[ 3 / biz / cgl / ck / diy / fa / g / ic / jp / lit / sci / tg / vr / vt ] [ index / top / reports / report a bug ] [ 4plebs / archived.moe / rbt ]

Due to resource constraints, /g/ and /tg/ will no longer be archived or available. Other archivers continue to archive these boards.Become a Patron!

/g/ - Technology

View post   

[ Toggle deleted replies ]
File: 57 KB, 866x522, AMD-Radeon-R9-390X.jpg [View same] [iqdb] [saucenao] [google] [report]
48286552 No.48286552 [Reply] [Original] [archived.moe] [rbt]

so what's the verdict on this? could this be true?


>> No.48286598


>> No.48286604

This is screaming fake so much it's almost palpable.

>> No.48286617

Just wait for the damn cards to be officially announced and you'll have your answer. Seriously what a shitty thread

>> No.48286631

This. Computex starts tomorrow and will probably be announced then.

>> No.48286695

>Cheapest model at 599
>top kek

>> No.48286709

no 380x?

>> No.48286775

285 rebrand

>> No.48286786

pretty sure AMD said E3 at some point.

>> No.48286900

Pretty sure AMD doesn't even know themselves anymore

>> No.48286975


Cheapest model will be a cut down fiji with 4gb of HBM at 600$ and will be as fast or faster than the 650$ 980ti

The un-binned fiji-X model with 4gb of HBM has already beaten the 12gb TitanX in all benchmarks and game runs at 4K and 1600p

The most expensive model will be the unbinned Fiji-X chip with 8gb of HBM memory which will be under 900$ and even faster than the 4gb model that beat the titan in all applications that draw over 4gb

the flagship enthusiast GPU from AMD will not only be the fastest card in the world it will also be cheaper than Nvidia's offerings by at least 150-200$

>> No.48287003

>AMD is just rebranding
>970 is too expensive for me
>960 doesn't worth upgrading from my 760

What should i do ?

>> No.48287032

Join me in sitting in a corner of solitude. guessing we will be waiting for next year for decent competitive cards at all price points

>> No.48287064

so is the 390 really not HBM?

I really want to go to 1440p and freesync since there's good monitors for that out now but if the 390 isn't hbm i'll just get a 980ti and a xb270hu or something

sorry amd but fuck man, you delayed the 300 series for so fucking long for what? one card with HBM that's $850?


R9 290 maybe, it's much better than a 960 and you can get one for $240~

>> No.48287092

Computex has already started because taiwan time

>> No.48287106

>so is the 390 really not HBM?
No fiji gpus /w hbm are getting different branding
kind of like titan

>> No.48287110


all 300 series cards are GCN 1.3 architecture with true audio, Freesync and DX12 support

a "rebrand" is when you take the exact same card and rename it for a new generation series for example: the 280 and 280X are rebrands of the 7950 and 7970 (same arch same lack of feature due to arch)

you need to understand what words mean kid

>> No.48287118

how much are those gonna be?

>> No.48287157


they are enthusiast grade so in the price bracket of the 980ti and titanX

binned to unbinned 4gb to 8gb they will be 600$ to 850$

but they will bet faster than the Nvidia cards thats why nvidia released the titanX early and then followed up with the 980ti undercutting it because they wanted to get their money fast before everyone switched to team red at least on the top end enthusiast level

HBM is a major game changer over GDDR5 and AMD beat them to it

>> No.48287174

>GCN 1.3

The fuck is GCN 1.3 even supposed to be?
Tonga/1.2 was just adding "delta color compression" and 4k hw video codec shit.

As far as I'm concerned, scaling Tonga to 3k+ ALUs is sufficient for a 380X/390/whatever, HBM or not.
I'll likely be upgrading to Pascal/490 next summer anyway, so I don't have any pretenses about the last 28nm gen lasting forever.

>> No.48287194
File: 76 KB, 500x464, 1432824554378.jpg [View same] [iqdb] [saucenao] [google] [report]

I have a 980 right now; i can get another of the same model for $400, only like a couple months old from a guy I know

I want to go to 1440p (getting either the ROG swift or the acer xb270HU)

should I grab it and sli?

Or wait for the HBM AMD cards, sell my 980, and grab one of those and a freesync monitor?

Although the only 1440p freesync 144hz ips that I know of just got recalled and was shit anyway and could only do 30hz-90hz with freesync (that asus one)

>> No.48287232


even if its just modifing the chip and the board to add a new feature set that is a major upgrade

A 370X modified with a True audio controller and DX12 support and DeltaV compression is no longer a "rebrand" of the 280X which was a re-brand of the 7970 its been changed

and frankly a card with the 280X's power and all that shit added on would be really awesome

>> No.48287251


normal i would shill AMD but in your case two 980's in SLI should work nicely with a 144hz 1440p monitor

so i would reccomend that you get a second 980 as cheap as you can along with a G-sync monitor

>> No.48287277


pascal and 400 series will both be die shrinking but on different process nodes

the Samsung 14nm process AMD will be using is superior however so they will come out ahead on the shrink

its gonna be hilarious when the 400 series AMD gpus are cooler and more power efficient than pascal and Nvidia ends up being the house-fire meme again

>> No.48287348


Last time Nvidia made such a huge jump (i.e new node and new memory) it resulted in Fermi.

