Home
Creators
Search Recent Missing Random
Posts
Search DMs Popular Hash Lookup Tags Random
Importer
Import FAQ
Register Favorites Keys Review DMs Logout
Community
Telegram ThePornDude
Home Creators Posts Import Register Favorites Logout

From Fermi to Ampere – Ranking Nvidia’s GPU Generations (Patreon)

Published:
2022-07-27 22:53:04
Imported:
2026-06
Tags:

Content

This week me and Dan will be discussing Nvidia's mainstream gaming architectures that have dominated during the past 10 years - Fermi, Kepler, Maxwell, Pascal, Turing, and Ampere.

Which generation impressed you the most for its time?  Which one had the biggest "win" over AMD?  Which one sucked the most?!

Feel free to rank them below, talk about just one, and/or give your thoughts on Nvidia's Graphics Odyssey in general.  This will be a fun open discussion that should adequately prepare us mentally for how to view Lovelace....and yeah we plan to do one for past RADEON Architectures as well :).

You have 24 hours (Till 7/28 Afternoon US Central Time) to submit your comments, questions, and thoughts below! You know the rules!

https://en.wikipedia.org/wiki/Fermi_(microarchitecture)

https://en.wikipedia.org/wiki/Kepler_(microarchitecture)

https://en.wikipedia.org/wiki/Maxwell_(microarchitecture)

https://en.wikipedia.org/wiki/Pascal_(microarchitecture)

https://en.wikipedia.org/wiki/Turing_(microarchitecture)

https://en.wikipedia.org/wiki/Ampere_(microarchitecture)

Comments

Benjamin Goodman

I am still using a 1080ti and if the next generation doesn't blow my socks off I'll still be using a 1080ti for a while longer.

Diplomaticus

Pascal, and specifically the 1080 Ti, really came at a time when AMD couldn't compete; the R9 Nano was innovative enough but Fiji couldn't get the performance close. Having said that, Pascal was impressive in it's own right, nearly doubling performance at the same TDP - I thought Ampere was good but I forgot NVIDIA are boiling us alive like frogs 😰

nathan moes

I feel that at least in the mobile/laptop space one of the worst generations was ampere. Many 2080 max-p laptops perform nearly the same as 3080 laptops, the only real increase in performance was with the 3080ti laptop gpu. Currently running that in the MSI GE76 raider and it is hardly much of a real upgrade.

Mozria

I spent entirely too much money on a 1080 non-Ti when they came out in 2016. I got it specifically for 1440p 144Hz in games that were a few years old at the time, and it did that really well until I upgraded to a 48" LG OLED. Now, even with an adapter to convert DP 1.4 to HDMI 2.1 it's really showing its age. Getting to the point: Do you think that gamers (meaning me lol) could find compelling 4K120-capable options in the mid-range next gen, or should I just grab a used 6800XT for cheap? Maybe I should keep waiting and give my 1080 another year. Pascal endures.

Cleansweep

I'd say Maxwell was probably the biggest win over AMD, just because of the GTX 970. Yeah, AMD came out with better cards that didn't have VRAM limitations, but that $330 MSRP for a 70 class card was enough to get Nvidia the PR win. Plus they hadn't ruined the GeForce Experience ecosystem at that point, so you got a ton of extra software features that made it an even more appealing package without needing a dumb sign-in. I do think that Nvidia shot itself in the foot a bit with Maxwell though, since it was good enough at non-AAA games that a lot of people held off from jumping to Turning like Nvidia wanted. It wasn't until the Ampere/RDNA 2 launch that the performance uplift from an upgrade became worth it for a lot of people.

Ryzza5

Looking back, it seems we used to get multiple series per architecture. Fermi was 400+500, Kepler was 600+700. But then we have Maxwell=900, Pascal = 1000, Turing=2000, and so on. Was there ever an explanation for this? Besides nVidia seemingly running out of model numbers and reusing old ones? If anything, slowing down the numbers could contribute to customers hanging onto their older cards for longer. 280 to 680 on the face of it seems like a bigger jump than 1080 to 3080.

Treble Sketch

How would you compare price per performance or even performance per watt with GPUs going as far back as Fermi? Looking back, trying to compare 70 tier GPUs from Fermi (215W) to Ampere (250W). Is it worth the higher power consumption for the amount of performance you do get?

CompressedAIBlocks

I think Fermi was the last time Nvidia truly tried (before Ampere). Granted the GTX 480 was cut down but these were purely for yield reasons. I felt after that Nvidia had a decent performance or efficiency advantage, oftentimes both so we never truly got their best foot forward after Fermi and gamers ate that all up which is where we're on a 83/17 split. It was truly a bad snowball effect for AMD post Fermi even despite some wins they had along the way. Despite this I often stuck with AMD because the value was insane. What do you guys think?

Mozria

My other post was a bit irrelevant to the discussion at hand, so here's another. I've noticed a preponderance of gamers in my social circle that bought into Pascal with the 1070 and 1070Ti, and most of them still aren't even interested in upgrading. Given how those cards perform, their 8GB VRAM allotments and the low prices they were often found for after the 2017-2018 crypto spike, do you think they represented a sweet spot in the market that's still relevant today?

kjm015

From my perspective, Pascal was the best Nvidia generation. For me, going from a GTX 750 Ti to a GTX 1070 made a world of difference back in 2017. The post-mining crash price drops in 2018 also made it really affordable for most people. You can see it still dominating the Steam Hardware charts even 6 years after launch.

SuperCrepen

What do you guys think about Nvidia's coolers over the years? I have always wandered why Nvidia makes these fancy cooler designs with nuts and bolts and windows, while AMD seems to just make a cheap plastic shroud and call it a day. The difference between Ampere and RDNA2 in this regard is massive, with Ampere having a sexy, sleek and innovative cooler while AMD's frankly look like shit. I have a GTX 980 and have always admired the sleek blower design with a window and other nice details, but every time I see an AMD reference cooler, the look like a black rectangle with some red on it. What do you guys think is the reason for this?

PostRetroism

@SuperCrepen the Ampere FE coolers looked uh cool, but are a PITA to take apart to replace a dead fan, repaste, or replace thermal pads. So Nvidia's coolers for me are a reason to buy an AIB card. AMD coolers - well if the keep the card cool and are easy to maintain, then they're fine by me.

10GHz Netburst

Maybe it's just because my GTX 670 was the first GPU I bought with my own money, but as badly as Kepler has aged, the fact that they moved basically the whole scheduler to software in one generation impresses me a lot.

Mang0 Bandit0

970 kid here, first built pc had strix model. Then pascal blew everything else away, i had to have 1070, since the upgrades seem like smaller jumps and as sff dood, these triple slot coolers blow....Im rocking 3080 fe and as well machine as it is and kool it stays, flow through is a btich to build around.

SavageSMC

I think maxwell was a big step for NVidia due to efficiency. Pascal was a bigger delta in performance, but they preserved a lot of the efficiency changes they developed for the maxwell architecture.