RDNA2 Beyond the Round, the most up-to-date information - graphics cards - HWzone Forums
Skip to content
  • Create an account
  • About Us

    Hello Guest!

     
    Please note - in order to participate in our community, comment and open new discussions, you must join as a registered member.

    Our members enjoy many advantages, including the ability to participate in discussions, enjoy raffles and promotions for members of the site, and receive our weekly content directly by email.

    Do not like being harassed by email? You can register for the site but do not submit your registration to the weekly email updates.

RDNA2 Beyond the Round, the most up-to-date information


Recommended Posts

 

Official launch of On 28.10.2020 (this is already old news) But, reviewers' tests went online on 9.11.2020, 

And ticket sales start two days later on 11.11.2020 (ask for fresh and new information for us).

 

There will probably be three models in the new Blaine 21,22,23.

The top model navi 21 is also known as Big Navi is as it seems Double that of 10 previous known by name

RDNA1 or RX 5700XT. This time it will have 80CU's compared to the 40CU's in the previous model.

Something large in area, around 500 mm squares of principle size, with 2 times processing units from the previous model as mentioned.

Electricity consumption size 300 - 350 watts. So we probably need to be mentally prepared for the fact that this is the new trend in the industry

Graphic accelerators.

 

Beyond that, this is a change Assimilated, similar to the same slow evolution as they do in Risen processors (detail

First interesting below). Kerry also expects to improve performance in comparing transistor versus transistor in Outgoing.

 

It's another move to a more advanced 7nm lithograph , Which allows a little more transistors to the field, but no less important,

Has a working frequency of about 10-15%. Based on leaks from The new chip comes in the latest working frequency lithography

Equivalent to the new consoles, which use the same Exactly RDNA2, which is around 2200-2300Mhz.

 

The chip will be mated to proper Vram memory, with 16GB of 256-bit bandwidth with . But in order to compensate for the unimpressive width

This (which is equivalent to the previous model), Implemented its next evolution in the GPU world and here we have the first improvement to come out:

It is called an infinity cach with a capacity of 128MB and it does wonders for memory traffic, saving an order of magnitude of half.

Therefore, a bandwidth of only 256 bits is sufficient and there is no need for more than that.   

** The lower model navi 22 (trimmed) will have 192 bits and volume Of 12GB Vram. Probably with 60CU's active.

*** The lowest model in the Navi 23 trio will have 40CU's with 128-bit bandwidth and volume Of 8GB.

 

The model codes of the card will be RX6900 / 6800/6700.

When the 6700 replaces the 5700 from last year and improves its performance by about 10-20% (based on one Benz that has already leaked into the network).

Talk from Talk about rubbing salt in my wounds - d'oh!

But inferior to those of the Empire 3000 series. Also as an alternative to DLSS, AMD has developed its own equivalent that knows

Perform similarly - Render lower resolution, and combined with a smart upscale makes it effective. The messages are that the quality

The visuals are lower than those of DLSS 2.0, but on the other hand the performance is better. Which is like saying

DLSS lower quality but faster performance.

An advantage though, is that it works on all titles and not just on a limited collection of titles developed for them.

It's an aspiration that NVIDIA does not have DLSS, so in those titles AMD will be able to activate its patent that works in every title,

And to record a significant performance lead in performance, over its Naduaya in the same title there is no parallel trick.

 

AMD is trying this time to maintain a thunderous, very strange silence, one that is not typical of them. Some say the default assessment is that

"Because of shame for not having anything to show", This time upside down. They mean upside down. 

 

There were some who heard (accidentally or not - you can no longer know in today's world) from people from AMD, that Big Going

"Give in". Because performance at a level of 3080 and above should be expected. 

It has been said that, at least in a presentation from a month ago, AMD officially wrote in the context of GPU, that the date 28.10.2020 constitutes

leadership performance in the GPU domain.

I do not know at this time whether to give such official and arrogant statements any weight, or to ignore them altogether.

Days will tell, another month or so, so for now we will ignore.

 

But we do know one thing for sure right now:

And it's that 3080 is faster than 5700xt (which is RDNA1), 90% coarse (4K). And less so at lower resolutions.

So as long as RDNA2 big navi version improves at a level of 90% (or more) above RDNA1, we will have a worthy and interesting confrontation with 3080. 

