Is the 3090 really the titan card?

[amback]

Maybe am over-thinking it or perhaps even confused but would like to explain what i mean.

when we head over to the nvidia page this is what it reads.

 

Quote

"3090 is a big ferocious GPU (BFGPU) with TITAN class performance "

titan class performace? that's a strange phrase to put there, specially when this gpu IS supposed to be THE TITAN. what's even more weird is that during the entire page of the 3090 the word titan only appears once, and only on the sentence "TITAN class performance".

if i have a ford, mod it and can get lamborghini class performance, i still got a ford with lamborghini class performance not a lamborghini.

 

let's go back to the presentation and see what jensen said about the 3090 been a titan  https://youtu.be/QKx-eMAVK70?t=2052

 

Quote

 

"Several years ago, we started building the Titan, pushing the GPU to the absolute limit.

to create the best graphics card of that generation. "

"It was built in limited quantities and only through NVIDIA. The distribution was limited. The demand surprised us. "

"Creatives were making 4K movies, rendering cinematics. "

 

"There is clearly a need for a giant GPU that is available all over the world."

"So we made a giant Ampere. Ladies and gentlemen, the RTX 3090."

 

 

The word titan in that whole presentation appear twice:

at minute 34.15 "Several years ago, we started building the Titan, pushing the GPU to the absolute limit, "

at minute 35.27 "10 times quieter and keeps the GPU 30 degrees cooler than a Titan RTX design."

 

notice at the end when he said

Quote

"There is clearly a need for a giant GPU that is available all over the world."

"So we made a giant Ampere. Ladies and gentlemen, the RTX 3090.

 

they made a giant ampere GPU and according to the nvidia website, that ampere giant gpu

Quote

"3090 is a big ferocious GPU (BFGPU) with TITAN class performance "

 

so is this a giant titan like gpu OR is this really the next generation titan, because the message seem very foggy coming from nvidia.

[Moonzy]

it's the 3090, nothing else, take it as it is

[Jurrunio]

If I recall correctly, GA102 GPU has 10752 CUDA cores and you can bet on Titan getting the full core count when 3090 already has 10496. 3090 should also have fully populated the 384bit memory bus, which means the titan is unlikely to improve here any further.

 

So yeah, 3090 is as close to a Titan as Geforce cards have ever been. There is indeed the GA100 core used in the top range Quadro and Tesla cards, but Titan and Geforce never received the biggest and baddest GPU since Maxwell (900 series) days

[LIGISTX]

We won't know if nvidia has plans for a higher tier card until they either launch it, or 4000 comes out. Either way, its a hell of a lot of GPU for the money.

 

Also.... a Ford GT will beat most lambo's on track, just sayin

[igormp]

Well, the GA102 on the 3090 is already pretty close to the fully unlocked chip on the family, as @Jurrunio said.

 

The only possibility for a Titan card would be a GA100-based chip, but this one would have fewer CUDA cores (8,192 tops, while the current A100 only has 6,912), would consume more power, and have no RTX since this chip is meant as an AI workhorse.

 

8 minutes ago, Jurrunio said:

So yeah, 3090 is as close to a Titan as Geforce cards has ever got. There is indeed the GA100 core used in the top range Quadro and Tesla cards, but Titan and Geforce never received the biggest and baddest GPU since Maxwell (900 series) days

We had the Titan V based on the predecessor to the A100, the V100 (which is still the most powerful available card for Cloud services currently since the A100 is still in closed alpha). The Titan RTX was better for games tho.

[Chris Pratt]

Simply, Nvidia's names are shight. Always have been. Not sure who decides what each thing is going to be called, but they probably should have been put out to pasture long ago. This time around, for some reason, Nvidia decided that instead of having an 80 and an 80 Ti, they were going to have an 80 and a 90. The 90 is the Ti. I'd say that they were perhaps planning to phase out the Ti moniker, but I've heard rumblings of something like a 3070 Ti, which frankly makes no sense. (Again, their naming sucks). We'll have to see how the lineup fills out over the next two years, but for the time being the 3090 is the 3080 Ti.

[igormp]

18 minutes ago, Chris Pratt said:

Simply, Nvidia's names are shight. Always have been. Not sure who decides what each thing is going to be called, but they probably should have been put out to pasture long ago. This time around, for some reason, Nvidia decided that instead of having an 80 and an 80 Ti, they were going to have an 80 and a 90. The 90 is the Ti. I'd say that they were perhaps planning to phase out the Ti moniker, but I've heard rumblings of something like a 3070 Ti, which frankly makes no sense. (Again, their naming sucks). We'll have to see how the lineup fills out over the next two years, but for the time being the 3090 is the 3080 Ti.

It's likely that we'll have a 3080Ti soon to fill the gap between the 3080 and 3090. Before Turing, the Ti cards only came some months after their regular counterparts.

[amback]

Another thing i would like to point out is the amount of VRAM in the 3090, which is 24GB, on itself is pretty outstanding. the problem is 2 years ago the rtx titan already had 24GB of VRAM, so if this is the next generation TITAN you would assume it would also get a VRAM upgrade aswell .

[Jurrunio]

26 minutes ago, amback said:

Another thing i would like to point out is the amount of VRAM in the 3090, which is 24GB, on itself is pretty outstanding. the problem is 2 years ago the rtx titan already had 24GB of VRAM, so if this is the next generation TITAN you would assume it would also get a VRAM upgrade aswell .

