Discussion NVIDIA Ampere RTX3000 series Discussion Thread(Updated Launch Specs)

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
NVIDIA Ampere is rumored to be announced in August and launch in September at Computex. It is said to be decent improvement over Turing with around 40% performance improvement through out the lineup. With the launch of RTX3080Ti(most probably that is what the RTX2080Ti successor is going to be called) we can expect 4K 60fps+ in 90%+ titles at ultra settings which was not possible with present high end cards well still 4K 100fps+ will out of reach for majority of demanding titles.

GA102(TITAN, RTX3080Ti):
5376 CUDA Cores
24GB GDDR6

GA103(RTX3080):
3840 CUDA Cores
10GB GDDR6

GA104(RTX3070,RTX3060):
3072 CUDA Cores
8GB GDDR6

GA106(RTX2660Ti,RTX2660):
1920 CUDA Cores
6GB DDR6

GA107(RTX2650):
1280 CUDA Cores
4GB GDDR6

I doubt there will be non RTX cards even on low end, I guess all the cards will be Ray-tracing enabled and come packed with RT Cores.
The info out there is very little but the potential leaks of specs or rumors are interesting. We did expect even more than this but 40% performance improvement over Turing is not too bad. I hope there is at-least comparatively bigger jump in Ray-tracing performance compared to Raw performance making Ray-tracing bit more mainstream.

Are we finally getting to the point where a single GPU can handle 4K@60Hz without any limitations.

UPDATED SPECS AFTER LAUNCH
RTX3090(TITAN)
10496 CUDA Cores
24GB GDDR6X
$1499

RTX3080
8704 CUDA Cores
10GB GDDR6X
$699

RTX3070
5888 CUDA Cores
8GB GDDR6
$499

That is pure Insanity in comparison to what we were expecting. Interesting. Lets see how the Performance goes in real world. On paper the specs are pure Insanity. If it delivers similarly in real world then we are in for a big WIN.
 
Last edited:
personally i think nvidia try to avoid hyping next gen geforce. because they still try to clear 20 series inventory. even when nvidia talking about the quadro part (which is very little from the presentation) he was highlighting existing RTX 8000 solution that they will be offering to the professional. maybe we will hear something in mid or late Q3 this year.
 
personally i think nvidia try to avoid hyping next gen geforce. because they still try to clear 20 series inventory. even when nvidia talking about the quadro part (which is very little from the presentation) he was highlighting existing RTX 8000 solution that they will be offering to the professional. maybe we will hear something in mid or late Q3 this year.
Yes that is what I am thinking. But if the entire lineup gets equal treatment that will get RTX8000 G2 and RTX TITAN G2 packing 7360-7424 CUDA Cores which is good. Really good. Lets see how its ends up being at the launch.
 
Last edited:
Ray Tracing on co-processor hmmmmmmm.........

Okay If anyone says that it will exist on add-on card going into separate PCIe slot and be bridged together using NVLink or something like that then nope that will not be the case. I can bet on it and no that is not going to happen.

Co-processor on the same board well that can be discussed. I don't think that will happen but will say that there is little possibility. Now coming to the point if they do implement co-processor dedicated to handle ray-tracing and related calculations and possibly bridge it to raster engine with 0 latency(which is very much complicated and hard to do but lets assume NVIDIA pulls it off) then That will enable the actual chip to have lot more CUDA Cores and raster performance than we expected and the performance jump will be over 100% in comparison to RTX2000 series.
Will that happen? For now possibility is very less. If it happens that will be insanity.
 
With all the current rumors NVIDIA may have managed to get the performance up by decent level. But the thing troubling me is the pricing of the high end GPUs. I think NVIDIA should bring its high end pricing down so that it doesn't make the mid range pricing look so far apart and odd. Because NVIDIA will have to bring down their mid-range pricing a notch or two if they want to stay in competition because with RX 5500 XT and RX 5600 XT. There is not a single RTX or GTX GPU below RTX 2080 which makes sense to be considered over AMD GPUs. That is a clean sweep from AMD in the Mid-range and below.

