What a lot of today's reviewers forget is that 1080p wasn't always the most popular resolution. When this card was released, 1024 x 768 was the most popular resolution. Many lcd's had that resolution. Expensive and larger lcd monitors (22" used to be the large lcd monitors, remember that?) were often only 1680x1050 or 1600x900.
@@JohnSmith-iu8cj Resolution mostly varied by lcd panel size. In 2008, according to the Steam survey, over 70% of pc gamers ran at either 1024 x 768 ( I ran that resolution on a 17" lcd) or 1280 x 960 (I never ran that resolution.)
True! These tests were meant to show what performance is like at a modern resolution, but 1080p did not become mainstream until around 2010 or so. However, higher resolution were already available on larger CRTs.
In a lot of ways, the 8800GTX was the 3090Ti of its time. Like the 8800GTX, the 3090Ti considerably raised the power envelope - from the generally accepted 250W to a staggering 450W - and was about 1.8X faster than the previous gen. 2080Ti as a result. And now, the 220W 4070 Super and 245W 7700XT are considered "power efficient" cards. I'm pretty sure things will eventually go up to 500-600W, now that we have finally hit the silicon barrier and it's getting harder and harder to shrink transistors and/or run them more efficiently.
Also,don't forget, at $600+ it was like x2-3 times more expensive than the GPUs that came before. Compared to $250 of the mainstream 8800 GT. Kinda like RTX 3090/4090 vs 3070/4070. Also keep in mind, people would be buying 2, 3 and sometimes 4 of these cards to use with SLi. So top tier GPU setups were around $2000, again similar to 3090s and 4090s of today. Just to highlight that nothing really changed in the past 18 years, and top tier PCs of today aren't really that more expensive than the top tier PCs of 2006.
@@harryshuman9637 Pretty much. That's exactly why I consider the 4070Ti to be Nvidia's flagship. Cards beyond the 4070Ti are Titan-class behemoths with ridiculous TDPs and terrible price-to-performance ratios. You're paying a lot for not a whole lot of performance. For example, the 4090 offers ~60% better performance than the 4070Ti, yet cost twice as much, as far as MSRP is concerned.
A big "Issue" is also how well those cards scale with additional power. You dont lose a lot of performance with limiting power yet you gain a good bit with just increasing the power budget. Unlike older gens where you would put in tons of power and get only slight gains.
@@Sam-K You are comparing gaming performance tho, but in 2024 you aren't buying top tier GPU for gaming. You are buying it for ML applications, and 4090 is around twice as fast as a 4070.
A VERY excellent video, and exactly the point of view I would have taken. It was clear some years ago, even before the AI craze, that CUDA was the distant leader of the supercomputing-on-a-GPU revolution, and that was all because as you said, they laid the inroads early on and basically formed the industry from the ground up around their products. It was hard for AMD or other competitors to break in, because it wasn't just the GPGPU industry they pursued, it was _NVIDIA's_ GPGPU industry. The 8800GTX and G80 GPU at the time was described by multiple publications as an "inflection point", but they couldn't have known back then how right they'd be.
The foundation for CUDA was a set of programming and debugging tools derived from the Open64 project in the early 2000s -- an open source compiler for Itanium and x86-64.
The same silicone doesn't equal same "specs", learned that the hard way. Power delivery, number of phases, cooling solutions for components outside of core.
@@JamesSmith-sw3nk I have a Quadro FX 5600, the pro 8800 GTX with 1.5 GB. It achieves exactly the same clocks as a stock 8800 Ultra and not 1 MHz more.
5% from datacenters in 2015 to 87% that is a massive change. They barely even need gamers anymore. I remember when the 8800GTX came out but didn’t pay much attention to CUDA as the gaming performance was so impressive.
More important than CUDA, the 8800 GTX was also Nvidia's first GPU to support OpenCL, the open GPGPU language that is equally fast/efficient as CUDA, yet works on all hardware from Nvidia, AMD, Intel, Apple, ARM, ... OpenCL is increasingly important especially in science, where new supercomputers regularly have GPUs from different vendors - the same OpenCL code works everywhere and developers don't have to waste years on code porting. 13:29 AMD's counterpart to CUDA is not ROCm, but HIP, which is their proprietary nonsense language that doesn't even work on most of their own GPUs.