Nope, not happening. Since GA102 is capped to a 384bit memory bus, it could hold 12 32bit GDDR6X or 24 16bit GDDR6X (clamshell mode) at most. Currently the only GDDR6X size available is 8Gigabit, which means with 24 of them 24Gigabytes capacity.

 

Not impossible to get 48GB after Samsung's 16Gb GDDR6X enters production, but for the near future 24GB is the best they could offer with GA102, even if Nvidia themselves wants more.

[igormp]

18 minutes ago, amback said:

Another thing i would like to point out is the amount of VRAM in the 3090, which is 24GB, on itself is pretty outstanding. the problem is 2 years ago the rtx titan already had 24GB of VRAM, so if this is the next generation TITAN you would assume it would also get a VRAM upgrade aswell .

There are only 1GB GDDR6X chips avaiable as of today, and nvidia is using 24 of those for the 3090, 12 on the front and another 12 on the back. 

 

When the 2GB modules become available, I doubt that NVIDIA will use those models on consumer cards and will only make their top end Tesla and Quadros available with those. Giving more vram to a consumer card means less Quadro/Tesla sales (which retail for much higher prices).

[Chris Pratt]

2 hours ago, igormp said:

It's likely that we'll have a 3080Ti soon to fill the gap between the 3080 and 3090. Before Turing, the Ti cards only came some months after their regular counterparts.

Why would there need to be something to "fill the gap"? Even the 3070 is more than most people need. The 3080 is already the overkill option. The 3090 is for specific workloads that need large amounts of VRAM. There's nothing really in between. For pure raw power without massive amounts of VRAM, you have the 3080. I'd you need the VRAM, you have the 3090.

[amback]

The more i see this the more i feel that we just got a new segment hierchy introduced to us.
maybe wont happen during this generation gpus, but i would now not be surprised if the next gen gpus - 4000 series, 5000 series -
you see the top end hierchy go xx70 xx80 XX90 XX90ti -TITAN-.

 

Mark this post, and let's see what happens.

[Moonzy]

23 minutes ago, amback said:

you see the top end hierchy go xx70 xx80 XX90 XX90ti -TITAN-.

if AMD was competitive on the high end i could see this happen

 

at this stage? nah

[jasonc_01]

Jen hsun said in the launch it was a titan level card.

[amback]

6 minutes ago, jasonc_01 said:

Jen hsun said in the launch it was a titan level card.

That means it can perform at the level of the titan. not necessarily that is the titan.

all nvidia has to say is "3090 is the titan".

[jasonc_01]

9 minutes ago, amback said:

That means it can perform at the level of the titan. not necessarily that is the titan.

all nvidia has to say is "3090 is the titan".

It replaces the titan. What else can the do to make it any more titan?

[johnsoner13]

46 minutes ago, amback said:

That means it can perform at the level of the titan. not necessarily that is the titan.

all nvidia has to say is "3090 is the titan".

Sorry but really? Can it be any more obvious? Lmfao. They don’t have to say “3090 is the titan” because it’s not, it’s the 3090 and it replaces the titan. They say “Titan class” performance because if there were to be a titan this generation, it would be the 3090, but they’re obviously trying to push away from the titan naming scheme for now. Your post/this entire thread is just useless speculation based off some nitpicking over word choice lmfao. Sorry, but you’re not the chosen one that discovered/predicted NVIDIA’s next big secret!

[igormp]

6 hours ago, Chris Pratt said:

Even the 3070 is more than most people need. The 3080 is already the overkill option.

If that's your case, fine. But don't generalize, nor think that those cards are made only for games (and even so a 3080 might not be enough).

Quote

The 3090 is for specific workloads that need large amounts of VRAM. There's nothing really in between. For pure raw power without massive amounts of VRAM, you have the 3080. I'd you need the VRAM, you have the 3090.

The vram and price gap between the 3080 and 3090 is way too big. There are tons of games that'll suffer on 4k with only 10gb of ram, specially the ones that allow mods.

And say what you want, but the amount of vram is a downgrade from the 11gb of the 2080ti, even if it's faster.

 

As I said before, nvidia usually releases the Ti cards later on, so it's likely that we'll see a refresh next year (it may be called Super instead of Ti).

[Tech-Fix]

8 hours ago, igormp said:

There are only 1GB GDDR6X chips avaiable as of today, and nvidia is using 24 of those for the 3090, 12 on the front and another 12 on the back. 

 

When the 2GB modules become available, I doubt that NVIDIA will use those models on consumer cards and will only make their top end Tesla and Quadros available with those. Giving more vram to a consumer card means less Quadro/Tesla sales (which retail for much higher prices).

Knowing this I should stop dreaming about a dual slot 3090, memory in the back needs a cooling, so it also needs space. And a 3080 with something like 16GB of VRAM should be far away, like at least 1 year from now... I really wanted more VRAM than 2080ti for 3D rendering work. Quadros are slower and way more expensive.

[igormp]

32 minutes ago, Tech-Fix said:

Knowing this I should stop dreaming about a dual slot 3090, memory in the back needs a cooling, so it also needs space. And a 3080 with something like 16GB of VRAM should be far away, like at least 1 year from now... I really wanted more VRAM than 2080ti for 3D rendering work. Quadros are slower and way more expensive.

Some motherboards have a 3rd slot that's also x16 capable, so you could have both with some space between one another, giving enough room to breathe.

 

Agreed on the expected date for the refreshes and Quadros.