...

The 3080 I was going to defend because the whole market was scalped to heck and back during that time, and most of the 30-series seemed like decent cards if you managed to get one, BUT THEN I remembered that the 10gb and 12gb versions had different core configurations and clockspeeds, in addition to the 30-series problem of not having enough VRAM. Selling two cards that have the same name but different VRAM amounts, but are actually different in more ways than just the VRAM, that definitely deserves a spot in the hall of shame.

4060Ti I think deserves it. You can say it's just priced wrong and/or named wrong or whatever, but Nvidia did what they did there, and it really seems like a 60-class contender in some ways saddled with a 50-class memory system.

For the 2080/Turing I probably would have done 2060(/Turing) for not being powerful enough to take advantage of it's namesake new feature. I bought my RX 5700 at that time and I remember the extra cost for RT performance that was unimpressive even at less than Ultra settings not making any sense to me personally.

Can't argue with the Thermi, and the FX 5800 is before my time. Dishonorable mention is good. I've got a bone to pick with a few other Nvidia cards, but if there's a focus on cards that are both recent and had significant impact on the market, I totally understand the omission. The Titan Z was dumb, but older and of little impact because it was so out to lunch. The GTX 1630 is overpriced and should not be a GTX, but no one cares and no one is buying it. GT 210 I dislike even as a non-gaming video adapter because they kept selling it well after iGPUs passed it in performance and standards, and I think they wanted good money for that junk up to and maybe even beyond the architecture getting dropped the main drivers.

The FX series was hot garbage in anything DirectX 9. I posted something like this in another one of these threads but all it takes is a picture (worth a 1000 words). Keep in mind the Radeon 9000 series came out BEFORE the FX series.

 

[bit_user]

Seems weird to bash the last three generations of Nvidia cards when they have established a near monopoly in gaming GPUs and one of their biggest problems is that employees are getting lazy because they are paid so much and their stock is valued "to the moon" without irony.

But this article is clearly written from a consumer gaming perspective, yet the valuation you're talking about is really being driven by the datacenter use of their AI-oriented products.

I agree that their last 3 generations were at least decent in all regards, other than price. RTX 2000 does jump out as delivering lackluster generational gains (as well as starting the climb on power consumption), but instead it introduced ray tracing and tensor cores to the masses. I think if they were priced a bit better, their underwhelming raster performance gains could've been overlooked. I don't think there was a real stinker among them.

 

[bit_user]

As a GTX 970 owner I can't say I was thrilled about that, but in practice I don't recall it being something I bumped up against on a regular basis. I'm sure it burned some people, but in my experience I ran into other bottlenecks before I ran into VRAM issues the driver couldn't manage.

What I recall reading about it suggested that it was less of an issue than how it was often portrayed. Games had to use almost the GPU's whole memory capacity, before they'd hit the slow memory bank.

Selling two cards that have the same name but different VRAM amounts, but are actually different in more ways than just the VRAM, that definitely deserves a spot in the hall of shame.

I hear that, but just want to point out that Intel is doing almost the same thing, with the A770 16 GB having faster core clocks and memory clocks than the 8 GB version.

4060Ti I think deserves it. You can say it's just priced wrong and/or named wrong or whatever, but Nvidia did what they did there, and it really seems like a 60-class contender in some ways saddled with a 50-class memory system.

It's not just the memory subsystem of the RTX 4060 Ti that's 50-class. Compare the CUDA cores, too! The 3060 Ti has 4864, while the 4060 Ti has only 4352!

It seems clear to me Nvidia originally planed it to be the RTX 4050 Ti or maybe the RTX 4060, but they had to upgrade it a tier because GPU prices collapsed and they couldn't justify charging so much for the original tier it was meant to be in, but their increasing chip costs prevented them from selling it at a lower price. Remember: they'd have done their original wafer buys for the RTX 4000 series during the worst days of the chip shortage, when TSMC had them over a barrel.