Both CUDA and HIP are dialects of C++, so I don't know why you're singling out HIP specifically, especially when HIP is able to run on more than just AMD GPUs.
I own a 8800 GT, which has basically a shrunk-down, slightly reconfigured G80, the G92, where the memory interface was cut down but counteracted with twice the TMUs. Was probably too good of a card for where NVIDIA placed it
i used to do distributed computing (seti@home) back in the day, and the first time i saw my 8800 working with cuda instead of the CPU, it was so fast i thought there was a problem with it because there was no way it could be that much faster than a CPU
Been using Nvidia since the beginning and enjoyed the ride. My first G80 was the 8800GTS and then the XFX Ultra XXX edition in july of 2007 to play Crysis. But my favorite G series was the final iteration on a 55nm G92 Die or my EVGA 250 GTS 1 gig. That GPU had refined higher clocks speeds and process. Thanks Fully buffered for the look back and informative video.
i remember the gtx 8800 i wanted one bad lol. but i could only afford the 7600gs:(. the card was still 225 bucks and i was earning 90 bucks a week as an apprentice mechanic 1st year lol
It definitely was the turning point no doubt about it. Thing is it really took over a decade for it to take off, and 15 years for CUDA to become the staple it is today.
I loved the amd opteron quad cpu setup, but that case is really expensive. This old 8800 GTX is some huge up in nvidia's game, cuda helped me in more ways than just fps improvements in older games.
Great video. I owned three 8800gtx's, and for about a week I remember I had 14th on 3D mark's scoreboard. Tri-sli stuttered a lot in games so I eventually just ran 2 way sli and used the 3rd card for PhyXs.
never owned 8800gtx but i did own BFG 8800gts 320 and 2X EVGA 8800GTin sli, they where way to hot.. Those where super exciting times for gaming with great games..
CUDA is the market default. It's like Adobe. Even if a program better than Photoshop and Premiere exists, they will ask in your curriculum Adobe products.
One day ill get around to setting up my dream beast, 2 G92 8800GT Gold cards in SLi with an Athlon II X2 example running over 4Ghz, a golden theme for a golden era, back when i was using an FX5200 and ol' Sempron where the only thing it was good at was forum and review binging the hardware i could never afford :')
@@FullyBuffered Got a 7950GT SLi system in mind too with an e6500K, still hunting for a second sparkle 7950GT though, need to sell on some pcs first before making any more :P
в 2007 году я смотрел на 8800, но предпочёл взять ati 19xx. К сожалению, года через 4-5 она сломалась, но до того момента я не особо страдал от отсутствия унифицированных шейдеров
I have sort of wondered if we'll move on from the "GPU" naming convention given it's just a general purpose vector processor that we've been running for a long time now.
definately the 2 Data-Center/A.I. landmarks for nVIDIA are *Tesla* (8xxx-lineup ) and *Fermi* (GTX5xx-lineup) architectures !! Both of them heavilly compute-based architectures that paved the way for the A.I.-era coincidentally ,these 2 have always been my most beloved architectures although i'm(was) just a videogamer ...
@@GrainGrown i'm an nVIDIA customer since 1999 ( *Riva TNT2 Ultra* ) i know very well how it's written , write in google : nVIDIA logo and check the letters ,the "n" is a small one and the other letters are capitals ,but they have the exact same size ,which ,of course i can't write it the same way using "MSword" since ,in "MSword" small & caps letters have different sizes ...😉 (*it can be written the way you say also , but this doesn't mean that the way i write it isn't correct as well )
I never really understood what that long plastic that extends the GPU is, what is it actually? Some older AMD Radeon HD blower type cards had those too.