 

now,

If the leaked data is correct, which is big if, that is, twice the size of the chip (the number of cores), accompanied by a frequency improvement of 10-15%,

And an algorithmic improvement accepted in the industry from generation to generation follows it, so there is something here that smells not bad at all and even with potential

Quite optimistic about achieving the goal. That 2 times size , Plus 10-15% frequency, and a few more percentages of generational improvement, it's more than twice.

 

This round I have a gut feeling, not beyond, but a strong feeling, that we are enjoying and the competition in the market will be effective.

More than that, this round I have a rough estimate, that AMD will bring it in double, even the most powerful processors in the world Zen3 (5000 generation),

And perhaps even for the first time in a decade, effective competition against NVIDA in the field of GPUs.

So another month we will know, not far away. 

 

 

 

 

 

 

Edited By nec_000
Post Reply Direct Link To This Post
Share on other sites
  • תגובות 340
  • Opens on
  • Last comment

Prominent participants in the discussion

Prominent participants in the discussion

Popular responses

I'm in favor of open source, and it's great that AMD is based on it and not on proprietary tools. The problem is that it does not rely on it because it supports open source, it relies on it because it has no choice. NVIDIA has a highly developed software department, and the industry is beating

Leave, he's a professional. He understands more than you. In fact, he understands more than you do about anything. In economics, in psychology, in everything related to technology and computer science. It's just a pity he did not learn how to write paragraphs in the forum without going down lines in line

I informed my wife that I was coming home early tonight, that she would make me a fresh watermelon with a Bulgarian in a family room / TV, next to it nuts, a bottle of cold Zero Cola, a Hapoel towel rising red, and that she would take the kids out somewhere ...

Attached Images

Quote of napoleon45

And no less important, wait for NV's contract as well.

 

What is left as a contra to bring?

Their GA102 chip is already 628mm in size, huge, there is really nowhere to grow further with it.

 

So what it can provide as a contra, will be in the 3080 Super or Ti style, with dual memory = 20GB.

She can also try to play with the range she has between 3080 and 3090, but it is not large,

 There's a whole 10% tiny margin to utilize it and that's it. It's nothing.

 

Namely:

If at some surprising length, let's say the need charged at the moment and not beyond,

RDNA2 is in the caliber of 3080/3090, in such a case for Nido ... except to provide a product with more

Memory and try to give a lower price, no extra maneuvering space. It is stretched to the edge with the 3000 series.

 

In the extreme case, the GA102 chip has a total of 2% of unused processing cores. We have also seen working frequencies, left no room

For substantial or significant haste. The tickets are marketed from the factory at a working frequency that is close to the maximum that can be reached

Via samsung 8nm. Maybe there is a tight squeeze, 3-5% more.

 

 This time for Nodia really there is no room to surprise or maneuver with, was and Comes out with something serious / competitive in front of her.

 

And produce something new, in a new, more minimized lithograph, like moving to , It's not from now until now.

This is a minimum year if done with great difficulty, at a huge cost, and a great risk.

 

 

In my opinion the caution, Could not hope for opportunity and time Better than those, when as it seems, Noida

Made (in retrospect) a mistake that bet on Samsung's 8nm. Since currently in the possession of Advantage at the production level of

Lithography. TSMC's 7nm is ahead of that of In a generation. It is around 25-30% denser, and more efficient

Energetic. This is a significant gap, which gives a significant advantage in her hand. It also allows for a high working frequency of about 10-15%.

 

While in the past the algorithmics / architecture of Nuida knew how to compensate for such gaps and not allow to get around

It is even used in one more advanced processor, this time and after narrowing the architectural gaps that AMD

Achieved, it will already be more difficult to perform.

 

 

 

Edited By nec_000
Post Reply Direct Link To This Post
Share on other sites

Thanks for the info..there is a lot of new here.

Especially the whole issue of dlss and RT ... I have not heard of it anywhere yet ...

Allow link to source?

This is what it sounds like to be good to be true ..

Hopefully the hype of Justified unlike the other times ...

 

By the way, why is 80 cu equivalent in NVIDIA?

They mainly use the term cuda cores .. and there are thousands of them ..

