Can't believe people are still shilling Navi and think AMD will offer $800 card for $250

hmb2 is only on radeon 7 cards, i dont think that navi will use hbm2 cause that would make no fucking sense

These are the reasons why I personally find it plausible at least:
1.) VEGA 56 MSRP at launch was $399 and meant to compete with the GTX 1080 on paper. It had the FP32 math throughput advantage.
2.) Said FP32 throughput was for the most part left untapped and even on modern games you still see it untapped. A navi architecture even if it's GCN would address this in some way and increase FP32 tapped.
3.) 7nm and navi combined means high clocks and higher yields for the same CU count.
4.) We know vega sucks ass at AA and if fixed on navi would stop dropping performance so much when turned on.

My guess for the $250 price tag is:
40 CU @ 1.8 GHz = 9.2 TFLOP FP32 with 80% usage over 50-60% usage over vega + 8GB GDDR6X vRAM

Attached: virgin_2060_vs_chad_Vega56.png (605x821, 29K)

Ultimate retardation. They could've made it 8GBs instead of 16, it's not like any game even uses more than 5GB. Don't talk to me about muh professional shit, no one is going to buy an AMD GPU for anything other than gaming.

Yes, I know, I was pointing out to OP one key reason Navi cards would be much cheaper to produce than Radeon 7, and thus why they could priced more aggressively.

Radeon 7's are salvaged defective workstation cards, which were designed to have four stacks of HBM2. It would have cost AMD more to engineer a way to reduce the memory to 8gb than just leaving the design as is and using the smallest sized HBM2 stacks, which is what they did.

290>390>480>580>590 were nothing but rebrands from AMD with 5%~ performance difference between each card for almost a decade now, what makes you delusional AMDrones believe that this time it will be different in the midrange segment with Navi or even with Arcturus? It won't for at least another 3 years so stop grasping at straws here.

End of summer or later. Might be next year if new consoles get delayed. We could see Raytracing coming to Radeon with Navi so RTX won't be a meme unlike some AMDtards make it out to be.

Attached: 54456.png (261x77, 5K)

And it still a lot more expensive than GDDR5.

I read that they are using 16gb because it uses 4X4gb stacks and if they use 3X4gb then bandwidth will be much less and won't be competitive.

AMD get paid a lot by Sony and MS to build Navi even Vega arch team was amputate for Navi developer, Navi will be mass production chip for Consoles, AMD hits die size around 200-250 mm^2 in 7nm, Arch improved + 7nm Navi 10 > Vega 64.

Thats just not how hmb works, if you cut stacks of it, youre also directly affecting its bandwidth, so 8gb of hmb2 would not also be slow but perform like shit.
Amds goal is not to sell this to the mainstream, but to professionals. They dont guve a flying fuck about people not buying them cause theyre used to it. Theyre already going to get enough money selling it to companies.

Because of VEGA. We know that with 2X MSAA the 56 is able to at least match the gtx 1080 performance in a growing number of titles. If navi can solve the AA problem on VEGA it would become gold.

Attached: 1080p.png (933x2222, 333K)

do you fucktard even know what a rebrand is?
390-480 was at the very least a die Shrink and 580-590 a process refinement.
Thats like claiming Zen+ is a rebrand of Zen 1
The performance gain is a letdown but stop claiming shit you know nothing about
>for at least another 3 years
your speculating there too my boi
No one expects a titan killer, but if navi will be again a mere 5% increase, the new consoles will be dead on arrival

>a die Shrink
node shrink, node shrink

Attached: 3541535434554.jpg (300x168, 6K)

Rebrands in terms of performance. There won't be a miracle that will save Rebrandeon or consumers from the ongoing assrape by Nshitia.

cant believe navi will still be gcn

they dont care about highend.
they supply both Ms and Sony where there's zero competition
>Rebrands in terms of performance
just stop

Attached: 1472569081370.jpg (770x711, 102K)

I think the price points might be legit, but I think it's total bullshit that they'll be anything other than mid-range and low-end cards for HTPCs and basic bitch 1080p gaming.

To be fair the 290 was a great high end part for its time, but rather than push it further with improvements in power efficiency and thermals they just cut down the die size to meet the same specs. Mining takes a lot of blame, but AMD figured out awhile ago that they can make oem tailored parts and just rebrand them as gayming 6 months later and idiots won't notice.

Attached: Make gpu please.jpg (1584x980, 259K)

590 was just released at higher price than 580, the new vega fits in nvidia price/performance. I think this makes nowadays prices fixed for at least another year. If something comes out, it'll be in the same price/performance.

>VEGA suffers from massive performance loss when using any AA more intense than CMAA JUST like polaris
God fucking dammit AMD, get your shit together.

Attached: Dirt_1080p (1).png (1295x1392, 45K)

>72fps min

>To be fair the 290 was a great high end part for its time
no it wasn't, the only thing it really did well was crypto-mining.

