Tuesday, July 7th 2020

NVIDIA GeForce RTX 3070 and RTX 3070 Ti Rumored Specifications Appear

NVIDIA is slowly preparing to launch its next-generation Ampere graphics cards for consumers after we got the A100 GPU for data-centric applications. The Ampere lineup is getting more and more leaks and speculations every day, so we can assume that the launch is near. In the most recent round of rumors, we have some new information about the GPU SKU and memory of the upcoming GeForce RTX 3070 and RTX 3070 Ti. Thanks to Twitter user kopite7kimi, who had multiple confirmed speculations in the past, we have information that GeForce RTX 3070 and RTX 3070 Ti use a GA104 GPU SKU, paired with GDDR6 memory. The cath is that the Ti version of GPU will feature a new GDDR6X memory, which has a higher speed and can reportedly go up to 21 Gbps.

The regular RTX 3070 is supposed to have 2944 CUDA cores on GA104-400 GPU die, while its bigger brother RTX 3070 Ti is designed with 3072 CUDA cores on GA104-300 die. Paired with new technologies that Ampere architecture brings, with a new GDDR6X memory, the GPUs are set to be very good performers. It is estimated that both of the cards would reach a memory bandwidth of 512 GB/s. So far that is all we have. NVIDIA is reportedly in Design Validation Test (DVT) phase with these cards and is preparing for mass production in August. Following those events is the official launch which should happen before the end of this year, with some speculations indicating that it is in September.
Sources: VideoCardz, TweakTown, kopite7kimi (Twitter)
Add your own comment

104 Comments on NVIDIA GeForce RTX 3070 and RTX 3070 Ti Rumored Specifications Appear

#76
Valantar
RandallFlagg
Uh no, that is *NOT* what you said........ You are now backtracking.

What you said was (emphasis added) :



Clock speed has nothing to do with IPC increase from new nodes.

If I have a microcode instruction that completes in 1.1 cycles, it will have to wait for the next (2nd) cycle for anything to be done with the result. This essentially means it takes 2 cycles to complete in a useful way. If I improve the time it takes to complete that instruction by 20% (a common claim by TSMC) it now takes ~0.9 cycles, it now went from being a 2 cycle instruction to a one cycle instruction. This *directly* impacts IPC.

You are welcome for the education.
Okay, I see I was a bit more categorical than I ought to have been. But nonetheless, any increase in transistor switching speed from a node change is likely to be used first and most critically to increase clock speeds. Why? Because this is easier to design for (just speed up the entire design rather than a specific part) and increases performance for everything, not just that one instruction. That instruction being sped up also of course only helps if the rest of the pipeline either is waiting for data from it or is sped up equally. While what you describe no doubt exists, it is a marginal effect compared to clock speed increases, and a high effort one too. I would also highly appreciate even a single demonstrable example of this, i.e. of a node change actually increasing IPC on a way that can reliably be ascribed to that and not to more explicit architectural tweaks.
Posted on Reply
#77
medi01
mtcn77
Think of it this way: LCD beats OLED in every manner apart from pixel transitions.
Wah? Black levels, cough? (and, for the love of god, don't mention FALD)
Posted on Reply
#78
mtcn77
medi01
Wah? Black levels, cough? (and, for the love of god, don't mention FALD)
Black level is a product of contrast. OLED isn't inherently bright. I will update if I can get the source remembered.
Posted on Reply
#79
medi01
RandallFlagg
If I have a microcode instruction that completes in 1.1 cycles
Hm, I remember attending classes explaining how basic components of that kind work. That was a while ago though.

Now you got me thinking, when Intel stated back in 90th that most common instructions are done in one clock, e.g. division... I need to think about it more, but indeed, what you state makes sense.

Say I have a block which is a chain of transistors that need to complete in 1 clock cycle. Having transistors switch faster, would allow me to build longer/wider chains and do crazier stuff in one clock. (how easily that translates into IPC gains is far from obvious, but clearly it could)

PS
Your posts come off somewhat arrogant, it would help if you would tone it down.
mtcn77
Black level is a product of contrast. OLED isn't inherently bright. I will update if I can get the source remembered.
Black level is absolute and OLED wipes the floor with LCD here.

Contrast is relative: black vs bright dot and LCD can be much brighter than OLEDs can afford (burn in) and achieve better contrast that way.

Still, our eyes do tell the difference.
Posted on Reply
#80
ARF
mtcn77
Black level is a product of contrast. OLED isn't inherently bright. I will update if I can get the source remembered.
OLED has defects like the legacy plasma screens.
VA LCDs have perfect blacks. I don't think any OLED has better black than the best VA.
Posted on Reply
#81
medi01
ARF
I don't think any OLED has better black than the best VA.
OLEDs have literally 0 lux blacks, what the heck are you talking about?

