[Rumor, Tweaktown] AMD to launch next-gen Navi graphics cards at E3

[Glo.]

^This. We need to wait till E3, maybe we will get a little more info on Navi and actually have something to talk about.

Its 9 days away. Anyone knows whether will be there any livestream, or what time will it be?

 

[RaV666]

I asked this beginner99, I will ask you the same thing.

If Majority of market having the option between GTX 1050 Ti, and 40% faster, and 20$ most of the time RX 570 still picked GTX 1050 Ti, what makes you believe that it would be anything different with Navi GPU?

And if so, what benefit would be for AMD to price their GPU at a discount, compared to Nvidia's offering, and still do not get any sales?

No. Nvidia ALWAYS Gets more sales. No matter what will happen Nvidia will always make more sales. Because they have mindshare, which AMD does not have.

There is no incentive for AMD to:
A) design powerful GPUs if they cannot sell them
B) to sell their GPUs at a discount compared to Nvidia, if they still cannot sell them.
Funniest part, is that if the rumors are 100% true, about AMD and Nvidia, Nvidia appears to be the savior of consumers, if they will really lower the MSRPs of their GPUs...
And what will you get if you will add this:

to 40W more? You get 268W of power. Isn't it 275W TDP/TBP?

COMPARE THE CUSTOM CARDS TO CUSTOM CARDS. ALWAYS. COMPARE REFERENCE DESIGNS TO REFERENCE DESIGNS. ALWAYS. And do it AFTER WE WILL HAVE ANY REVIEWS of GPUs discussed.

There is no TDP and TBP rating official from AMD. Only Rumored ASRock designs. Why do you fail to grasp this concept? They are NOT reference designs.

First of all you a re very very sad person to go on the lookout for the highest power consumption custom model , than end up not getting the result you wanted XD.
So you took extreme case, and failed. 268W is not 275W.
Thats because you lied.
Second of all , the discussed 225W TBP was amd based , so no, not a custom liquid cooling model or something along those lines.
You can also rest assured that those extreme models are faster than base or FE models, and it was not the model amd was comparing itself to.
EOT.

 

[Glo.]

Second of all , the discussed 225W TBP was amd based , so no, not a custom liquid cooling model or something along those lines.

EOT.

And where, the ****, AMD has announced ANYTHING about Navi, apart from series name?

Nowhere. Nothing officially, apart from Radeon RX 5700 Series exists. No SKUs, No performance, Prices, no TBP/TDP. Only Rumored(!) TDP and TBP based on ASRock's information, which even this is UNOFFICIAL.

What are you basing on, your assumptions? Imagination?

 

[RaV666]

And where, the ****, AMD has announced ANYTHING about Navi, apart from series name?

Nowhere. Nothing officially, apart from Radeon RX 5700 Series exists. No SKUs, No performance, Prices, no TBP/TDP. Only Rumored(!) TDP and TBP based on ASRock's information, which even this is UNOFFICIAL.

What are you basing on, your assumptions? Imagination?

You are one of those people that cant have a logic argument.Because the argument always changes.
This is what you wrote before
"There is not one single Navi GPU that has higher TDP than 180W. "

Well, how you can write such a thing , if there is nothing official ?
Arent you contradicting yourself here ?
OR, you were discussing TDP`s and TBW`s on those unofficial numbers just a page or two before ?

BTW there is one sku and one performance XD, well even two "performances" known.

 

[DrMrLordX]

If they can really beat RTX 1070 across the board by 10%, that's a $399 card at most if AMD wants it to sell. If it matches RTX 1070 in most games and sometimes falls short by 5% or so, then it's a $329-$349 card if they want it to sell.

AMD also has to do two other things to sell their products:

1). Advertise them well.
2). Deal effectively with the price/performance positioning of their past products.

To expand upon point #2, it makes a certain amount of sense to ask, "how many people are willing to pay $x for a dGPU? What performance % did they expect with reference to the top dGPU on the market when they made that purchase?". AMD isn't selling a 2080Ti killer here, so they are limited in what they can charge, mostly based on how much people are willing to pay for a given % of the top dGPU's performance (in this case, the 2080Ti).

If AMD expects people to start paying a price for midrange-and-lower cards that scales upwards with the halo card prices without giving them a better cut of the performance pie, then they will probably suffer from reduced sales, and they'll wind up killing off what's left of the consumer dGPU market faster than if they tried to hit price points from years past.

RX Vega 64 and 56 were overpriced as gamer cards, so obviously AMD shouldn't be using them as a price reference. They should be looking more at Polaris (particularly the RX 480 and RX 580).