For gaming the 290 was pretty meh at launch, it was within 5-10% performance of the GTX 780.

And 8 months later, we got the GTX 970 and GTX 980 which 4+ years later, the RX 590 is barely 10% faster than the GTX 980, and during the same time Nvidia has released the GTX 980Ti, the GTX 1070/1070Ti/1080/1080Ti and the RTX 2060/2070/2080/2080Ti

All of which beat the 590.

There are still driver issues on the 2060 but it shows how an optimized GPU architecture can efficiently utilize most of its FP32 compute. It only has about 6 TFLOPS of it (~same as the RX 580) yet it competes with a 10 TFLOP VEGA 56

Attached: cumblr_mhz8n3EZ531rvyooxo1_500.png (500x405, 115K)

j-just wait for drivers.
wait

>the RX 590 is barely 10% faster than the GTX 980, and during the same time Nvidia has released the GTX 980Ti, the GTX 1070/1070Ti/1080/1080Ti and the RTX 2060/2070/2080/2080Ti
>All of which beat the 590.

Attached: 1479161204343.jpg (480x480, 14K)

GDDR6 is still hugely expensive.

you don't need anything more on 1080pee.

its not $300 expensive, the 2 stacks on vega 64 were $150

>its not $300 expensive, the 2 stacks on vega 64 were $150
Source me the list prices for 2Gbps 4-Hi stacks.

I mean, HBM is expensive. I think between that and not having to care about FP64 compute, prices should drop significantly for Navi, even if $250 is way too low imo.

FP64 costs basically nothing for GCN.

is Navi confirmed to be GCN based?

>The HBM2 memory is probably around $150, and the interposer and packaging should be $25.
gamersnexus.net/guides/3032-vega-56-cost-of-hbm2-and-necessity-to-use-it
vega vii has 4 stacks, $150 x 2 = $300, it may have gotten cheaper but not by much and the interposer is still required.

Well great, I haven't been using a 1080p monitor since ~2011.

yes

GCN 6
so we really can't say too much about how good or bad that is.

Vega is GCN5, so Navi is their first GCN6 cards.

Everything is GCN-based, it's their SIMD ISA.
You can even read the manual on it, it's very comprehensive at that.
The ACTUAL list prices you cockmongler.
You know, the ones vendors publish under seven circles of NDAs.

you obviously cant get those unless you're amd, you fucking retarded autist. point is, its around that price so much more expensive than gddr6.

I WANT AMD to succeed desu but in this case Nvidia is now one step ahead of them with RTX. My guess is they now have full HW async compute. For AMD to compete they have to do something that will unfuck their GCN architecture. Like many anons have stated huge performance loss from AA is one serious problem AMD needs to address.

Attached: Comb16012019101036.jpg (752x1162, 170K)

>its around that price so much more expensive than gddr6.
Show me the actual list prices for both you moron.

WHOOPS, wrong AMD slide. Meant to post the one with MSAA

Attached: 15021640962kesgb4nqh_10_2.png (630x410, 178K)

If Intel gets competitive desktop GPUs and iGPUs at one point will Sony and Macroshit consider Intel for their PS6/Xbawx 3?

>Dx11

What is the point of this bench btw?

No.

Rtx isn't really the problem. In fact for raytracing to really take off amd needs to adopt it as well. The 1000 series hamstrung both companies. Amd couldn't compete, and it was such a jump in performance that nvidia had to pull the raytracing gimmick to make it seem like they weren't having trouble following it.

Why?

Intel is unpleasant to work with, see the OG Xbox.

Surprised rtg hasn't respun big polaris on 7nm
Maybe that's what navi is + tweaks?
Basically the xbox one x gpu on roids would actually be a neat low power midrange gtx 1080 tier gpu

>Maybe that's what navi is + tweaks?
That's a no.

Rtg doesn't have the funding for decent shit hence the gcn repspins since 2011
Blame fucking bulldozer still has knock on effects 9+ years later.
If it didn't suck as much we would have seen faster much better gpus from them but nvidia and Intel royally fucked them for over a decade
290 and 390 where good cards that nobody besides me bought same with Vega 56
Nvidia will always rule the niche gamer desktop market until x86 and (custom builds) finally die when everything goes soc arm soon (not just phablets and crap tops)

Navi is still gcn though as is Arcturus (with dxr rt shit slapped on)
Maybe next gen gcn will unfuck the front end bottlenecks and arch fuck ups like Zen 2 did for Zen 1 and bulldozer

>everything goes soc arm soon
?

>Rtg doesn't have the funding for decent shit hence the gcn repspins since 2011
That they do, at least now.
GCN is their SIMD ISA.
The other IP blocks aren't tied to it.
GCN GPUs wholly inherited their geometry setup from TS3.
>Maybe next gen gcn will unfuck the front end bottlenecks
By that you mean Navi, and probably.
They don't have anything else that's obvious to do.