Plasma had issues with whites, not blacks.
Posted on Reply
#82
mtcn77
medi01
Black level is absolute and OLED wipes the floor with LCD here.
You don't understand and this is the killer fact: black isn't itself a 'level'. It is darkness, hence a reflection(shadow) of light level. If brightness isn't on par, it is just dark unconditionally.
I honestly speak far and wide, bringing little evidence. There was a side by side display comparison. It was the same with igzo-ltps difference.
Posted on Reply
#83
medi01
mtcn77
fact: black isn't itself a 'level'.
What the heck, of course it is, it's zero lumen:
en.wikipedia.org/wiki/Lumen_(unit)

As for "but some dudes on youtube make convincing argument that Samsung's Q-Lies are better than they are", welp, that's in the eye of beholder, mine call out BS.
Posted on Reply
#84
mtcn77
medi01
What the heck, of course it is, it's zero lumen:
en.wikipedia.org/wiki/Lumen_(unit)

As for "but some dudes on youtube make convincing argument that Samsung's Q-Lies are better than they are", welp, that's in the eye of beholder, mine call out BS.
I see my post needs further explanation...
Posted on Reply
#85
medi01
mtcn77
I see my post needs further explanation...
No, I'd appreciate if you would not try to redefine "better black" one more time.
Posted on Reply
#86
mtcn77
medi01
No, I'd appreciate if you would not try to redefine "better black" one more time.
Brightness:

Champion - LCD & LED

When it comes to brightness, LED TVs have a modest asset. LEDs are just really good at getting extremely bright. OLED TVs can get bright, too, but cranking OLED pixels to maximum brightness for extended periods not only reduces that pixel’s lifespan, but the pixel also takes a little while to return to total black.
Any objection?
Posted on Reply
#87
Valantar
mtcn77
Any objection?
Yes. All that says is that at sustained high brightness, it takes OLEDs a bit of time to get back to total black (while an LCD's ability to quickly dim would not be similarly affected). That still doesn't remove the OLED's ability to reach total black, nor change the fact that LED-backlit LCD panels can never reach total black.
Posted on Reply
#88
mtcn77
Valantar
Yes. All that says is that at sustained high brightness, it takes OLEDs a bit of time to get back to total black (while an LCD's ability to quickly dim would not be similarly affected). That still doesn't remove the OLED's ability to reach total black, nor change the fact that LED-backlit LCD panels can never reach total black.
I see that needs further explanation...
Posted on Reply
#89
Valantar
mtcn77
I see that needs further explanation...
Yes, you do need to explain further how 0 lumens is somehow more bright than a value above 0 lumens, and/or how the rate of dimming from a light source somehow correlates to its minimum light output.
Posted on Reply
#90
mtcn77
Valantar
Yes, you do need to explain further how 0 lumens is somehow more bright than a value above 0 lumens, and/or how the rate of dimming from a light source somehow correlates to its minimum light output.
It is intriguing that LCD can achieve comparable or even slightly better MPRT and ACR than OLED, although its response time and contrast ratio are generally perceived to be much inferior to those of OLED.
You have to harrass your eyes with total darkness, in order to take OLED into the lead. Which is false as we know it... 300lux 0.5 m² panel will cast ambient light at a level of 150 candles. It will be equal to two 75w bulbs. Have fun with nonexistent contrast levels.

K, so 16:9 aperture ratio 43" panel 300cd/m² bright will emit the same as a 150w bulb when displaying white image.
Posted on Reply
#91
medi01
mtcn77
You have to harrass your eyes with total darkness, in order to take OLED into the lead. Which is false as we know it... 300lux 0.5 m² panel will cast ambient light at a level of 150 candles. It will be equal to two 75w bulbs. Have fun with nonexistent contrast levels.

K, so 16:9 aperture ratio 43" panel 300cd/m² bright will emit the same as a 150w bulb when displaying white image.
Phenomenal.
You've just rendered OLED and FALD TFT irrelevant.
In a thread about 3080.