 

[Glo.]

You are one of those people that cant have a logic argument.Because the argument always changes.
This is what you wrote before
"There is not one single Navi GPU that has higher TDP than 180W. "

Well, how you can write such a thing , if there is nothing official ?
Arent you contradicting yourself here ?
OR, you were discussing TDP`s and TBW`s on those unofficial numbers just a page or two before ?

BTW there is one sku and one performance XD, well even two "performances" known.

If I will say that I am not contradicting myself will it change anything?

I also was the one who was adamant about Navi design having 40 CUs while others believed it to be 48-56 CUs. And yet, nothing is still official .

 

[RaV666]

If I will say that I am not contradicting myself will it change anything?

I also was the one who was adamant about Navi design having 40 CUs while others believed it to be 48-56 CUs. And yet, nothing is still official .

And that is why we are discussing things, based on best knowledge we have so far, not only on fully official verified and crosstested information.
Because its fun.
I also believe that these things have 40Cus, even tho , oh the horror, it is not official information .
There was a leak in wccftech like 8 months ago with all the info pretty much. 1080 performance and 40CU`s.
The series name is RX 5000, but amd showed RX 5700 SKU in both special 3dmark pcie bandwith test and in strange brigade.
The assumption being, that is the strongest model (probably 225W) . But it may not be the strongest model.
Anyhow, seems like nvidia reacts with new 2060/2070/2080 super models and adjusts prices.
Of course , this is also a leak and not anything official...

 

[Glo.]

And that is why we are discussing things, based on best knowledge we have so far, not only on fully official verified and crosstested information.
Because its fun.
I also believe that these things have 40Cus, even tho , oh the horror, it is not official information .
There was a leak in wccftech like 8 months ago with all the info pretty much. 1080 performance and 40CU`s.
The series name is RX 5000, but amd showed RX 5700 SKU in both special 3dmark pcie bandwith test and in strange brigade.
The assumption being, that is the strongest model (probably 225W) . But it may not be the strongest model.
Anyhow, seems like nvidia reacts with new 2060/2070/2080 super models and adjusts prices.
Of course , this is also a leak and not anything official...

If those leaks were correct, why nobody has come up with RDNA name for the Architecture?

Why for months everybody referenced RX 3060/3070/3080 as a name for Navi GPUs, which turned out to be incorrect?

And lastly. Why has dr Lisa Su said blatantly at the event that there is not many people who know what Navi is?

 

[RaV666]

If those leaks were correct, why nobody has come up with RDNA name for the Architecture?

Why for months everybody referenced RX 3060/3070/3080 as a name for Navi GPUs, which turned out to be incorrect?

And lastly. Why has dr Lisa Su said blatantly at the event that there is not many people who know what Navi is?

Well, "RDNA" name they could end up with like a day before , its just marketing name for real architecture name navi.
The whole RX 3xxx series was started by adoredtv, and 90% of other people copy him.
wccftech leak came way before that.And they didnt state name, they just said it will be new architecture, which it KIND OF is, but not fully.
As for miss Su, the misinformation in recent months mostly by adoredtv reached so many people that she probably meant exactly this.It doesnt mean that probably hundreds if not thousands of people have to have some insider knowledge.Too many people in different companies worked on or around it for like nobody to know.
Real leaks happen, whole chiplet thing, redgaming tech knew about VII even tho most people didnt believe v20 was coming to deskktop.Rtx name before that.etc etc.

 

[Glo.]

Well, "RDNA" name they could end up with like a day before , its just marketing name for real architecture name navi.
The whole RX 3xxx series was started by adoredtv, and 90% of other people copy him.
wccftech leak came way before that.And they didnt state name, they just said it will be new architecture, which it KIND OF is, but not fully.
As for miss Su, the misinformation in recent months mostly by adoredtv reached so many people that she probably meant exactly this.It doesnt mean that probably hundreds if not thousands of people have to have some insider knowledge.Too many people in different companies worked on or around it for like nobody to know.
Real leaks happen, whole chiplet thing, redgaming tech knew about VII even tho most people didnt believe v20 was coming to deskktop.Rtx name before that.etc etc.

You really believe that Architecture Name(!) they would make one day before the event, and not months before?

About adored. He also spread the rumor that Navi RX 3080 will cost 250-300$. I presume he was also correct on this?

Anyways. 9 days to go. According to Robert Hallock, if Computex was 80% Zen 2, and 20% Navi, E3 will be other way around: 80% Navi and 20% Zen2 Gen 3 CPUs.