My personal suspicion is AMD saw the way the wind was blowing way before Nvidia did and this will pay off either with Fury or (more likely) the 400 series.

>> No.48287360

nothing official from amd yet

>> No.48287412

that's the 380 though

>> No.48287444


the game changing shit in on their fury cards which most consumers cant afford

for the mid range people 200-500$ this 300 series wont mean much unless their refresh improves the lower tier card as well

but artic islands is gonna be an unholy fuck you to gaming since by then HBM will have large enough yields (probably use hbm gen 1 on low end card and gen2 on the top tier cards is my guess)

>> No.48287455

>the Samsung 14nm process AMD will be using

About a month ago, Samsung's SEC filing showed that they would also be making chips for NVidia. Looks like both companies are jumping ship away from TSMC.

>> No.48287481

Thank god maybe TSMC will fucking learn instead of sitting with their dick in their hand

>> No.48287513

it seems like everything good that was supposed to happen in 2015 is happening in 2016 instead

>> No.48288139

>Looks like both companies are jumping ship away from TSMC.
Good, they've been fucking shit for the last few years.

>> No.48289980


What is disappointing is that Fiji still probably uses TSMC instead of taking advantage of GloFo's superior 28nm process.

>> No.48289995
File: 955 KB, 360x360, 1423541127484.gif [View same] [iqdb] [saucenao] [google] [report]


>> No.48290019
File: 72 KB, 311x1047, 1432844491796.jpg [View same] [iqdb] [saucenao] [google] [report]

It begins...


>> No.48290114

>R9 380 4gb $199
290 rebrand for $200?
Nvidia is dead.

>> No.48290123

Intel won?

>> No.48290188

380 is a 285 with double the RAM

>> No.48290219

So they kill off 4GB Hawaii entirely? Why would they do that?

>> No.48290221

Is it full DX12 support? Because my GTX 460 has DX12 support...

(also, adding one or two features doesn't make it new, I can't put a new radio in a car and call it a new model)

>> No.48290225

because that already exists

>> No.48290249

4GB Hawaii will be 8GB Hawaii with better power efficiency in 390 and 390x. HBM Fiji cards will have another branding (Fury?)

>> No.48290260

It's all rumors at this point.
AMD didn't release anything yet.
Wait for official announcement.

>> No.48290305

why is amd waiting so long? people like me are gonna get impatient and hesitant and just get a 980 ti. It'd be one thing if from leaks we had a generally good idea of what was coming but we have no fucking clue
no idea if 8gb HBM will be availabe
no idea if they will have an HBM card priced to compete with a 980 ti that will outperform it
no idea if amds drivers will even fucking work for the first 6 months
no idea if it will fully support windows 7-10 with no major bugs for any system

and i have good reasons to doubt all of those will happen

>> No.48290350

Why would you pay the extra cost for 8GB? These cards don't really benefit from 8GB.
4GB Hawaii has to find a place in the new lineup

>> No.48290357

why? just get a 4gb hawaii now. they dont need to rerelease it. it already exists

>> No.48290385

Rebrands should be cheaper than the cards they 'replace'.

>> No.48290395

true. I'd be really surprised if you couldn't get a 290x for $250 by the end of the month though

>> No.48290443


>why is amd waiting so long?

It is better to wait and get it right than rush a release and be slaughtered by the press and other marketing.

People still think the 290x runs at 94c despite every single AIB cooler fixing this.

>> No.48290508

AMD is waiting for E3 so they can announce it along with new PC games that might actually need $600 cards.
Nvidia just rushed the 980ti release to be 2 weeks before that. You can't actually buy the thing yet because no store has any in stock.

>no idea if 8gb HBM will be availabe
pcgameshardware puts 4GB model in June, 8GB Model in August

>> No.48290517

People are in for a big surprise when they find out how much heat and power the 980 ti uses.

>> No.48290571

A few years ago nobody even looked at TDPs and 250W was normal for a high end card.
Why do people care now?

>> No.48290575

yeah, guru3d put the max temp under load at 84c. and as loud as the 7970. I'm gonna drop all my cash on the asus strix version as soon as it's available

>> No.48290675

>Why do people care now?
Nvidia shills and their markering campaigns are out in full force.

>> No.48290686

but right now theyre shitting on nvidia cards, lel

>> No.48290720

But anon, I did. A month ago. Got my 290x Tri-X on Newegg.com on sale for $260

>> No.48290739


>250W TDP
>probably over 300W power consumption
>maxwell very cool and efficient

>> No.48290823

It matters to nvidia shills, that's why i mentioned it
the 980 ti is a titan x, not a upgraded 980

>> No.48290881

Because jewvidya. Remember when **actual** housefires where powered by them? Shills are in such a full force now trying to make the AMD housefires meme to stick, remember some months ago when one of them made a thread about how their 290x or something burnt the mobo and was caught lying and stealing pics from a german or swedish forum where they were depicting it was actually a 670? They don't even have proof of it ever happening so they have to resort to their own hardware faults to kill competition and jack up their prices by 100% when it happens.