RX 5500 XT priced around GTX1650 performs at level of GTX1660
RX 5600 XT $25 more than GTX1660 Super performs at the level of RTX2060
RX 5700 priced around RTX2060Super performs better than RTX2070
RX 5700 XT priced around RTX2060Super again and performs better than RTX2070Super

Seriously NVIDIA needs to bring down pricing throughout the lineup. I hope they price upcoming GPUs in a more sensible way. Because even now if they go on pricing higher what is stopping AMD from increasing the price as well and that will be a big hit for us consumers.
 
There is not a single RTX or GTX GPU below RTX 2080 which makes sense to be considered over AMD GPUs. That is a clean sweep from AMD in the Mid-range and below.

RX 5500 XT priced around GTX1650 performs at level of GTX1660
RX 5600 XT $25 more than GTX1660 Super performs at the level of RTX2060
RX 5700 priced around RTX2060Super performs better than RTX2070
RX 5700 XT priced around RTX2060Super again and performs better than RTX2070Super
I would not say that's the case at all. The 5500 XT is generally shown to perform about on par with the 1650 SUPER, which is significantly faster than the 1650, but still typically priced lower than a 5500 XT 4GB, at least in the US market. The 1650 SUPER currently starts around $160, while the 5500 XT 4GB starts around $180...

1650 SUPER: https://pcpartpicker.com/products/video-card/#c=476&sort=price
5500 XT: https://pcpartpicker.com/products/video-card/#c=478&sort=price

In some games the 5500 XT manages to be a little faster, but in others, its x8 PCIe interface can cripple performance on systems without PCIe 4.0 support, especially when VRAM is exceeded on the 4GB model. As for the 8GB model, pricing starts at $200, but at least in current titles, performance is still barely any better than the 1650 SUPER available for $40 less. Meanwhile, the original 1660 tends to offer over 10% more performance than those cards for only about $10 more than the 8GB 5500 XT...

1660: https://pcpartpicker.com/products/video-card/#c=439&sort=price
1660 SUPER: https://pcpartpicker.com/products/video-card/#sort=price&c=450

And for $230, one can get a 1660 SUPER, a card that can be around 25% faster than the 5500 XT in demanding games. Both the 4GB and 8GB models of the 5500 XT are priced poorly in my opinion, especially considering they are only slightly faster than AMD's $200 cards from 3-4 years ago. The RX 480 and 580 were very competitive against Nvidia's offerings at the time, but the 5500 XT not so much. Around this price segment, Nvidia is arguably offering better value at this time.

As for the 5600 XT, AMD's decision to overhaul the cards days before launch with a BIOS revision that significantly increased clock rates has resulted in some models not being able to clock as high as others. So while it's possible for many models to perform on par with a 2060, that's generally not the case for the lowest-priced cards, and one will probably be spending around $280 for a card that can properly handle the higher core and memory clocks. And of course, it's now possible to find some 2060s for around $300. The average selling price of a 5600 XT tends to be a little lower, but the 2060 does offer some additional features like dedicated RT hardware and a better video encoder. Both have their merits, but AMD is arguably only about on par with Nvidia in terms of value in this price segment.

The 5700 appears to be largely out of stock now, and I wouldn't be surprised if manufacturers mostly quit making them in favor of the 5600 XT, which doesn't perform too far behind, and the 5700 XT, which doesn't perform all that far ahead. With faster cards likely launching in a matter of months, we may see 5700 XT pricing crunched down to a lower price segment, leaving less room for the non-XT variant.

As for the suggestion of the 5700 performing faster than a 2070, that's rarely the case, and on average it's only around 10% faster than a 2060, with the 2060 SUPER in turn being faster than it. The 5700 XT is faster than the 2060 SUPER and 2070, but again, on average it tends to not be as fast as a 2070 SUPER. And the 2070 can now be had for a similar price around $400.

