NVIDIA Announces GeForce Ampere RTX 3000 Series Graphics Cards: Over 10000 CUDA Cores

[SN2716057]

And now we wait for reviews and...<drumroll>.. availability.

 

[MxPhenom 216]

most definitely 3060 with 4608 Cuda if everything doubled and moved one tier above is making the unreleased consoles look like RTX 3050.
2080 x2 = 3070, 2080Ti x2=3080. except the memory bandwith, that remains the same and probably bottlenecking.

Memory bandwidth remains the same? Not at all.

 

[Makaveli]

For all the malcontents who didn't even bother to read the graphs and watch available videos, here is a preview of 3080 with same performance numbers:

lol this video has been posted 3 times now and the last two post are on the same page!!!!!!!!!

Try reading thread before posting!

 

[Ibotibo01]

I will probably sell my RTX 2060. Please AMD release RDNA2 as soon as. It is odd RTX 3070 has 5888 cores and it gets RTX 2080 Ti performance. GTX 1060 has 1280 cores but it's performance same with GTX 980 which has 2048 cores. Ampere's 20TFLOPS GPU is equal Turing's 13TFLOPS GPU??

 

[PowerPC]

I will probably sell my RTX 2060. Please AMD release RDNA2 as soon as. It is odd RTX 3070 has 5888 cores and it gets RTX 2080 Ti performance. GTX 1060 has 1280 cores but it's performance same with GTX 980 which is 2048 cores. Ampere's 20TFLOPS GPU is equal Turing's 13TFLOPS GPU??

That's why you should never measure GPU performance in TFLOPS. It's all about more specialized hardware with GPUs.

 

[chodaboy19]

Quick question:

The fan on the top side, is supposed to suck air from the bottom and exhaust it to the top. But, when I look at the shape of the blades, it looks like the air would flow in the opposite direction? I'm a bit confused...

 

[Vya Domus]

You might be right, but considering the explicit implementation of DirectStorage, actual VRAM usage for games using this is likely to decrease, not increase. Heck, current VRAM needs for games even at 4k are massively bloated due to duplication and games streaming assets based on HDD response times and transfer rates, with huge portions of in-VRAM assets never being used at all. DirectStorage aims to solve this (didn't MS claim a 2.5x improvement or some such for the XSX?), so I'm really not worried at all. As long as the game is stored on a fast NVMe SSD, it'll likely perform admirably.

DirectStorage aims to reduce IO overhead not necessarily memory requirements, 1 GB of textures are still going to be 1 GB of textures, they'll just load more efficiently. Just because an engine no longer needs to load as many things ahead of time doesn't mean the memory wont fill up with something else, in facts that's the goal, to allow for an increases in the amount of assets used.

 

[efikkan]

Never have we witnessed more rumors about an upcoming GPU architecture, and never have they been so wrong.
I believe it was just a couple of days ago some video claimed to know everything about Ampere… (most of it was wrong)
The approximate CUDA core count have been known to those with real access since the beginning, yet most leakers missed this up until yesterday, so they are obviously serving false information. This is the moment to go back and look at the various "leaks" and evaluate if it's based on something real or just fake news or click-bait.

Let this be a lesson about using more common sense and skepticism.

 

[medi01]

2080 x2 = 3070

Huang said 3070 = 2080ti, 3080 = 2x2080 (DF said, 65-90% faster, not 100%)

 

[Vya Domus]

That's why you should never measure GPUs performance in TFLOPS.

You can always use TFLOPS for that, pick any two random GPUs and compare their TFLOPS ratings then the actual performance. I'll be you anything that probably 80% of the time the GPU with more TFLOPS will be faster in the real world as well. This time around Nvidia is doing something finicky with the way they count CUDA "cores", I put that in quote marks because they were never real cores (same with AMD's stream processors), it's the SM/CU that's the real "core" of the GPU. But for some reason this time around they chose to be even more inconsistent as to what that means. Probably to make it look more impressive.

It's all about more specialized hardware with GPUs.

Nvidia would sure like you to believe that. Shading languages don't run on specialized hardware, they can't, they need generic all-purpose processors.