>> No.48291100

hitherto exclusively lurking noob here, with an only somewhat related question but you guys sound so cool talking about all these things and i have so much faith in you...
i'm currently running an amd a8-6600k apu and using a sapphire amd radeon hd 7850 for graphics, and i accept that my system is shit.
should i buy the sapphire r9 270x 4gb i found for £150, or is it not worth it?

>> No.48291143

>cheapest is $599
>around $770 in straya
Well fuck me, there goes my plan to buying "new" AMD card. I'll just get the 290x instead.

>> No.48291475

>$770 in straya

Where are you getting that price? The 980ti is $650 in the US and it $1000+ on PCCG and Mwave.

>> No.48291685

The R9 270X is basically a rebranded HD 7870 GHz edition. No, definetly not worth the upgrade for you.

Also the 7850 is fine, your Problem is the CPU.

>> No.48291728


The picture was stolen from overclock.net where someone had bought a used watercooled card that had leaked onto the pcie connector and corroded it.

>> No.48291754

EVGA 980 is $499 non-OC'd, might get

would it even be possible to step up to the 980ti classified when its released?

>> No.48291761
File: 923 KB, 3191x1760, 1433162646240.jpg [View same] [iqdb] [saucenao] [google] [report]

>named after some weeb volcano
>power hungry
>should've been released a year ago

Fermi #3

>> No.48291795

Wait for the lightning.

They call it the classified because it's secretly shit.

>> No.48291852



hell fucking no

My MSI 560ti straight up stopped putting out video after 6 months and the replacement they sent me kept blue screening.

Neither of my friends who have EVGA 780ti and 980 have had any issues.

>> No.48291867

Pls be true. We need you AMD ;__;

>> No.48291888

if Sapphire hasn't released shit, then nothing's official

>> No.48291980

I'm kinda in a similar boat with my GTX 680. It's not worth upgrading if I stay on 1080p, but it'll be worth it for 4k but I just don't feel like these cards will be enough to push 4k / high-ultra / and 60 fps. I simply refuse to play with eye cancer 30 fps
>drop the settings
When I pay 600 bucks for a fucking card, I expect it to run shit on the highest settings.

>> No.48292017

Those guys should really try to grow beards.

>> No.48292027

>implying it won't be the next 8800 GTX
HBM is gonna be gud

>> No.48292186


its going to be a fucking amazing card, but the 4GB meme is going to cripple sales.

Its dead before it even left the womb.

>> No.48292241

>The interesting news is however that in its current state – the Radeon R9 Fury X GPU does not beat the GTX 980 Ti in terms of gaming performance.

Its ogre. AMD is done. Now ill just wait for the hybrid watercooled 980ti.


>> No.48292307


>linking to clickbait
just get the fuck out

>> No.48292327

good lord, those curryniggers probably just spin a wheel to make their next clickbait speculation article.

they've already claimed the 390X/Fury is at least a dozen different things, and they've posted "leaked" benchmarks showing it to beat the Titan X.

Just wait until reputable reviewer get a look at everything.

>> No.48292343

>Check their sources
>"The cards also can't run in their current form, as no BIOS is present."
I'll wait for the benchmarks before trusting clickbait shit.

>> No.48292346

>we have received multiple reports

In other words they have NO source, not even a shitty one.

>> No.48292384

They're hit and miss but they've been right about lots of things since ive been keeping tabs.

If nothing else, the fact they dont have their shit together this late in the game sounds alarming.

Am I supposed to buy this thing end of June and hope for a day one patch?

>> No.48292394
File: 52 KB, 400x400, wccftech.jpg [View same] [iqdb] [saucenao] [google] [report]

>believing hadjitech rumors

>> No.48292403
File: 6 KB, 200x190, peppe.jpg [View same] [iqdb] [saucenao] [google] [report]

>mfw reading comments

What a cancerous community.

>> No.48292426


I'm in pretty much the same situation. I plan on moving to 1440p soon (might wait for the new 144hz Gsync IPS from Asus) but don't know if I should get another 980 or sell it and get a 980Ti. I wouldn't be able to tolerate microstutter and I can't find any credible looking information on whether or not that's still an issue with SLI.

AMD is a no go for me because I use Linux as much as I can and can't be asked to deal with their shitty drivers there. The only reason I'm waiting is to see if their cards force a bigger price drop from nVidia.

>> No.48292586

never read the comments there, it's as if /v/ and /g/ had a retarded kid

>> No.48292602
File: 189 KB, 530x901, 1369706448175.jpg [View same] [iqdb] [saucenao] [google] [report]

I just want to upgrade from my 280x with out buying nvidia shit

>> No.48292613

AMD's Computex conference starts at 10pm EST today, so hopefully we find out what's up then instead of relying on conflicting "leaks".

>> No.48292625

>I wouldn't be able to tolerate microstutter and I can't find any credible looking information on whether or not that's still an issue with SLI.

Its not an issue of SLI, its an issue of xdma. We will never get rid of it until we switch the rendering mode.

Multi GPU is shit, dont do it if you have the choice.

>> No.48292631

Will it be streamed?

>> No.48292642

No, we won't get official information until 16 June.

>> No.48292654

Sorry, an issue of AFR, not xdma.