I think it was just like in the very old days when "full length" ISA cards sometimes had a little extension to lock into support brackets at the front of cases. Maybe Nvidia anticipated the sag problems of the future even if it wasn't strictly necessary with these cards?
gtx10 from 9 series brought the performance leap that you mentioned from 7950 to 8800gtx, funny that when they released the 1070ti, the only difference between it and gtx1080 was gddr5x and 128 more cuda cores, basically cutting an 8800gtx...
Its hilarious that the 8800GTX can now get spanked and DESTROYED BY AN AMD RADEON RX 780m lmmfao! 780m (8700G ) is the performance today of an Radeon R9 280X but lol in the cpu !! and is very close to the R9 290 desktop & obliterates the Flagship (ATT) HD 6990 6GB dual GPU CFX card $1500 GPU in 2010
Which I pulled a 9800+ from someones old build years ago and it still works. I keep it in my collection of older GPUs I use for my Win Xp 32Bit play everything I can get working parts.
your GPU test is being bottlenecked by the CPU, your imput lag is also horrible whats clearly seen at 9:02 when moving the mouse, in my opinion it is unplayable like this, yes there are some casual players who dont know the feel of actual 5ms frametime 144Hz and up gameplay, who are fine with it and even playing on a controler, but i could not game like this.
it'll get to the point where it'll be a being whos bloods electric and its now lobbying for rights that superceed human rights to become an electric god....at that point our blood will become a mere color to be used in a twisted pallet when it subverts us in a matrix like future un the guise of ART.... sounds crazy i know..
We all helped create the worlds most valuable company, even though nvidia has basically told gamers to get rich or touch grass, I'm happy I threw my hat into the ring 24 years ago. 7h3 m0r3 y0u buy 7h3 m0r3 y0u 54v3
What a lot of today's reviewers forget is that 1080p wasn't always the most popular resolution. When this card was released, 1024 x 768 was the most popular resolution.
Many lcd's had that resolution. Expensive and larger lcd monitors (22" used to be the large lcd monitors, remember that?) were often only 1680x1050 or 1600x900.
Yes!!! I had a 1440*900 panel :D
1680x1050 was the standard for new tfts at the time. Old tfts had mostly 1280x1024
@@JohnSmith-iu8cj Resolution mostly varied by lcd panel size. In 2008, according to the Steam survey, over 70% of pc gamers ran at either 1024 x 768 ( I ran that resolution on a 17" lcd) or 1280 x 960 (I never ran that resolution.)
@@JohnSmith-iu8cj i still have my 1280x1024 dell monitor. it's been my primary monitor for 20 years now
True! These tests were meant to show what performance is like at a modern resolution, but 1080p did not become mainstream until around 2010 or so. However, higher resolution were already available on larger CRTs.
In a lot of ways, the 8800GTX was the 3090Ti of its time.
Like the 8800GTX, the 3090Ti considerably raised the power envelope - from the generally accepted 250W to a staggering 450W - and was about 1.8X faster than the previous gen. 2080Ti as a result.
And now, the 220W 4070 Super and 245W 7700XT are considered "power efficient" cards. I'm pretty sure things will eventually go up to 500-600W, now that we have finally hit the silicon barrier and it's getting harder and harder to shrink transistors and/or run them more efficiently.
Also,don't forget, at $600+ it was like x2-3 times more expensive than the GPUs that came before. Compared to $250 of the mainstream 8800 GT. Kinda like RTX 3090/4090 vs 3070/4070.
Also keep in mind, people would be buying 2, 3 and sometimes 4 of these cards to use with SLi. So top tier GPU setups were around $2000, again similar to 3090s and 4090s of today.
Just to highlight that nothing really changed in the past 18 years, and top tier PCs of today aren't really that more expensive than the top tier PCs of 2006.
@@harryshuman9637 Pretty much. That's exactly why I consider the 4070Ti to be Nvidia's flagship. Cards beyond the 4070Ti are Titan-class behemoths with ridiculous TDPs and terrible price-to-performance ratios. You're paying a lot for not a whole lot of performance.