That's a 10/10.
Posted on Reply
#92
mtcn77
medi01
Phenomenal.
You've just rendered OLED and FALD TFT irrelevant.
You should feel proud.
As expected, when the ambient light is weak, OLED shows a much higher ACR than
LCD. But as the ambient light gets stronger, two ACR curves get much closer. At 300 lux
(moderate indoor lighting), LCD shows comparable ACR to OLED (140.1 vs. 150.6). If we
slightly increase the peak brightness of LCD (by increasing the backlight intensity) to 800
nits, two ACR curves crossover at 90 lux (typical lighting condition in office building
hallway or toilet lighting). It means below 90 lux, OLED (with 600 nits of peak brightness)
exhibits a higher ACR, but beyond 90 lux the situation is reversed for the LCD with 800 nits
of peak brightness.
You dare disagree with me? :)
Posted on Reply
#94
mtcn77
medi01
11/10
Have it your way.


This is what happens when you have developed your whole skill tree on astroturfing.
Posted on Reply
#95
Valantar
mtcn77
Have it your way.


This is what happens when you have developed your whole skill tree on astroturfing.
a) How is this even remotely relevant to the thread? Start your own thread with this as the topic if it's that important to you.
b) You have this nice tendency to quote random things without providing sources. Is inserting a link really that difficult?
c) A graph with unnamed axes? Yeah, that tells us nothing. Percent of what?
d) "Harass your eyes with total darkness"? Our eyes are capable of discerning a wider dynamic range than any TV is capable of displaying, and any camera is capable of capturing. Having sufficiently low black levels is essential for any display panel that is going to be used in a typical home (non-controlled) environment. I certainly watch a lot of stuff with very dim room lighting.
Posted on Reply
#96
mtcn77
Valantar
a) How is this even remotely relevant to the thread? Start your own thread with this as the topic if it's that important to you.
b) You have this nice tendency to quote random things without providing sources. Is inserting a link really that difficult?
c) A graph with unnamed axes? Yeah, that tells us nothing. Percent of what?
d) "Harass your eyes with total darkness"? Our eyes are capable of discerning a wider dynamic range than any TV is capable of displaying, and any camera is capable of capturing. Having sufficiently low black levels is essential for any display panel that is going to be used in a typical home (non-controlled) environment. I certainly watch a lot of stuff with very dim room lighting.
Posted on Reply
#97
Valantar
mtcn77

Congratulations on quoting a post and then addressing exactly 0% of the contents of said post, I guess?
Posted on Reply
#98
mtcn77
Valantar
Congratulations on quoting a post and then addressing exactly 0% of the contents of said post, I guess?
All your questions are exaggerated criticism. Don't you think you need to refrain a little?
I don't like biased discussions. If you don't like my references, how about you provide a little yourself, so the debate wouldn't be one sided...
Posted on Reply
#99
Valantar
mtcn77
All your questions are exaggerated criticism. Don't you think you need to refrain a little?
I don't like biased discussions. If you don't like my references, how about you provide a little yourself, so the debate wouldn't be one sided...
You haven't provided a single reference, you have provided unattributed quotes with no sourcing. Where are the quotes from? Do you have a link? Is embedding it really that hard (there's a button for that in the toolbar above the post editor)?

As for my points being "exaggerated criticism": I would love for you to explain what exactly is exaggerated about it. The question of relevance to the thread topic is a straightforward question - this is a topic about rumored GPU specifications; you are talking about TV panel technologies. How is the latter relevant to the former? Asking you to provide sources for your quotes is a straightforward thing too. Sure, I guess you could call me biased against unsourced claims - I'll gladly accept that. Beyond that, I'm simply asking you to show us where you are getting your data from. There is no bias in this. The same goes for pointing out that you posted an image of a graph with unlabeled axes and no explanation - are we supposed to guess what the numbers represent? Asking you to provide some more information on what exactly you are trying to say is not bias. As for my latter point, I was just pointing out the absurdity of you simultaneously arguing that a) QLEDs have better contrast than OLEDs and b) that the total black of OLEDs "harass[es] your eyes", which must be understood as the low black level being harsh and difficult to look at in an otherwise normally lit room, which would mean that the OLED's image has a higher dynamic range than the QLED.

I am not the one making new claims here, thus I have no obligation to provide proof for my claims except if necessary to counter any proof you might bring to the table. As the one making new claim, the onus is on you to show evidence for these claims. So far your "evidence" has consisted of a couple of graphs with zero indication of what the graphs actually show. Rejecting this isn't bias, it is just reflective of you failing to present a comprehensive and convincing argument. And as I have strongly implied, this doesn't belong in this topic whatsoever, and should be moved to a topic dedicated to this discussion.
Posted on Reply
#100
mtcn77
Another unaccredited wall of text with no reference.

It seems astroturf pays big time. Why "trolling" has a negative connotation is beyond me.
Posted on Reply
Add your own comment