These Graphics Cards Shouldn't Have Existed 

What was one called? GeForce 7950GX2? Was that thing any good?

@@cfb36 here are two more, the 9800GX2 (decent card) and the GTX295

@@cfb36 gtx 690

@@lbochtler Also the GTX 690 wich at the time was actually very good. and not that expensive compared buying two separate 680s.

GTX 690?

Wacky times. Kindoff sad, but im also happy that Gpus are so heavily standardised now.

Yes, it also kinda sucked because you had to use crossfire which wasn't supported too well and it used 700 watts of power just on it's own

@@MrTurbo_ 700?! dang... and here we are looking at today's which use less than that thinking it's highly excessive lol

I bought two of the R9 295x2 cards during early Covid for 200 bucks to mess around with and could barely power them with a 1500 watt power supply. OCing would usually trip the overcurrent protection lmao

The 295x2 was utterly stupid and pointless and wasteful, but I am glad such a ridiculous feat of engineering was made.

Have a GT 730 like that. Came with a prebuilt system the rest of was actually good except for it. Always found it strange.

That’s actually it’s intended use case

I have a 630 that still runs all day every in my Plex server. Spent a decade before in a crappy Lenovo workstation and just won't die

I too have a 630, works fine for me as I don't need GPU power.

My XPS 8700 has a GT 630 and has not shown any problems. I didn't even know it has a bad reputation and I use it every day.

I also regret the Matrox Millennium II. I think I got it for CAD reasons, but was not a great gaming card. If I recall it only supported Direct3D while 3DFX and OpenGL was killing it on the gaming front.

Old memories are being flooded into me from CompUSA days. Was the Matrox the one that had a "warning" label promoting how other cards might only be good at 3d games but that their card was good at both 2D and 3D games?

Millennium 1 was great... 2 and Mystique... not

Same.

Yep I had one of those. I had to buy it because I made the mistake of buying a Compaq Deskpro (NLX Form Factor) and basically at the time it was the only card that would fit. It could run Homeworld... just. I supported it with a Voodoo II (? I think) 8MB sister card eventually. When I built my first PC by hand it was like the gods had sung in my direction! Suddenly I was in a land of AMD CPU goodness with "proper" graphics - I know it was an ATI GFX card, but can't remember which one. a Radeon of some sort I think. But for the time it was powerful. The jump from my Compaq to my home build was massive. The Matrox Mill 2 was just a little underpowered for the day. Only 4MB ram, 8 with an upgrade. The Voodoo sister card added "proper" 3d abilities (lets face it - Quake). My timeline is messed up 'cos I'm getting games and tech out of sequence... it was a long time ago!!!!

I know, I freaking LOVE the FuryX. We'd be lucky to see GPUs like it again, they were high end GPUs in a slim package that were just silent at all times.

It was badass in every way... except the VRAM capacity.

@@sergeantsarge7081 made up for it in speed for sure but a couple more gbs of vram it would have been a 1070ti killer

this guy is the only LTG "Host" that i cannot stand. hes always letting his opinion get in the way of facts (Not to mention his jokes suck). the Fury X was a great card.

What other manufacturer/card comes with water cooling? Underrated card

Meanwhile, my Vega56 OCs the HBM2 memory from 800 to 1000mhz flawlessly. Cha-ching!

@@blunderingfool i wished my Vega 56 could do that 😩

Then you didnt read up on it back then. I had both a Fury and Fury X and although the overclocking was very limited, you could overclock the HBM. However, it could only be done in fixed increments (55mhz or somethinf weird) and if your card couldn't handle that next jump, you couldn't OC it. I was lucky enough that both my Fury and Fury X would allow 1 or 2 steps of memory OCing.

@@blunderingfool Vega 56 with 64 bios is just plain amazing. Not particularly efficient however, but damn good for the price

I rocked a R9 fury until I scored a 3060 last year

hhhhh

*SecretLab!*

@@infernaldaedra tf?

@@infernaldaedra ooookay 👀

Cum

Yeah, I knew there was a Voodoo 5, soI was a bit surprised they mentioned the 4 as the last one. The voodoo was a beast but it was overkill for the time with those dual gpus. I had a Riva TNT 2 and while it was good, it couldn't match those, but hey, it was cheaper

@@smokeduv The 4 technically released at the same time the voodoo 5, so it's the just as new but definitely not their best card. I have a Voodoo 5, it's a cool card.

I miss my 5500

I remember Voodoo drivers offering *tons* of settings that could be adjusted. I was mind-blown trying to customize its performance.