>> No.48292658

gonna need sauce for that steaming plate of bullshit

>> No.48292661




That was the plan, but they may try to take some wind out of Nvidia's sails since they released the 980Ti

>> No.48292694

The AMD sponsored E3 event with a counter on the Website and Lisa su stating "more information in 4-5 weeks at FAD" (which was ~3 weeks ago) isn't enough?

>> No.48292705

I doubt it, they wouldnt waste so much money for Sponsoring that E3 event if it werent for a release.

>> No.48293074

There's 3-4 tiers of DX12 support.

Nvidia's cards are only partial Dx12. All of AMD's new cards will be fully DX12 compliant with limitless bindings.

>> No.48293462


Ssshh, don't say that - it makes Nvidia fanboys on /g/ very angry.

>> No.48293621

I was about to say, I thought xdma was a proposed solution to that problem.

>> No.48293649

xdma was developed to solve the bandwidth issues caused by hardware bridges

>> No.48293660

Ah, thank you anon.

>> No.48293714

it uses the PCI bus instead of relying on a ribbon which is way more inefficient, and the reason why SLI will always suck compared to crossfire x

>> No.48293718

AMD has no full DX12 support

Only Nvidia has full DX12 Feature Level 12_1 support, the highest feature level tier

But of course, keep lying about it, AMDPOORFAGS

>> No.48293719

Well, frametimes with CF over xdma are really better than with any other multi GPU setup, but you still get microstutter. Its inevitable.

Also AMDs CF support seems to be almost non existant in the last few months.

>> No.48293744

[citation needed]

>> No.48293763

what are you talking about? 15.4 beta resolved issues and generated a XFire profile for GTA5, and the latest Beta was for TW3 and Projects CARS, both now have XFire profiles as well

>> No.48293766

Well, that's not actually true. This device: http://www.amd.com/en-us/products/graphics/workstation/firepro-display-wall/s400#features is actually built to do just that. Completely micro stutter free.

Expensive as hell though, and requires special cards and hardware connectors.

>> No.48293799
File: 70 KB, 310x400, 1421244558336.jpg [View same] [iqdb] [saucenao] [google] [report]

>still believing rumors from geforce forums

you got it backwards again, nvidiafag

>> No.48293800

>Also AMDs CF support seems to be almost non existant in the last few months.
>2 nvidia sponsored games have delayed xfire support
>amd has bad xfire support
i love this meme
who cares about multi-gpu hacks anyway, just get a single card for the same money you would have gotten 2 cards.

>> No.48293858

While I think you're a faggot this guy is right.

As an active CF user who loves his cards, I cant recommend it.

Cough up the money on a single more powerful card. Its worth it.

>> No.48293872


>> No.48293878


Yeah, it's looking like a 980Ti is in my future, unless AMD has a nuclear bomb up their sleeve that's good enough for me to deal with their horrific Linux drivers.

>> No.48293890

>Still believing lies about AMD DX12 support


>> No.48293900

What the hell is the difference between pro and XT? I hope it better not be the water cooling part.

>> No.48293922

>still believing bullshit spouted by brandfags circlejerking on a brand's forum
>rejecting the fact that AMDs cards are DX12 compliant

> http://www.pcworld.com/article/2873545/dont-panic-directx12-wont-require-a-new-graphics-card-after-all.html

>> No.48293934

You don't support 12_1, you're not compliant

>> No.48293941

In project cars crossfire still scales negative. Not that this game is a good example for anything but terrible devs though. CF support is, at the moment, slower than SLI support.

That has absolutely nothing to do with crossfire or gaming Setups.

>> No.48293952

>a single more powerful card
What if it doesn't exist? Titan X and 980 Ti didn't exist when I bought 290X CF. The 980 came out but it was much, much slower. There were no options other than CF or SLI.

I made the mistake of choosing CF, but it was multi-GPU either way.

>> No.48293962

pro is cut down and xt is full

>moving goalposts

pathetic, if you actually read the whole article you would see that Nvidia cards aren't 12.1 compliant either

>> No.48293963

Well, AMDs open source Linux drivers are not far away, probably it'll get better with them.

>> No.48293972

So what difference does HBM make vs GDDR5? What improvements will I see?

>> No.48293993

You're wrong, Nvidia is fully 12_1 compliant

But stay mad, faggot


>12 API with Feature Level 12.1

>> No.48294006

Its not that its really better with SLI. Probably a little bit less shitty at the moment.

If theres no other Option in form of a stronger single GOU card you have to go multi GPU, yeah. Just arguing that its retarded to go multi GPU instead of choosing they available stronger single GPU card.

>> No.48294025

Lower tdp, more space on top for die, smaller PCB, lower wattage, more headroom for OC'ing. And the chip won't be memory starved compared to gddr5. So really, artic islands is a smarter purchase.

>> No.48294037

>nothing to do with Xfire

I know anon. I didnt mean to imply that it did. I only meant to show that it was possible, even though you have to use a fairly exotic set up.

I dont know anon. I'm not calling anyone stupid for buying a certain card, only stating that getting the most powerful single gpu solution is the best option in terms of frame times and "feel".

pcb size being smaller as a result of no ram chips taking up space

reduced heatsink complexity and improved efficency as a result of a single localized heat source.

vram bus obvs.