Full architecture disclosure? I wonder if they will talk about larger L1 caches in Navi...

 

[RaV666]

You really believe that Architecture Name(!) they would make one day before the event, and not months before?

About adored. He also spread the rumor that Navi RX 3080 will cost 250-300$. I presume he was also correct on this?

Anyways. 9 days to go. According to Robert Hallock, if Computex was 80% Zen 2, and 20% Navi, E3 will be other way around: 80% Navi and 20% Zen2 Gen 3 CPUs.

Full architecture disclosure? I wonder if they will talk about larger L1 caches in Navi...

Architecture name they have for YEARS its "NAVI" .But for whatever reason they wanted something new marketing wise, probably they wanted to get rid of GCN moniker, they tried this with vega already.But somehow people forgot about whole NGCU .
About adored i dont know why u are using phrase "he was also correct on this" , because that implies he was right about rx 3xxx, and he wasnt so i dont know what are u refrerring to.
As for the price, i have no idea. i THINK it should be 399/299.But thats only what i would do.
Regarding cache sizes, i was under the impression that L2 doubled.

 

[Glo.]

About adored i dont know why u are using phrase "he was also correct on this" , because that implies he was right about rx 3xxx, and he wasnt so i dont know what are u refrerring to.

The question mark, at the end of that sentence implies it is rhetoric question.

 

[Kenmitch]

^This. We need to wait till E3, maybe we will get a little more info on Navi and actually have something to talk about.

Yep. Arguing on about the unknown is silly, but reading it can be amusing.

 

[maddie]

Pure speculation here. 2 Navi designs incoming?

If the (40 CU ?) larger is 255mm^2, then we should estimate the smaller (20CU ?) at ~ 160mm^2. Assuming 7nm, and NOT 7nm+, with the same defect density as Zen2 (0.6 to get a leaked yield of 70% full die Zen2), we get a yield of 26% full die for the bigger model and 42% for the smaller one. These are very poor yields for a lower priced product assuming no redundancy in the full die. From this, we should expect that there will be extra circuitry in even the full die (CU, ROPS, even extra cache blocks) that will not be activated simultaneously. Maybe that's one reason why it's relatively large for a 40 CU 7nm GPU.

Based on Polaris, I really expected comfortably less than 200mm^2 even with larger caches plus AMD previously mainly went for higher density in GPUs.

I remember when the 4870 models arrived, AMD was found to have doubled up certain difficult elements and only used 1 as needed in case of defects. This allowed them to achieve much higher effective yields on the 55nm process.

Edit:
corrected the smaller die yield for the given.

 

[prtskg]

Here is AMD's Bridgman comment about Navi at Phoronix -

You could call it a hybrid but not in that sense... we used to talk about GCN as an ISA, but it seems that most people outside AMD think of GCN as a micro-architecture instead (ie an implementation of the ISA). RDNA is GCN ISA but not what you think of as GCN architecture.
https://www.phoronix.com/forums/for...-in-linux-5-3-mesa-19-2?p=1103266#post1103266
Whole discussion -
https://www.phoronix.com/forums/for...n-rx-5700-navi-support-in-linux-5-3-mesa-19-2

 

[Glo.]

Here is AMD's Bridgman comment about Navi at Phoronix -

You could call it a hybrid but not in that sense... we used to talk about GCN as an ISA, but it seems that most people outside AMD think of GCN as a micro-architecture instead (ie an implementation of the ISA). RDNA is GCN ISA but not what you think of as GCN architecture.
https://www.phoronix.com/forums/for...-in-linux-5-3-mesa-19-2?p=1103266#post1103266
Whole discussion -
https://www.phoronix.com/forums/for...n-rx-5700-navi-support-in-linux-5-3-mesa-19-2

Directly translates to: GCN ISA with cleen sheet design. Not comparable to previous versions of AMD GPUs.

 

[Glo.]

Pure speculation here. 2 Navi designs incoming?

If the (40 CU ?) larger is 255mm^2, then we should estimate the smaller (20CU ?) at ~ 160mm^2.

Nailed it .

 

[RaV666]

The question mark, at the end of that sentence implies it is rhetoric question.

I am pretty sure that question mark at the end of a sentence means only that it is a question...
Also, i dont think it "implies" the question part .
And .Well. The rhetoric part would work if we knew the price.

 

[beginner99]

No. Nvidia ALWAYS Gets more sales. No matter what will happen Nvidia will always make more sales. Because they have mindshare, which AMD does not have.