Voodoo 5 5500 was released in 2000, but ATI had their Fury MAXX dual GPU card in 1999, so the 3dfx card was not the first in that regard. Voodoo 4 4500 was in fact the last card from 3dfx as it was released a few months after the Voodoo 5 5500. Also, both used the same chips, the VSA-100, only the 5500 had two of them. Finally, the fact that a 5500 would beat a TNT2 (and even the first gen GeForce) did not help it much when it released right after GeForce 2 and ATI's first Radeon, both of which were faster and more advanced (they were DX7 cards with hardware Transform and Lighting, while the Voodoo was DX6). In other words, too little too late and I don't even think that it was cheaper than its competitors.

Remember when Titan's allowed you to choose whether you wanted good FP32 or FP64 performance up to the Titan Black? And then Nvidia was just like, "lol Not anymore. :)"

It may not have been the fastest card, but it did look cool and had great support up until around 2002-3 which was relatively decent for the time. Would've been better off with a GeForce 2 though for sure.

Yeah, the V4 4500 is practically just a single-chip version of it.

@@classic_jam Oh yeah, I would have gotten way more life out of a GeForce 2. I think the card I ended up replacing the V5 with was a Ti 4600.

Still has one and it still works

yeah I still have 2 of them yeah but it was SLI on the card

I remember being intrigued by the All-In-Wonder cards cause of their video inputs, but glad I avoided these headaches.

And here i am gaming on a quadro k5000 and tesla k80 (though the tesla only helps in phys-x games) (And no im not a dumbass that bought a workstation card thinking its better tgen a gaming card due to its price. I use them for scientific computation, 3d modeling and cad, gaming is just something they get to do once in a while.)

Same here and you know what is the real kicker? At the time people where saying that the 1080 was like the 4090 is now: Too hot, too expensive, and way too much card to justify. Yet years down the road that investment has paid off in spades because you bought a great part and now its aged gracefully to a good or ok level part. Those nay sayers bought "just enough" or only ok level parts that now have little useful life in newer titles without dramatic compromises.

@@lbochtler Isn't the K6000 the last Quadro card with Windows XP support, and also (mostly) the last Quadro card to have not-shit FP64 performance?

@@Hybris51129 Too hot? I got a Titan Xp which is basically just a 1080 Ti Special Edition, and it runs amazingly even in modern Blender workloads. lol

My 7yo has my old gtx 1080 in his PC Honestly a bit overkill for a 7 year old, but I wanted to build him a PC for his birthday and I didn't want to sell it and contribute to the environment of people taking advantage of others while the shortage was going full tilt.

I actually had that card when it was new at the time. The Rage 128 was worlds better than the Rage 2 and was actually competitive. The tuner even took a hit from lightning blowing out that portion of the card. Surprisingly the 2d/3d still worked well inside Linux if you didnt install or setup the tuner. I eventually replaced it with a All-in-Wonder Radeon due to the fried tuner for my main Windows machine.

I had this experience with my Riva TNT2. It was a top of the line card for it's time, but was given to me in 2001 for my first PC, by which time it was starting to be obsolescent. It ran games just fine up until Homeworld 2 launched, and I didn't have support for DX8 or 9, which was a limiting factor for me at the time (i forget which one it was). I ended up getting the Radeon 9250 SE not knowing it was basically an entry level display output not worth half a damn, and while it WAS faster than that TNT2, it ended up being unable to run Oblivion at launch just a year or two later. I was lucky my step-dad had bought a 9800 PRO All in Wonder that he was no longer using, because I was able to adopt it as my own for a massive performance upgrade, and my first proper experience with a "high end" card, even though it had already been replaced with the X800 and X850 by then. I still regret getting that 9250 to this day, it was my shortest lived GPU EVER, with even my 9 month old 5700xt outliving it before being replaced with a 2080ti and sold during the GPU shortage to someone building a new rig to survive the pandemic with. I think it only lasted me 6 months before being unable to run a new game on it and getting swapped for that 9800, which it's important to note WE ALREADY HAD, and my step-dad had just left it sitting out near his desk. I didn't see it until after we had wasted $90 on that shitty low end GPU, and asked him if I could have it since mine was garbage. He knew we were going to go get one and literally could have solved it ahead of time lol. I miss those days now, you could wait a year while saving up lunch money change for a parts upgrade, and get double to triple the performance sometimes just by being patient and sticking with the hardware and games you were running decently together at the time. It's not like I couldn't play starcraft just fine on a potato by that point in a pinch xD. Still, even having money sometimes my family would be dumb about stuff and make me wait because I was a kid, while we had hardware sitting around idle gathering dust. Now my step-dad is better about it, he asks me the second he decommissions anything if I want any of it before it gets sent to a refurbisher/recycler to be salvaged and probably resold for cheap to those who can't afford new hardware.