Only the 980ti and Titan X are FL1 supported. Not even that anon but saying that the company is "fully compliant" is misleading.

>> No.48294042

>posts the brand new card that NOW has 12.1 compliance

you really are retarded, GCN 1.2 is fully DX12 compatible, and the arch has been out since last year, keep grasping at straws you fucking high schooler

>> No.48294063

So theoretically it should OC like a beast?

>> No.48294064

Wrong, the entire Maxwell GM20x GPU family is FL12_1 compliant

They all share the same architecture and all have the features required, ROVs and CR

>> No.48294067

Probably none really. Current graphics cards don't seem to be suffering from some terrible memory bandwidth bottleneck, not yet and not at resolutions people typically use (1080p-1440p or even lower). As GPUs get faster they will need faster memory to feed them of course and GDDR5 would start becoming an issue, but right now cards like the 980 Ti and Titan X pretty much prove that there's no pressing need for HBM. Still, it's good that manufacturers are getting on the HBM train before it becomes an issue.

>> No.48294080

GM20x family supports FL12_1, why you so stupid?

GM200, GM204, GM206 fully supports FL12_1, they all share the same architecture

Are you retarded or something?

>> No.48294085


Please inform yourself friend.

>> No.48294112

The VRAM? Yeah. The chip? Nobody knows, depends if they removed all the compute stuff and implemented better power management.

With HBM you just need a smaller MC and save some power, thats it on the chip site. Massive memory bandwidth and lower latencys will also offer benefits, but that has nothing to do with how high the chip can clock.

>> No.48294133

Nvidia didn't update it for the older GPUs

You can run the query and determine it yourself


>ROVsSupported : 1
>ConservativeRasterizationTier : D3D12_CONSERVATIVE_RASTERIZATION_TIER_1 (1)


You don't even understand simplest thing that the entire GM20x family supports FL12_1

>> No.48294135

Theres still the power saving and better latencys...

Also many HPC tasks are massively bandwidth starved.

>> No.48294152

>the most powerful single gpu solution is the best option in terms of frame times and "feel".
Getting a smooth and even 33.3 frame time isn't better than 16.6 with some spikes. Getting 35FPS on a single GPU isn't smoother than VSync'd or FPS locked 60FPS. If a multi-GPU setup can get 60+ FPS where single-GPU is way below you can pretty much eliminate frame time concerns by locking FPS to 60 or using VSync.

No offense, but have you actually used a multi-GPU setup recently and tried to solve frame time issues or are you just talking out of reviews and graphs you've seen online?

From what I've seen NVIDIA at least makes an attempt to support games when they come out. There have been games that were out for MONTHS before AMD released even preliminary CF support. Witcher 3 has been out for 2 weeks and in development for years, it's the biggest release this year apparently and AMD does not have proper CF support. It works, to a playable degree, but it has lots of clearly visible issues still.

>> No.48294168

Sure, but those won't show in game benchmarks, not really. As for HPC, are HPC cards with HBM coming out right now even?

>> No.48294187
File: 24 KB, 656x341, dx12-amd-nvidia-table.png [View same] [iqdb] [saucenao] [google] [report]


>> No.48294205

Fake and wrong table, Kepler supports Binding Tier 2, even an AMD biased developer says it's wrong

>> No.48294209

I'm actually they put an AIO LC cooler on it.

I'm not sure why people rail against them so much. It gives even the damn R9 295X ocing headroom.

A forum post is not confirmation of support for an entire SDK.

>> No.48294239


>ResourceBindingTier : D3D12_RESOURCE_BINDING_TIER_2 (2)


>ResourceBindingTier : D3D12_RESOURCE_BINDING_TIER_2 (2)

Told you that table is wrong, but believe whatever you want, AMDPOORFAGS

>> No.48294240

No, that won't show in Benchmarks. But if Fiji is a compute monster it would be a huge win for AMD nontheless. They wouldnt have competition, eycept for KNL, until ~late 2016.

We dont know if there will be HPC cards with HBM. But Fiji could be sold as JPC card later this year as soon as large enough HBM stacks become available.

>> No.48294254

>Current graphics cards don't seem to be suffering from some terrible memory bandwidth bottleneck
That's most of the bottleneck
Memory in current computers is awful
I wonder if HBM for APU's is going to work out.

>> No.48294271

Why shouldnt HBM for APUs work out?

>> No.48294286

if its "fake" then link the article that says this, or post the correct one faggot, otherwise you're responses paint you as a butt mad little kid getting pissed that his brand gets shit on

>> No.48294293

>slower than the GTX 980Ti


>> No.48294294

I think you need to take a look at this.


>> No.48294307

Because it would make them incredibly expensive.
It might work out if they make some kind of unified memory for both CPU and GPU and just not include DDR3/4 slots.

>> No.48294313

>want to finally upgrade from potato tier
>would be happy with 1080p
>new cards coming out "SOON :')"
>just wanna play games on at 60fps without my system spewing flames

>> No.48294346
File: 144 KB, 960x869, Muh_baby.jpg [View same] [iqdb] [saucenao] [google] [report]

>have you actually used a multi-GPU setup

I'd say I have.