So much like with the 5600 XT, the 5700 XT does tend to offer a little more performance per dollar than the competition, but only if you are willing to trade away some additional features for it. Overall, I would say the 5600 XT and 5700 XT are reasonable options, but in terms of value they don't really go much beyond Nvidia's offerings around their price segment. So, I certainly wouldn't say it's a "clean sweep" for AMD in the mid-range right now.
 
With all the current rumors NVIDIA may have managed to get the performance up by decent level. But the thing troubling me is the pricing of the high end GPUs. I think NVIDIA should bring its high end pricing down so that it doesn't make the mid range pricing look so far apart and odd. Because NVIDIA will have to bring down their mid-range pricing a notch or two if they want to stay in competition because with RX 5500 XT and RX 5600 XT. There is not a single RTX or GTX GPU below RTX 2080 which makes sense to be considered over AMD GPUs. That is a clean sweep from AMD in the Mid-range and below.

RX 5500 XT priced around GTX1650 performs at level of GTX1660
RX 5600 XT $25 more than GTX1660 Super performs at the level of RTX2060
RX 5700 priced around RTX2060Super performs better than RTX2070
RX 5700 XT priced around RTX2060Super again and performs better than RTX2070Super

Seriously NVIDIA needs to bring down pricing throughout the lineup. I hope they price upcoming GPUs in a more sensible way. Because even now if they go on pricing higher what is stopping AMD from increasing the price as well and that will be a big hit for us consumers.

AMD definitely interested with increasing price structure. hence why $275 RX590 in late 2018 happen when same performance tier (slightly behind) can be had for $250 back in 2016. at this point we can only wait. btw since this is nvidia true shrink since pascal it might be possible the fastest ampere can end up in the range of 80% increase over RTX2080Ti.
 
AMD definitely interested with increasing price structure. hence why $275 RX590 in late 2018 happen when same performance tier (slightly behind) can be had for $250 back in 2016. at this point we can only wait. btw since this is nvidia true shrink since pascal it might be possible the fastest ampere can end up in the range of 80% increase over RTX2080Ti.
If Ampere really gets 80% performance increase then that will be insanely powerful and finally we get true 4K no compromise performance using single GPU. But at what price. I guess it will still be priced same if not less. I don't see NVIDIA going up in price because that will be a bad move.
 
Okay Seriously.

If Top of the Line AMD Big Navi GPU is said to have 200% performance improvement over RX 5700 XT. Where does NVIDIA Ampere Flagship will land up in performance. Because RTX2080Ti is only around 50% up on performance over RX 5700 XT. That means Big Navi Flaghip will have 100% performance increase over RTX2080Ti. That looks bit unreal. Don't care what the source is. That huge performance gain is pure insanity. Where does NVIDIA Ampere sit in this crazy not so practical chart of expectation.
 
Okay Seriously.

If Top of the Line AMD Big Navi GPU is said to have 200% performance improvement over RX 5700 XT. Where does NVIDIA Ampere Flagship will land up in performance. Because RTX2080Ti is only around 50% up on performance over RX 5700 XT. That means Big Navi Flaghip will have 100% performance increase over RTX2080Ti. That looks bit unreal. Don't care what the source is. That huge performance gain is pure insanity. Where does NVIDIA Ampere sit in this crazy not so practical chart of expectation.
My bad. I had to go through multiple articles as they were saying 200% performance improvement over RX 5700 XT but only 50% performance improvement over RTX2080Ti which made no sense as it should 100% performance improvement over RTX2080Ti and then realized that they were saying that it will 200% of RX 5700 XT an not over which means 2 x performance or simply put 100% performance increase over RX 5700 XT. But then still their point of it being 50% performance improvement over RTX2080Ti doesn't sit right.