For example, the 4090 offers ~60% better performance than the 4070Ti, yet cost twice as much, as far as MSRP is concerned.
more like 8800 Ultra, imo
A big "Issue" is also how well those cards scale with additional power. You dont lose a lot of performance with limiting power yet you gain a good bit with just increasing the power budget. Unlike older gens where you would put in tons of power and get only slight gains.
@@Sam-K You are comparing gaming performance tho, but in 2024 you aren't buying top tier GPU for gaming. You are buying it for ML applications, and 4090 is around twice as fast as a 4070.
i have the Quadro version of the 8800 GTX. Same deal but has 1.5 GB VRAM instead. Absolutely massive for the era.
A VERY excellent video, and exactly the point of view I would have taken. It was clear some years ago, even before the AI craze, that CUDA was the distant leader of the supercomputing-on-a-GPU revolution, and that was all because as you said, they laid the inroads early on and basically formed the industry from the ground up around their products. It was hard for AMD or other competitors to break in, because it wasn't just the GPGPU industry they pursued, it was _NVIDIA's_ GPGPU industry. The 8800GTX and G80 GPU at the time was described by multiple publications as an "inflection point", but they couldn't have known back then how right they'd be.
Very true - thanks for the kind words man :)
The foundation for CUDA was a set of programming and debugging tools derived from the Open64 project in the early 2000s -- an open source compiler for Itanium and x86-64.
Interesting - I did not know that!
yeah and unfair for other tech companies like AMD intel.. Anti competitive BLACK BOX TECH
CUDA is not OPEN = ANti Consumer
Your videos are so well produced and researched. I wish you made them more frequently.
Many thanks :)
If you own a 8800gtx, use MSI Afterburner and set it to the 8800 Ultra's clocks. They are the same gpu. The Ultra has a better cooler.
Ultra has a newer revision G80 and lower latency memory.
The same silicone doesn't equal same "specs", learned that the hard way. Power delivery, number of phases, cooling solutions for components outside of core.
@@Mr371312 It often can. When it comes to overclocking, it's ALWAYS:..
"Your mileage will vary."
@@JamesSmith-sw3nk I have a Quadro FX 5600, the pro 8800 GTX with 1.5 GB. It achieves exactly the same clocks as a stock 8800 Ultra and not 1 MHz more.
Binning
Extremely informative and interesting :)
Hi there Timmy
Whoa Timmy, nice to see you again bud. Miss your videos
Thanks for the kind words Timmy - good to see you around! :)
5% from datacenters in 2015 to 87% that is a massive change. They barely even need gamers anymore. I remember when the 8800GTX came out but didn’t pay much attention to CUDA as the gaming performance was so impressive.
Indeed! At the moment gaming has become a side hussle for them...
More important than CUDA, the 8800 GTX was also Nvidia's first GPU to support OpenCL, the open GPGPU language that is equally fast/efficient as CUDA, yet works on all hardware from Nvidia, AMD, Intel, Apple, ARM, ... OpenCL is increasingly important especially in science, where new supercomputers regularly have GPUs from different vendors - the same OpenCL code works everywhere and developers don't have to waste years on code porting.
13:29 AMD's counterpart to CUDA is not ROCm, but HIP, which is their proprietary nonsense language that doesn't even work on most of their own GPUs.
Both CUDA and HIP are dialects of C++, so I don't know why you're singling out HIP specifically, especially when HIP is able to run on more than just AMD GPUs.
Interesting - thanks for the comment!
Great video! is nite to see you back!
Thank you!
I own a 8800 GT, which has basically a shrunk-down, slightly reconfigured G80, the G92, where the memory interface was cut down but counteracted with twice the TMUs. Was probably too good of a card for where NVIDIA placed it
A great card indeed!
i used to do distributed computing (seti@home) back in the day, and the first time i saw my 8800 working with cuda instead of the CPU, it was so fast i thought there was a problem with it because there was no way it could be that much faster than a CPU
The 8800GTX was the most important thing to happen to computers since the IBM 5150.
Been using Nvidia since the beginning and enjoyed the ride. My first G80 was the 8800GTS and then the XFX Ultra XXX edition in july of 2007 to play Crysis. But my favorite G series was the final iteration on a 55nm G92 Die or my EVGA 250 GTS 1 gig. That GPU had refined higher clocks speeds and process. Thanks Fully buffered for the look back and informative video.