 

[TheoneandonlyMrK]

Never have we witnessed more rumors about an upcoming GPU architecture, and never have they been so wrong.
I believe it was just a couple of days ago some video claimed to know everything about Ampere… (most of it was wrong)
The approximate CUDA core count have been known to those with real access since the beginning, yet most leakers missed this up until yesterday, so they are obviously serving false information. This is the moment to go back and look at the various "leaks" and evaluate if it's based on something real or just fake news or click-bait.

Let this be a lesson about using more common sense.

Samsung 8nm wasn't fake, lot's of nonesense with snippet of truth, some for memory speed and allocation right, just these new split 32bit shader core's are new and un-rumoured, they look familiar in theory to someone else's design.

 

[MxPhenom 216]

Quick question:

The fan on the top side, is supposed to suck air from the bottom and exhaust it to the top. But, when I look at the shape of the blades, it looks like the air would flow in the opposite direction? I'm a bit confused...

View attachment 167458
View attachment 167459

They probably have the fan spinning in the opposite direction than what you're thinking.

 

[medi01]

When are reviews due, and when does RDNA2 drop?

"Leaked" as november, by the dude who got a number of things right.
He touted big navi to be much faster than A104.

 

[Deleted member 24505]

Smashed AMD in the GPU wars

 

[medi01]

I will probably sell my RTX 2060. Please AMD release RDNA2 as soon as. It is odd RTX 3070 has 5888 cores and it gets RTX 2080 Ti performance. GTX 1060 has 1280 cores but it's performance same with GTX 980 which has 2048 cores. Ampere's 20TFLOPS GPU is equal Turing's 13TFLOPS GPU??

Good point.

 

[Darksword]

If 3070's are $499 the used prices on 2080Ti's is about to be hit super hard.

 

[Fluffmeister]

"Leaked" as november, by the dude who got a number of things right.
He touted big navi to be much faster than A104.

Ah yes, the fabled "Big Navi", it does appear to be the second coming of Christ for our resident fans that only buy AMD.

Fret not, here is a sneak peak of it's potential:

Spoiler

 

[MxPhenom 216]

If 3070's are $499 the used prices on 2080Ti's is about to be hit super hard.

Best sell now

 

[Krzych]

lol this video has been posted 3 times now and the last two post are on the same page!!!!!!!!!

lol generic copy-paste complaining has been posted like 210 times now and on the same pages!!!!!!

 

[RevengeFNF]

I will probably sell my RTX 2060. Please AMD release RDNA2 as soon as. It is odd RTX 3070 has 5888 cores and it gets RTX 2080 Ti performance. GTX 1060 has 1280 cores but it's performance same with GTX 980 which has 2048 cores. Ampere's 20TFLOPS GPU is equal Turing's 13TFLOPS GPU??

First, 3070 does not have the same performance of the 2080 Ti, it is faster than it.

Second, 2080 Ti have more memory bandwidth than the 3070. That's why 3070 needs a lot more Cuda Cores. You can't just compare the Cuda Cores, you need to compare everything.

 

[Oberon]

The 3070 supposedly beating the 2080 Ti at $499 just underscores how ridiculously priced that GPU was in the first place, but also really puts the pressure on AMD. If Nvidia's numbers pan out I'm struggling to see how AMD could keep up with the 3080 even with a similarly power hungry card (320W!) and the reported (and likely best-case scenario) 50% perf/W increase of RDNA 2.

A 50% perf/watt uplift from Navi puts a 5700XT class card at 3080 performance for around 300W based on some back of the envelope math.

 

[TheoneandonlyMrK]

First, 3070 does not have the same performance of the 2080 Ti, it is faster than it.

Second, 2080 Ti have more memory bandwidth than the 3070. That's why 3070 needs a lot more Cuda Cores. You can't just compare the Cuda Cores, you need to compare everything.

So why then are they saying that the 3070 is for 1440p it's interesting, reviews will tell all.

 

[Vya Domus]

Second, 2080 Ti have more memory bandwidth than the 3070. That's why 3070 needs a lot more Cuda Cores.

That's not how it works, even in the slightest. You don't trade one for the other.

 

[yotano211]

On reddit, people are selling some 2080ti's for 450-500, last week it was around 850-900

 

[Fluffmeister]

On reddit, people are selling some 2080ti's for 450-500, last week it was around 850-900

Nvidia are literally competing with themselves at this point.