Amd basically started the soc/apu meme on desktop/laptops pc
Basically a wave of ultra cheap fairly powerful mini pcs are coming from amd and arm
Intel already losing their grip hard

the 2060 does not replace the 1070.
NVIDIA's mainstream XX60 cards now start at $350 when they used to be $250.
You are not fooling anyone here.

>correnction blatant lie
hurr must be an ayyyamPOOO shill

Attached: 666.png (1531x902, 842K)

Amdrones on suicide watch.
>AMD doesn't know what the fuck they are doing
>shills claim radeon vii is a "workstation" GPU, but it's got gimped FP64
>HBM is still just a shitfest that doesn't work unless they stuff a bunch of stacks together
>AMD advetises radeon vii as gaming card while amdrones keep regurgitating the bullshit that it's not
>AMD supposedly losing money on radeon vii so it's rumored they're just going to sell 5000 units or whatever
Oh boy, another year without competition for jewvidia.

>AMD doesn't know what the fuck they are doing
That they do.
>HBM is still just a shitfest that doesn't work unless they stuff a bunch of stacks together
What.

It's funny i see some guy call out that cuck Scotsman and they got -150 downvotes

Are ayyymds really like this?

Attached: ahhhhhhhhhhhh.png (396x408, 165K)

Because it has yet to be disproved that at least some of that info is close to the truth. We just have to wait more.

this is excluding VAT, r-right?

This. Just keep waiting ... they will launch stuff at Computex ... for sure this time ...

*destroys your fanfic with facts AND logic*

Attached: based.jpg (400x400, 64K)

tl;dr me on adored, did underage fags really pay a random on the internet for speculating hardware?

>patreon.com/adoredtv
He gets $2.3k a month from amdrones and then insults people when his "leaks" are wrong.

Attached: 324.png (573x164, 18K)

>its just a rebrand!
>die is literally 1.5x the size

Attached: 22.png (812x794, 71K)

At least put some effort into your bait. Radeon 7 is a gaming version of Radeon Instinct MI50. This is obvious from their spec sheets, and remains factually true no matter which features AMD disables for the gaming version.

It's also not possible to even speculate on the profitability of Radeon 7 without having some idea of the volume of MI50 cards they would be making if Radeon 7 didn't exist, and therefore what percentage of the dies used for Radeon 7 are scavenged dies that would otherwise have not made the cut for MI50 and been thrown out.

even worse, it sucks 1.5x the power for the same performance

I'm sure it's just a coincidence that the Ryzen 3xxx demoed at CES was apparently a Quality Sample of an 8c/16t 65w 3.7ghz/4.5ghz model while Adored's supposed leaks have an R5 3600 at 8c/16t at 3.6ghz/4.4ghz?

And that Adored's supposed leak was right about chiplets in Ryzen when he entire tech press disagreed (and even got Adored to question it)?

And that Adored's supposed leak was right about 12-core and 16-core models, since AMD has confirmed that 2 chiplet models will be made, again, while the entire goddamn tech press was yelling about how 12-core and 16-core models were not happening?

I'm sure these are all just a series of coincidences. I guess we'll see when the full list of specs gets announced.

Im still waiting on the CES announcement .... maybe he meant CES 2020?

I don't know how people think this is impossible or unlikely. It's a fucking node shrink on a brand new architecture. Just look at the defective MI50s with the Radeon 7. The die is considerably smaller than RX Vega yet it outperforms it at just about the same power. They could release mid tier dies like Polaris and definitely get 1080 performance, why the fuck is that something groundbreaking when moving from 14nm to 7nm? In fact, you could expect at most Radeon 7 performance if they play their cards right.

Attached: Everyone's so stupid.png (359x414, 150K)

did you just compared 590 with the 980?

wasnt him that said
>AMD WILL TALK ABOUT THEIR APU'S AND MOBILE PARTS AT THEIR KEYNOTE?
only for amd to release them before even ces started?
and
>AMD WONT TALK AT ALL ABOUT GPU'S AND DESKTOP CPU'S?
we all know what happened here

''AMD doesn't exist to force Nvidia to lower prices''

They do.

No, i compared it to the 290, and by extension the 590 since they're the same shit.

The will because now the can and they will eat all the marketshare they can get.

>$800 card
>only 5000 being made
>release navi after all have been sold
>now NVIDIA is getting BTFO in price-performance ratio

They're making way more than 5k of these.

Attached: 2.png (463x168, 17K)

no they aren't
tweaktown.com/news/64501/amd-radeon-vii-less-5000-available-custom-cards/index.html

Of course, rumors which could be spun as negative for AMD are axiomatically true without the need for any evidence whatsoever, but rumors which could be construed as positive for AMD remain false, even after they are publicly confirmed, right?

|
|>
|
|3
|

he said no announcement, he was right. apologize to tim RIGHT NOW

>tweaktown
Maybe some actual sources this time?