There is no incentive for AMD to:
A) design powerful GPUs if they cannot sell them
B) to sell their GPUs at a discount compared to Nvidia, if they still cannot sell them.
Funniest part, is that if the rumors are 100% true, about AMD and Nvidia, Nvidia appears to be the savior of consumers, if they will really lower the MSRPs of their GPUs...

And you simply ignore my point about 570 vs 1050 ti power draw while at the same time posting a chart showing the huge difference. 1050 ti outsold 4/570 for mostly that reason at it fits nicley into prebuilts for example or is quieter in a htpc.
There is more than just performance. if performance/$ is similar, performance/watt will matter a lot.

 

[Karnak]

But Performance/$ wasn't similar between the 570 (4 GiB) and the 1050Ti.

 

[Glo.]

And you simply ignore my point about 570 vs 1050 ti power draw while at the same time posting a chart showing the huge difference. 1050 ti outsold 4/570 for mostly that reason at it fits nicley into prebuilts for example or is quieter in a htpc.
There is more than just performance. if performance/$ is similar, performance/watt will matter a lot.

You do realize that you are talking about 40% faster GPU(RX 570) that costed for months the same, or LESS than GTX 1050 Ti? Since half of last year, RX 570 was going for 130-150$. At the time GTX 1050 Ti was costing 160-170$.

How is that even comparable in Performance/$? That is what I was talking about. GTX 1050 Ti outsold RX 570, even when RX 570 was cheaper to buy than GTX 1050 Ti.

Why? Majority of market are HTPC users?

 

[JDG1980]

Architecture name they have for YEARS its "NAVI" .But for whatever reason they wanted something new marketing wise, probably they wanted to get rid of GCN moniker, they tried this with vega already.But somehow people forgot about whole NGCU .

That's because Vega didn't really do any better than Fiji in terms of perf/TFlop, so it was obvious that it wasn't really a new architecture but just another minor iteration on the original GCN microarchitecture.

 

[JDG1980]

I asked this beginner99, I will ask you the same thing.

If Majority of market having the option between GTX 1050 Ti, and 40% faster, and 20$ most of the time RX 570 still picked GTX 1050 Ti, what makes you believe that it would be anything different with Navi GPU?

The GTX 1050 Ti has no 6-pin connector. This means it's suitable for use in cheap OEM systems with subpar PSUs that cannot handle the RX 570.

You underestimate the number of systems where power consumption/TDP actually does matter. AMD has been significantly worse on this metric since Maxwell (at least if you care about AAA gaming and not GPGPU).

No. Nvidia ALWAYS Gets more sales. No matter what will happen Nvidia will always make more sales. Because they have mindshare, which AMD does not have.

You could have said the same thing about AMD vs Intel back in the Bulldozer days. But it turns out that wasn't true. The problem wasn't mindshare, it was that AMD just didn't have competitive products. Once they brought out Ryzen, the "mindshare" problem in the enthusiast market mostly went away. But to win back that mindshare, they did have to provide significantly better perf/$.

 

[maddie]

You do realize that you are talking about 40% faster GPU(RX 570) that costed for months the same, or LESS than GTX 1050 Ti? Since half of last year, RX 570 was going for 130-150$. At the time GTX 1050 Ti was costing 160-170$.

How is that even comparable in Performance/$? That is what I was talking about. GTX 1050 Ti outsold RX 570, even when RX 570 was cheaper to buy than GTX 1050 Ti.

Why? Majority of market are HTPC users?

Exactly this and it says that for the near future at least AMD either settles for lower margins than the competition and work at rebuilding their brand or they abandon any attempt to reclaim significant marketshare. As it is right now, they have gotten some gains recently but this is probably driven by excellent present prices on 570 & Vega models. Abandoning this to price closely to Nvidia is foolish long term, in my opinion. It sucks for them, but it is what it is.

In my country, pretty much no one wants AMD or even know them, and you have to search all over for any model at ridiculous prices. Nvidia models however are ubiquitous. I remember when polaris came out, no one stocked them, but there were 290 models still selling for ~$ 3300TT = $500US. A graphics card is Nvidia almost by default. luckily it's easy for us to buy stuff in the US and ship. Quite cheap actually, so I buy all my stuff from US retailers.

 

[railven]

^This. We need to wait till E3, maybe we will get a little more info on Navi and actually have something to talk about.

Like Glo. said a few pages back, it's a clean sheet design and you can't make any assumptions/guesses on it, because nothing like it has existed before! I mean...oh wait.

This is perhaps the worst hype thread ever. There is literally nothing to spin and we're at like page 37. Moment you make any kind of prediction you're shut down because 'CLEAN SHEET DESIGN'!!!