>> No.48294350

Table is WRONG period, it says Tier 1 but when you actually query the hardware, it's Tier 2 for Kepler & Maxwell 1

But believe whatever lies you want to believe, that table is WRONG and thats it

>> No.48294360

>That's most of the bottleneck
That doesn't seem to be supported by practical evidence. 256b 970s and 980s perform just as well if not better than 512b 290X's. 384b Titan X's and 980 Ti's completely slaughter 290X's simply because of the much, much faster GPU despite the lower memory bandwidth. OCing GPU memory hardly has any impact at all on FPS in games, where increasing core clock does. Caching goes a long way.

>> No.48294366

[citation needed]

>> No.48294369

It should be cheaper than Intels ESRAM L4$. If you can sell the product has middle/high end class the additional cost for HBM and Interposer shouldnt be a Problem. Also the 512/1024b MC youd need for the HBM is ridiculously small.

>> No.48294371

>shows actual fucking SDK from the developers

>> No.48294381

So then, you're saying that games where you'd get a solid 70+ FPS aren't buttery smooth if you lock them to 60, or at least much better?

>> No.48294397
File: 17 KB, 637x486, baby-crying-2.jpg [View same] [iqdb] [saucenao] [google] [report]

>its wrong because I say it is!!!!!
> I have nothing to back up my claims but their wrong!!1!!1!

since your English is shit I assume its your second language and you probably look like pic related you fucking babby

>> No.48294398

No man, I think you misunderstood me. Games capped at 60 feel great, but still not as good as a single card.

That's all I was trying to say.

>> No.48294399

Hawaiis 512b MC is clocked way lower than Maxwells(or Keplers) 256/386b MCs.

Together with DCC GM 204 has about the same effektive bandwidth for gaming as Hawaii, GM 200 even more.

>> No.48294419

I already shown the table is wrong,


Not my problem you AMDPOORFAGS are not educated enough and don't know how to read at all and understand simple English

>> No.48294448

dude, post a reputable source. you haven't shown anything as being wrong, you're simpyl claiming the table is wrong

support your argument with facts or shut up

>> No.48294449

where did you show the table was wrong? oh that's right, you didn't, were just supposed to believe the bullshit FUD you're trying to spread

>> No.48294452

Are you talking about the right table? I linked directly to the Microsoft SDK. Which is not wrong.

How much did they pay you to take a Rosettastone English course and shitpost here?

>> No.48294479

But it's better than a single card at 35FPS or 40FPS. That's my point.

It's not as good as a single card at 60, but if that 'single card' is a mythical beast that does not exist you don't have any better options.

290X memory easily goes over 6GHz. As a 290X CF owner I can tell you that OCing memory from 5GHz to 6.2GHz doesn't do shit to in-game performance. You can measure it if you really want to, but you sure as fuck won't actually notice. Meanwhile throwing another 100-300MHz on the GPU is much more noticeable.

>> No.48294481 [DELETED] 

>>48294452 table is WRONG

When will dumb fags even learn?

>> No.48294483

you should get your money back for your shitty English lessons

>> No.48294486

>>48294187 table is WRONG

When will dumb fags even learn?

>> No.48294497

Like I said, I agree. What I said was "I'm not calling anyone stupid for buying a certain card, only stating that getting the most powerful single gpu solution is the best option in terms of frame times and "feel"."

I didnt mention frame rates and assumed it was <60

>> No.48294502

Maxwell GM20x GPU is the most advanced and compliant DX12 GPU with Feature Level 12_1 support

Anything else does not support FL12_1, period

>> No.48294507

I know that Hawaiis memory can be OCd fairly well (unless you got shitty elpida memory), but that wasn't my point. I just pointed out that Maxwell doesn't come with less (effective) bandwidth.

>> No.48294509

Yeah if it's <60 it's pretty shit. I hope nobody buys multi-GPU setups to play <60.

>> No.48294514


kek, post link since you insist this is the truth

>> No.48294519
File: 1.02 MB, 810x7800, 1426754955116.png [View same] [iqdb] [saucenao] [google] [report]


The Nvidia damage control has gotten out of hand recently.

Then again the nvidia fans think the 980ti was released out of the goodness of nvidia's heart rather than the fact nvidia are worried about losing the performance crown (being able to say you are the fastest is a powerful marketing tool).

See also: AMD claiming the 295x2 is the fastest card on the planet. This isn't a false claim either but does overlook the dual gpu nature of it.

>> No.48294527

That's fine, but my point is that current cards, both Maxwell and Hawaii do not show signs of being heavily bottlenecked by memory bandwidth.

I mean, I increase b/w by over 20% and shit all happens, clearly the GPU isn't starving for that shit.

>> No.48294541

It also overlooks the fact that this year they had proper CF support on release for like 2 games (Hardline and GTA V).

>> No.48294549

Nice backpeddling.

How's that google translated English workin out for ya?

You're trying so hard lol. Its pretty impressive.

I understand though, just trying not to loose your job.

yea. agreed.

What's happened? I used to enjoy the friendly amd/nvidia shitposting and even participated. But this shit is another level.

>> No.48294570


>>48294239 was already posted, this was queried straight from Kepler and Maxwell1 card with the Win10 drivers

Kepler and Maxwell1 supports Resource Binding Tier 2 yet the table says Tier 1, which is WRONG and clearly fake

Not my fault a moron like you can't understand simple English

>> No.48294577

Thats, to a certain degree, also caused by the bandwidth these GPUs are designed around. Just look at maxwell v2 for example. Fuck huge L2$ to hide latencys. With a higher bandwidth memory system you can spare some serious die area here.