Thank you!
i remember the gtx 8800 i wanted one bad lol. but i could only afford the 7600gs:(. the card was still 225 bucks and i was earning 90 bucks a week as an apprentice mechanic 1st year lol
Your videos are like a treat
Thank you! That's great to hear :)
It definitely was the turning point no doubt about it. Thing is it really took over a decade for it to take off, and 15 years for CUDA to become the staple it is today.
Thanks for the comment!
Excellently presented as always.
Thanks! :)
Hard work took a decade to pay off, but it paid off big.
I got a pair of 8800 GT in an XPS I got recently so it show how far we have come from the pair of SLI 8800GT to where I am now with my 3080 12gb
Nice!
I loved the amd opteron quad cpu setup, but that case is really expensive.
This old 8800 GTX is some huge up in nvidia's game, cuda helped me in more ways than just fps improvements in older games.
Just found out your channel, love the calming way you speak and love the deep dive into Nvidia's GPUs
Great video. I owned three 8800gtx's, and for about a week I remember I had 14th on 3D mark's scoreboard. Tri-sli stuttered a lot in games so I eventually just ran 2 way sli and used the 3rd card for PhyXs.
Thanks!
EVGA SR-2 FTW!
never owned 8800gtx but i did own BFG 8800gts 320 and 2X EVGA 8800GTin sli, they where way to hot.. Those where super exciting times for gaming with great games..
This card excited me so much when it came out that I bought 3 of them for 3way SLI. THAT WAS A MISTAKE.
Haha that would have been an awesome setup back in the day!
@@FullyBuffered I felt like a king for two generations. 😂
CUDA is the market default. It's like Adobe. Even if a program better than Photoshop and Premiere exists, they will ask in your curriculum Adobe products.
One day ill get around to setting up my dream beast, 2 G92 8800GT Gold cards in SLi with an Athlon II X2 example running over 4Ghz, a golden theme for a golden era, back when i was using an FX5200 and ol' Sempron where the only thing it was good at was forum and review binging the hardware i could never afford :')
Do it!
@@FullyBuffered Got a 7950GT SLi system in mind too with an e6500K, still hunting for a second sparkle 7950GT though, need to sell on some pcs first before making any more :P
Thanks for a informative video!
Thank you!
i have collected 3 GTS8800 and 2 8800 Ultra sadly they die for unknow reason. but my 4 GTX8800 still working great at 2024 inside my retro gaming pc.
Great video!!
Thank you!
Casual SR-2
Nvidia became a trillion dollar company not because they made good hardware, but because they made useful software... See: Microsoft
Informative. Thank you.
Thank you! Glad to hear!
в 2007 году я смотрел на 8800, но предпочёл взять ati 19xx. К сожалению, года через 4-5 она сломалась, но до того момента я не особо страдал от отсутствия унифицированных шейдеров
I have sort of wondered if we'll move on from the "GPU" naming convention given it's just a general purpose vector processor that we've been running for a long time now.
Very true!
old gpus are so funky and fun
8800 GTX was a groundbreaking product that pushed the boundaries of what was possible in GPUs.
definately the 2 Data-Center/A.I. landmarks for nVIDIA are *Tesla* (8xxx-lineup ) and *Fermi* (GTX5xx-lineup) architectures !!
Both of them heavilly compute-based architectures that paved the way for the A.I.-era
coincidentally ,these 2 have always been my most beloved architectures although i'm(was) just a videogamer ...
True!
It's written NVIDIA, not "nVIDIA".
@@GrainGrown i'm an nVIDIA customer since 1999 ( *Riva TNT2 Ultra* ) i know very well how it's written , write in google : nVIDIA logo and check the letters ,the "n" is a small one and the other letters are capitals ,but they have the exact same size ,which ,of course i can't write it the same way using "MSword" since ,in "MSword" small & caps letters have different sizes ...😉
(*it can be written the way you say also , but this doesn't mean that the way i write it isn't correct as well )
Loving the new haircut. I always thought you were gorgeous but now... woof. 😊
Pretty sure it’s “Ai, Ai, Ai” 😂😂😂 or at least that’s what so many CEOs keep saying at every shareholder meeting.