Pretty much all games were single thread back then so you weren't missing much. Core 2 just had insane single thread performance to begin with.

I've had an 4830 paired with an athlon x2 6200+ and 4 gb of ram - that was an amazing budget combination back in the days. A 400 € build that supported me for many years - great times

I am still rocking a vega 64 nitro and it still kicks ass at 1440p

I mean, it WAS a 4K card when it was released, so yeah, it can handle older games in 4K. But I was using mine for 1080p by the time I replaced it with the 7900 XTX.

Could say the same about the Gigabyte Aorus 3080, well minus the decibels but definitely a space heater

I had a pair of hd7970s, and for a short while, tri-fire 7970s. The only game that worked well was battlefield 3 and 4. Which, at the time, was all I played. The hd7000 series was a fantastic value for a lot of years though. I really liked my 7970, I still have it somewhere.

Wait... a NVMe SSD on a GPU?!!?

@@drakethedragon457yep. Had to Google for the model number. It was the Radeon Pro SSG. They advertised it as the first GPU to break the 1TB memory barrier. LOL.

I believe they did a video dedicated to that gpu, it sounds so good! Especially now that vram is at a premium for games

It was the most direct Direct Storage Access feature so far lol.

The 3070ti is still a great card. Just be happy you have such a great pc.

Really? i have the RTX 3060 Ti and this card is really good!

Same here. I spent $800 AUD on an RTX 2060, it was overpriced, but I'm still glad I have it.

Other youtubers already did that.

Remember when Titan's allowed you to choose whether you wanted good FP32 or FP64 performance up to the Titan Black? And then Nvidia was just like, "lol Not anymore. :)"

I can echo these statements. Bought 2 GTX670's and water blocked them and OC the snot out of them. It seemed pretty slick at the time but the micro stuttering and SLI compatibility bugs ruined years of gaming for me.

You legend!

Polaris came AFTER Fiji. In other words, the Fury X was the 300 series. The 400 and 500 series came after.

How was it a stepping stone if they literally decided to ditch the HBM and never use it again in consumer segment products?

@@dark666king stepping stone can also be a learning experience for them

@@dark666king they used it again in the Vega cards and Radeon VII. This was at a time when GDDR5 was very long in the tooth, and VERY power hungry. nVidia had sophisticated memory compression techniques AMD didn't, so they had to use brute force to feed their GPUs (290/390 had a massive 512 bit memory bus which I don't think has been seen again). They hit a memory bottleneck, so HBM was their only way to get more bandwidth in, and they hoped having more cache and WAY higher bandwidth would make up for the lack of VRAM. It kind of did. With newer memory, and their own newer memory compression algorithms, they don't need that super high, raw bandwidth to feed their consumer cards anymore. It wasn't so much a failure, as other technologies more cost effective came along. They still use HBM on some enterprise/compute cards, but games just don't need the bandwidth HBM offered.

@@dark666king Vega 56 and 64, Radeon VII all used HBM

Wasn't the 9700 pro plagued with texture issues and buggy colour pallets?

@@CaptainVKanth Seriously? In fact, the 9700 brought some innovations that later became standard in GPUs. I had a 9600 when the 4600 had a cooler problem and damaged the GPU, and the only problem was the ATI driver at the time, which was very unstable (catalyst if I remember the name). But at that time the problems with GPUs were diverse, such as not correctly recognizing the AGP speed (here a great advantage of the 9700, which was 8x and the 4600, only 4x), fluctuating voltage, memory recognition, faulty drivers in Windows XP and many others. Wild times huahuaah.

@@marcoskatsuragi Yeah, my friend had one back in the day saying it's much better than the GeForce 4. But he mentioned texture bugs when playing Need For Speed Hot Pursuit 2. He had to wait for a driver update to fix it. I think some AIB manufacturers also has to release bios updates as well.

If you still have the 750 Ti, keep it. It will run on any OS (XP and up) in any system configuration as long as it has a PCI Express slot of some kind. It can also natively drive CRTs.

I had S3's first 3D card (the VIRGE). I don't recommend anything S3 makes :)

@@MrGencyExit64 That's actually what the Stealth III S540 was replacing, a ViRGE GX of some variety. Given my poor experience with the ViRGE, I don't know why the hell I wanted another S3 card. 🤣🤣

I remember getting my first Voodoo card. Orchid Righteous 3D. That and Tomb Raider and I was completely blown away!

Who cares about resolution? The quality in graphic fidelity compared to software render blew me away.

@@BigFatCone Oh you are certainly correct there. GL Quake made the software renderer look like a half baked prototype. The early days of 3D acceleration were a lot fun. Stuff evolved so quickly.