We Start to hit the limits of GDDR5, more effective bandwidth than with GM200 you simply cant squeeze it out with reasonable expense.

>> No.48294584


>What's happened?

Well it has been slowly escaliting with the 290x slowly gaining on a 980 in recent drivers, but the real catalyst is AMD saying the MOTHERFUCKING FURY is aimed at being a premium brand like the titan series.

Nvidia fans really, really have been getting salty over the whole 3.5 thing and the fact the 980 is generally seen as overpriced. Now nvidia has just pissed off all the titan x owners with the 980ti and the performance crown in geopardy the fanboys are starting shitpost as quite a few reasons to say "muh nvidia" are being eroded.

>> No.48294590

The table we were talking about was this one: https://msdn.microsoft.com/en-us/library/windows/desktop/dn899127%28v=vs.85%29.aspx#invariable_limits

Not that shit you keep pointing too. Why are you trying so hard to ignore this?

>> No.48294604

Yeah, that's what I'm saying. With current GPUs it's not really an issue (obviously, the GPUs were designed to work with GDDR5), but it will become one in the future, so it's great that we're moving to HBM before it's a significant bottleneck.

>> No.48294612

Can wait to see the shit storm that brews once the original shitwrecker finishes working on Zen.

Gonna be fun times.

>> No.48294618

>Why are you trying so hard to ignore this?

Whats that for a question? Did you expect reason from someone using terms like amdpoorfags?

Seriously, you should get insta-banned for amdpoorfags, nvidiots and anything like that.

>> No.48294624
File: 777 KB, 1280x720, Certfied shit-wrecker, world saver, anal annihilator.png [View same] [iqdb] [saucenao] [google] [report]


>> No.48294625

>-R9 380 4gb $199
>-R9 390 8gb $299
I would be very okay with this. Thought about buying a r9 290, but if these prices are true, i would buy one of those.

>> No.48294655

trying to shill novidya so hard

>> No.48294664

I'm normally an amd shill as well but I dont really see a problem with that suggestion.

Its bad performance/$ but if he has the money it's his choice.

>> No.48294694

if the 380X is going to be $250, I might just have to buy one

>> No.48294703

I guess it's not gonna happen but I'd like it if AMD released a GPU more powerful than the Titan X at a lower price.

If only because I feel bad for them and would like seeing them get good press.

>> No.48294720

If a mere 285 is going to be the new 380, I'm gonna be mad.

Fuck that.

>> No.48294757

it will be a more efficient 285 with 4GB of VRAM, it will perform better than the 285, what I'm wondering is how big will the shader count be on the 380X and if they'll offer more than 4GB VRAM on it, if it does and it like oc'ing, it will be a monster and destroy the 970

>> No.48294767
File: 25 KB, 650x425, Jim-Keller.jpg [View same] [iqdb] [saucenao] [google] [report]

I just wanna shake this mans hand, look him in the eye and ask him what it means to be a god.

>> No.48294779

When are the 300x series coming out?

>> No.48294805


>> No.48294823

I hope he wrecks all the shit. Only halfway there to 2016.

>> No.48294852

Possibly later this month, we'll have to wait for AMD's E3 conference to be sure.

>> No.48294872

Well, only a god can kill a Titan

>> No.48294891
File: 721 KB, 249x172, awe.gif [View same] [iqdb] [saucenao] [google] [report]


>> No.48294897
File: 104 KB, 645x803, lisa-su-amd-ceo-645x803.jpg [View same] [iqdb] [saucenao] [google] [report]

herro we have new glaffic cald soon, prease wait tir we show dem befow shitpost and guessing

xie xie

>> No.48294928
File: 127 KB, 799x1198, JHH_Feb1.2008_preview.jpg [View same] [iqdb] [saucenao] [google] [report]

ah yers, grad to see yrou truday

>> No.48294933

Since /g/ hates Samsung with a passion will they turn their back against AMD as well if Samsung ends up buying them?

>> No.48294946

If Samsung buys AMD I'll consider it AMD selling out.

I've supported AMD for so long having them get purchased by Samsung would just be a slap in the face.

>> No.48294961

>If Samsung buys AMD I'll consider it AMD selling out.
Are you stupid?

>> No.48294972

that's a pretty big jump, if they pair it with >4GB it will crush the 970

>> No.48294982

But Samsung are the good guys. They are the only company who has the resources to fight against Apple and their disgusting Apple-ification propaganda.

>> No.48294990


>> No.48295005

They're fab partners... What the fuck are you on about?

>> No.48295129

Doesn't matter. My 660 ti still works.

>> No.48295154

Nice blog post.

>> No.48295169

4chan is full of shit posts so who cares if I blog.

>> No.48296952


>This slide originates from a Brazilian forum and it's wrong on several quite important counts, for example: none of GCN chips support rasterizer ordered views, conservative rasterization, or tiled resources tier 3 (ie. "volume tiled resources" with Texture3D support), as shown by the D3D12 feature checking tool. GCN 1.1/1.2 cards support tiled resourced tier 2 (with Texture2D support) and feature level 12_0, GCN 1.0 "only" supports tiled resources tier 1 and feature level 11_1. Double precision floats are supported across the entire GCN range.