Edited By amit14
  • I liked! 1
Post Reply Direct Link To This Post
Share on other sites
Quote of amit14

Especially the whole issue of dlss and RT ... I have not heard of it anywhere yet ...

RT has been introduced by In fact on several occasions, about six months ago:

 

 

Edited By No-One
Post Reply Direct Link To This Post
Share on other sites
No-One Quote

RT has been introduced by AMD In fact on several occasions, about six months ago:

 

 

It is known..I meant a section that is better than Turing ..

 

By the way, at the moment it does not seem that the amp is much more impressive in this section ... minor improvement..already a serious impairment in performance ..

Post Reply Direct Link To This Post
Share on other sites
Quote of amit14

Still serious damage to performance ..

When working with RT also DLSS2 (and there is no reason not to, because you get a good image at least like the rotated original resolution) there is no real performance impairment, run RT without DLSS and say there is performance impairment it goes into small, these two technologies were developed and launched together for a very clear reason Run together to prevent performance damage.

 

Post Reply Direct Link To This Post
Share on other sites
No-One Quote

When working with RT also DLSS2 (and there is no reason not to, because you get a good image at least like the rotated original resolution) there is no real performance impairment, run RT without DLSS and say there is performance impairment it goes into small, these two technologies were developed and launched together for a very clear reason Run together to prevent performance damage.

 

True ... but even with touring it's like that ... my intention is that there is no real improvement in these aspects between the two ..

And right now not every game with Rt including dlss .. although I guess that's the direction ..

Post Reply Direct Link To This Post
Share on other sites
Quote of amit14

Even at the moment not every game with Rt includes dlss

True, a vivid and ugly example of this is COD MW, which apparently after the global community set up a shriek against the creators, they managed to put in the latest brand - COD BO COLD WAR support for DLSS2 as well, since it makes no sense to run RT in MULTI games, even if the player is not FPS chasers.

 

But like you said, this is really the trend and I believe it started from the 30XX series of The games will contain RT and DLSS2, or DLSS2 alone, but not RT as a single feature and will cause the game performance to crash.

Post Reply Direct Link To This Post
Share on other sites

 

In the meantime another source that says similar things in some, and more different (the amount of cores of the cheaper models, and electrical consumption).

 View:

 

 

Edited By nec_000
Post Reply Direct Link To This Post
Share on other sites
Quote of danielzy

I just do not see a situation, from a business point of view, to maintain such silence.

They are late for launch for about two months, two full months in which they lose tens of thousands of potential customers.

The only logical reason for this is that their product is simply significantly inferior.

There is also a matter of schedules. Amad (and we know this from reports from half a year ago) originally intended its launch to happen at the end of the year, just before the next generation of consoles (which is really the October-November region).

 

We may be living in a romantic age where there is a need for "gladiatorial battles week after week," but in practice these things are more businesslike.

 

What's more, anyone interested in a market offer somewhere would have waited a few weeks to stock up anyway, and would not have thrown in paying 3090 on 3080 sold at NewEgg or. So even if Amad had released a launch two days later, it would have been locked into its purchase. 

 

Anyway, as you wrote in the text, RDNA2 is not "DOA" like , And Nabidia understands this. Therefore the pricing "dripped" backwards in the main segments.

We will probably get one ticket in the 3080 area and one ticket in the 3070 area, which is more or less in expectations and an impressive achievement for (Given its current state).

 

I am more interested in its Reconstruction technology. There's a similar thing about the PS5, which is supposed to replace the "Checkerboard rendering" method used on games like Detroit (and technically much inferior), and will probably function much better like DLSS. The remake of the Demons Souls is supposed to be used to mimic a situation "Dynamic" with 60FPS. I have no idea beyond that because it is not talking about anything, but would be interesting to see. 

 

In my opinion, as long as Amad will provide a worthy competitor to the 3080, with a similar offering to Ray Tracing and DLSS, there is a lot to talk about. 

Post Reply Direct Link To This Post
Share on other sites

The information becomes more and more refined as time goes on.

In the next video more details, and more accuracy.
But also quite a few informed assessments, at those points where there is no sufficient information out there.
Worth watching in my opinion:

 

Edited By nec_000
Post Reply Direct Link To This Post
Share on other sites
  • djelectric locked this discussion
guest
This discussion is locked for new replies.

×
  • Create new ...

At the top of the news:

new on the site