@@mirage809 It's been crazy being along for the ride since the 80's. I don't think I got my Voodoo 1 until Quake 2 was a thing but the point still stands.

Yeah a friend of mine got his hands on one of those prototypes as well for a mac, I remember it had 2 cpus

For those wondering, we ran small NN models for a bunch of internal tasks + some client apps. Most models were rather small and the biggest model we worked with was 1GB in size and using 8gb while running (but utilising almost 96GB when training on the in-house rig). A few days before I got fired we launched a new upgraded model that used 12-13Gb while running so for a short while, the purchase was worth it

Oh wow, that's such bad luck. At least you picked up a new job pretty quick. Hopefully you were able to make up some money by selling the 3080. Either way, sorry to hear about the shitty situation

@@CreativityNull Oh yeah I was so happy I got to sell the 3080 for the price I did. I don't remember the exact numbers but I sold it for roughly 50$ more than the price I've seen it go around that time. Homey who came over to check it and buy it even gave me a little more extra cash. But my 3080 had a backplate thermal pad mod which was worth way more than the 50$ and had an above average silicon and memory dies - when oc'd it was within 1% of the avg stock 3090 in synthetic benchmarks, but I didn't even run it OC'd because I preferred how quiet it was. Also literally no one whom I talked to at the time, except the final buyer, believed that I never mined on it lol. People think it's a given unfortunately, even though I wrote in the ad text that I don't sell to miners, for obvious reasons

@Transistor Jump Europe, and technically it was 2360$ back then

So sorry to hear that man. Glad that you're OK and I hope you'll put your GPU to good use! Maybe some VR someday?

Lol I am still using single gtx 970 to play all games. Not a single game I played was a problem for me. And seeing this new gpu prices, I think I’ll never be able to buy “next gen gpu” at their launch dates anymore. But still it was fun journey.

I had my bout with Crossfire and even multi-brand mGPU. There were probably only ever actually 3 games that ever did exist with flawless Crossfire/SLI scaling and they were running the Mantle API which I don't think Nvidia supported. The absolute king of the crop was BF4 running mantle where 2 FuryXs scaled better than 100% and the main thing is that it was a perfectly buttery smooth experience with absolutely ZERO stuttering even on an old 3rd gen i7, - If games could've continued to scale like that then I'd be wanting liquid cooled (Vega 64LC anyone?) GPUs in crossfire in my rig to this day. It was actually pretty amazing for gaming/productivity if all you played was BF4 (that was me) and then when it was time to rendering something in Blender you had a load of powerful GPUs at your disposal (Back then AMD actually rendered faster than nvidia in Blender, a trend that's no longer the case). To this day BF4 has offered the smoothest stutter free experience I have ever seen and that was running 2 liquid cooled FuryXs that ran silently. So the promise land of crossfire did exist, it's just that shortly after Mantle was abandoned and DX12 never lived up to it's mGPU promises.

I wish multiple gpu scaling was a thing still. I know it's stupid in terms of power and price, but it looks so damn cool

@@YashPatel97335 I would still be using one of my 970s if they didn't start giving me weird graphical glitches, I suspect it was due to overheating from dried up thermal paste, but I never did verify it. Got a 3070 just before the scalpathon of 2020/2021/2022 took place

@@QueenSaffryn cool. The point is that 970 and 980 are still very good cards if you want to hold out on buying new cards. And you can always get one generation behind cards at very reasonable price just like you got 3070.

I remember going out and buying a 980 (upgrading from a 660 Ti) because I had just bought Arkham Knight and wanted to play it at the best settings with the "PhsyX" stuff enabled. I don't regret it for a minute. Arkham Knight looked amazing with long overdue use of the Batmobile.

The Titan X (which is just a 980 Ti with 6 more GB of VRAM) is probably the most "definitive" legacy Nvidia card available in my arrogant opinion. Needs external power, but beyond that, can work in any OS from Windows XP up and can also drive any monitor including CRTs. And even today, the card is surprisingly powerful.

And that PCI variant is now the most expensive one because of the system compatibility it has. People who want those cards to relive that era of GLide gaming don't want to get the AGP version that's stuck using a handful of AGP 3.3v motherboards, when the PCI version is ~90% of the performance in D3D/OGL (and has no penalty in GLide) and works on any board with standard PCI. Those cards also have the quirk of using PCI signaling internally and using AGP's PCI fallback mode at 66MHz to get it working as AGP. The PCI Voodoos have no translator bridge, but they do have the usual PCI bus transaction chips.

Write them an email, and say you will trade it for a shirt with the autograph of linus