Gamers and nerds, unite!
just showing off the evga sr2 lol
Can AI explain why one man needs so many spatulas?
I never really understood what that long plastic that extends the GPU is, what is it actually? Some older AMD Radeon HD blower type cards had those too.
Put in or remove card from server easily
Stiffening bracket for mechanical integrity in transport. It is intended to be used in a case that has slots to support the weight from both sides.
I think it was just like in the very old days when "full length" ISA cards sometimes had a little extension to lock into support brackets at the front of cases. Maybe Nvidia anticipated the sag problems of the future even if it wasn't strictly necessary with these cards?
I have the first 7800gtx dual slot with that extender, from dell. It has no cuda
@@JohnSmith-iu8cjthose were older architecture. Nvidia only support CUDA with unified shaders architecture.
gtx10 from 9 series brought the performance leap that you mentioned from 7950 to 8800gtx, funny that when they released the 1070ti, the only difference between it and gtx1080 was gddr5x and 128 more cuda cores, basically cutting an 8800gtx...
WOOWW your hair is so nice!
Its always nice when you post a video, highlight of my day everytime!
Many thanks for the kind words! I'm glad to hear that! :D
Its hilarious that the 8800GTX can now get spanked and DESTROYED BY AN AMD RADEON RX 780m lmmfao! 780m (8700G ) is the performance today of an Radeon R9 280X but lol in the cpu !! and is very close to the R9 290 desktop & obliterates the Flagship (ATT) HD 6990 6GB dual GPU CFX card $1500 GPU in 2010
& only 12CU's
Thanks cuda core, really cool
2032
Aipu
Also, it's became the basis for 9800,that became later the 250!😊😊😊
Which I pulled a 9800+ from someones old build years ago and it still works. I keep it in my collection of older GPUs I use for my Win Xp 32Bit play everything I can get working parts.
@@Warbob11
Got 3 of 'em now.
8)
Indirectly, those all use G92 and not G80. But G92 is a more efficient refresh with upsides and downsides.
@@classic_jam
Maybe.
I might make a video,going from 8800 to the 250.
Just got to setup a platform.
yeah nvidia the anti-consumer - anti competitive POS GPU Company
your GPU test is being bottlenecked by the CPU, your imput lag is also horrible whats clearly seen at 9:02 when moving the mouse, in my opinion it is unplayable like this, yes there are some casual players who dont know the feel of actual 5ms frametime 144Hz and up gameplay, who are fine with it and even playing on a controler, but i could not game like this.
I can imagine! Thanks for the comment
It's "H100", not "Haayitch100".
Now nvidia is greedy company.
more like always
@@bp-it3ve earlyer they holds off price appetites . Now they make excuses and lies with "we will do low prices after"
Sounds like something a poor would say.
@@LawrenceTimme sounds like something a rich man would say. And if high prices are normal for you, then I feel sorry for you.
@@bp-it3ve They used to keep their appetites. Now they are lying by saying “we will lower prices after Covid”.
туда
Please, let's try to forget about those cringe, disgusting "gamer" GPUs. Everyone knows that nvidia is really an AI server company.
That's certainly their trajectory...
i think all this ai crap should be a crime against humanity.. i do not approve of the ai crap happening these days..
it'll get to the point where it'll be a being whos bloods electric and its now lobbying for rights that superceed human rights to become an electric god....at that point our blood will become a mere color to be used in a twisted pallet when it subverts us in a matrix like future un the guise of ART.... sounds crazy i know..
You can certainly make some strong arguments against it...
We all helped create the worlds most valuable company, even though nvidia has basically told gamers to get rich or touch grass, I'm happy I threw my hat into the ring 24 years ago.
7h3 m0r3 y0u buy 7h3 m0r3 y0u 54v3
Why were you a share holder?