Lets say
RX5700XT = 100fps(base line)
RTX2080Ti = 150fps(50% performance over RX5700XT)
Big Navi Flagship = 200fps(100% performance over RX5700XT but only 35% improvement over RTX2080Ti)

Still am unable to understand from where they are getting that 50% performance over RTX2080Ti number.
 
well as usual things are get crazy as we nearing new GPU launch. and most often if nvidia coming out with something really good some people will start take that as a definite sign that AMD have something under their sleeve making nvidia very nervous. i still remember back then where some people very insisting that polaris will at least going to match nvidia GTX1070 with the price of $300 ($100 cheaper than 1070). they come to this conclusion after matching all the rumor and some of the limited info AMD telling the public back then like how polaris will be 100% more power efficient than their previous gen GPU.

so this time we got something like RDNA 2 is 50% more power efficient than the original RDNA. combine with increase IPC (which i believe still pure speculation at this point?) and 80CU big navi will at least end up like 60% faster than RTX2080Ti. and that is not the biggest navi that speculated that can go as high as 128 CU.

so be wary of weird expectation.
 
well as usual things are get crazy as we nearing new GPU launch. and most often if nvidia coming out with something really good some people will start take that as a definite sign that AMD have something under their sleeve making nvidia very nervous. i still remember back then where some people very insisting that polaris will at least going to match nvidia GTX1070 with the price of $300 ($100 cheaper than 1070). they come to this conclusion after matching all the rumor and some of the limited info AMD telling the public back then like how polaris will be 100% more power efficient than their previous gen GPU.

so this time we got something like RDNA 2 is 50% more power efficient than the original RDNA. combine with increase IPC (which i believe still pure speculation at this point?) and 80CU big navi will at least end up like 60% faster than RTX2080Ti. and that is not the biggest navi that speculated that can go as high as 128 CU.

so be wary of weird expectation.
Hahahahaha.........................

Calling it weird is an understatement.

Everything coming our recently is being more than weird and sometimes to the point of being physics defying.
 
  • Like
Reactions: renz496
well as usual things are get crazy as we nearing new GPU launch. and most often if nvidia coming out with something really good some people will start take that as a definite sign that AMD have something under their sleeve making nvidia very nervous. i still remember back then where some people very insisting that polaris will at least going to match nvidia GTX1070 with the price of $300 ($100 cheaper than 1070). they come to this conclusion after matching all the rumor and some of the limited info AMD telling the public back then like how polaris will be 100% more power efficient than their previous gen GPU.

so this time we got something like RDNA 2 is 50% more power efficient than the original RDNA. combine with increase IPC (which i believe still pure speculation at this point?) and 80CU big navi will at least end up like 60% faster than RTX2080Ti. and that is not the biggest navi that speculated that can go as high as 128 CU.

so be wary of weird expectation.
I think people should make up their mind if the upcoming NAVI flagship will have double the performance of RX5700XT or will it have 50-60% more performance than RTX2080Ti. Both statement cannot be true. Secondly 80CU and going as high as 128CU hmmmm. Nah that 128CU not gonna happen unless they wanna give us room heater no matter how efficient 7nm may be.
 
I think people should make up their mind if the upcoming NAVI flagship will have double the performance of RX5700XT or will it have 50-60% more performance than RTX2080Ti. Both statement cannot be true. Secondly 80CU and going as high as 128CU hmmmm. Nah that 128CU not gonna happen unless they wanna give us room heater no matter how efficient 7nm may be.

well wccftech just come up with interesting leak:

https://wccftech.com/amd-radeon-rx-...raphics-card-16-gb-vram-q4-2020-launch-rumor/

this probably one of the modest one about big navi so far.

  • no performance expectation yet. driver development probably not truly begin yet hence all the talk we heard about performance most likely pure speculation.
  • will have 16GB of VRAM as reference but might use GDDR6 type of memory (the rumored GDDR6X?) and not HBM.
  • expected to launch in sometime in Q4 with reference design only. custom board in Q1 next year.
 
Okay so RTX3080 to be 20% more powerful than RTX2080Ti. That makes it 50% more powerful than RTX2080 which is expected and have been in discussion for long. Nothing new. Besides being from "reliable source" well, lets see.

Anyhow that was the expected jump in performance. Nothing to go crazy over. If it was 30% jump over RTX2080Ti that would have me bit more interested. Not saying that 20% is bad, but it was expected.