>Kepler and Maxwell-1 do support resource binding tier 2, and do not support typed UAV load for additional texture formats - this leaves them at feature level 11_0. Maxwell-2 is currently the only card which supports feature level 12_1.

>AMDPOORFAGS here believe HUEHUEHUE lies

>> No.48297150

>nvidiashills reposting nvidiashills

Fuck off.

>> No.48298644

>source some HUEHUEfag forum
>which in turn points to some Hungarian forum

Yeah that's not a fucking source you imbecile.

>> No.48299314

Went dual 7950s one time, it was fucking awful.
I'm so glad I managed to profit by selling my cards on fleabay when buttcoins were popular
Glad I have my Titan X now.
AMD never again

>> No.48299708
File: 85 KB, 430x480, 11292776_715509701905836_384015936_n.jpg [View same] [iqdb] [saucenao] [google] [report]



>> No.48299737

Yikes that is....incredibly racist. Could you delete this post?

>> No.48299748

Stop posting that clickbait currynigger site.

>> No.48299848

>have the only pic on the gpu
>don't give a fuck about the the confidentiality/embargo/delays

>clickbait currynigger site

>> No.48299984


>> No.48300027


Are you a fucking retard? That picture is two weeks old, a Dev over at DICE posted it on his twitter.

>> No.48300177

No, the already stated 4gb hbm, and they showed what it looks like which is not OPs pic.

>> No.48300239

sauce pls

>> No.48300285

>pic of the card = pic of the chip with the pcb

>> No.48301104


Well, Sapphire is the one manufacturing all amd reference shit, so yeah, if there was a real leak it wouldn't come from Asus or Currynigger leak sites...

>> No.48301232

speaking of amd gpus, should i buy a 290x now before the 390 comes out? will it go up or down in price?

>> No.48301442


Just wait, perhaps the 390x will be like the 290x of the lineup, a really beefy Tonga, with 8gb of GDDR5, and 380 a 290 rebrand, but we just can't fucking know before AMD actuall SAYS ANYTHING...


>> No.48301515

i have a full build MINUS GPU ready to go. im so close to just buying a 290x... fuck this shit why are gaymd so fucking slow I hope nvidia shits on them at this point

>> No.48301571


>why are amd so slow

Oh geez, it's not like when they released the whole 200 series it was regarded as a housefire that needs a nuclear reactor, while still wrecking shit TWO FUCKING YEARS LATER, with their 295x2 still being the strongest card on the GPU market, and giving in to the 900 series even more marketshare, i'd rather them take their time to perfect it, and release a banger that will make all the 980's go to ebay and get amd back massive markeshare, than force it out, have shit marketing, and bleed money.

And let's face it, I have a 290x, and most of the time i just browse reddit, watch youtube and netflix and play games maybe one hour a day or some shit, and you can do that on your IGP

>> No.48303698 [DELETED] 

So in theory 4GB HBM is worth how much in DDR5 GB in performance?

>> No.48303953

it isn't really as the older 280x had the same,however with tonga architecture it will actually perform way more that the numbers say.

I highly doubt you'll see 4GB 380x, 2GB definitely,seeing from pass release of hawaii pro to hawaii Xt( 290(x)) the RAM spec didn't change at all.

>> No.48303996

also with the 280 and 280x,same vram spec

>> No.48305309

wew lad

you sound like an utter faggot

>> No.48305327

reading your posts i always imagine a pig squealing

>> No.48305522

That console fps on TW3 must be nice with your 290x

>> No.48305847
File: 524 KB, 429x501, tfw no fiji.png [View same] [iqdb] [saucenao] [google] [report]

>tfw no fiji

>> No.48305889

Not him but I have a 290 and it runs about 50-60 on Ultra with hairworks on.

>> No.48305976


Strictly speaking AMD users get a better experience with hairworks on as CCC lets you limit the amount of tessellation (hariworks uses x64 which is stupid no matter the gpu) whereas Nvidia does not support this option.

NOTHING needs to use x64 tessellation, x16 is pretty much the limit of human eyesight.

>> No.48306491

There's simply no way Hawaii with 40% more shaders at the same clocks and HBM can be slower than GM210 without some really major fuckups.

Unless, we've been fed the wrong spec info and Fiji doesn't have 4096 shaders, but much less.
Heck, it might not even be Hawaii + 40% like it was leaked, it might just be fully scaled sub 600mm^2 Tonga, though that's unlikely.

>> No.48306501

i was thinking the same

>> No.48306575


Stock watercooling might suggest very high base clocks.

>> No.48306593

Unlikely, the PCB apparently only has 6 phases for the GPU, dismissing absurdly high clocks.

What the watercooling is for is to make the card small, there's simply not enough room on that small card to fit a huge heatsink and keep it cool.

>> No.48307392

Saying that 7950s are a shit experience compared to a titan X is a waste of breath.

>> No.48307692

I'm excited

>> No.48308062


Name (leave empty)
Comment (leave empty)
Password [?]Password used for file deletion.