http://www.anandtech.com/show/10469 /amds-tuesday-radeon-rx-

>1: Based on data running ’Total War: Warhammer’, ultra settings, 1080p resolution. Radeon Software 16.6.2 74.2FPS vs Radeon Software 16.7.1 78.3FPS; Metro Last Light, very high settings, 1080p resolution, 80.9FPS vs 82.7 FPS. Witcher 3, Ultra settings, 1440p, 31.5FPS vs 32.5, Far Cry 4, ultra settings, 1440p, 54.65FPS vs 56.38FPS, 3DMark11 Extreme, 22.8 vs 23.7
>gimping
are you a fucking retard?

>shift some of the power load off of the PCIe Graphics (PEG) slot connector, presumably in order to bring power consumption within PCIe spec. Note that AMD doesn't say anything about reducing the total power consumption, and given option #2, it's reasonable to assume that this involves holding the power requirements as-is and shifting the load to the external 6-pin power connector. Based on earlier data this would potentially put the 6-in connector further over spec, but the vast majority of PSUs are very tolerant of this going out of spec.


How is this gimping exactly?

>>55437820
/thread

>>55437820
DELETE THIS

>>55437820
Also not crippled compute like Nvidiots. RX 480 24 h/s, GTX 1080 21 h/s. Muh ethereum mining.

>>55437820
Source? And does this mean it actually increases performance, or that it decreases by about 1 FPS?

>>55437861
lrn2read

>>55437861
Footnotes in OP article, citing the new driver's optimizations for a few titles. It's actually separate from the power consumption fix, which is going to come in two options, either offloading excess power draw to the much more capable 6-pin connector, or reducing total power draw so that both the PCI-E slot and 6-pin stay within spec.

>>55437861
Slightly dropping the voltage from just the peak boost state alone lets the card hold its clock longer so it performs better.
That isn't addressing the card pulling more power over the slot, but it is slightly lowering total power.

There are a lot of interesting things going on with Polaris. Not bad mind you, just software things to be ironed out. Programmable VRM controller. the die self analyzes and makes compensations for its perceived average transistor quality, adjustable power targets and setting certain power consumption ratios for the die and memory separately. 4th gen GCN has more new features than any other prior gen.

File: 2ecafb820f.jpg (226KB, 1098x1484px) Image search: [Google]

226KB, 1098x1484px

>>55437870
>>55437888
So it does increase performance
Nice

>>55437888
So they are giving you 3 options, how nice:
A, Gimp your card
B, overload you PCIe power bus
C, Overload the 6 pin power bus

AMD Shills will actually defend this.

>>55437925
B sounds pretty good desu

>>55437925
>>55437940
I mean C,
C!

>>55437832
your shitty chinese psu that you bough doing cheap shitty build with amd parts will fry your whole PC, Enjoy from AMD aka "Another Major Disappointment"

AMD is so incompetent, run by a bunch of idiots. Can't believe they fucked up so bad with the 480

>>55437953

A shitty psu will nuke everything regardless of Nvidia or AMD - they are rarely capable of delivering what the box says the unit can.

>>55437958

Fucked up so bad the card is selling like hotcakes.

>>55437820
>>55437789
>LESS THAN ONE WEEK AFTER LAUNCH
>ALREADY 3% BETTER

FUCKING
B A S E D
A
S
E
D

>>55437789
>AMD KEEPS KEKING NVIDIA
>3% PERFORMANCE INCREASE A WEEK AFTER LAUNCH
>NO END IN SIGHT
>NVIDIA KEKS ON SUICIDE WATCH

How can you be so stupid to mess up something like basic graphics card power?

It's pretty fucking simple, the card should only draw power according to the specs, but AMD was so fucking stupid they missed this basic fact.

This fix will inevitably cause performance issues and still overdraw from the 6-pin.

File: 1467721874829.jpg (275KB, 2067x1602px) Image search: [Google]

275KB, 2067x1602px

>amd is unable to gimp
>they attempt and get a 3% in performance
sorry nvidiots, this was unavoidable.

AYYMD GIMPING PERFORMANCE BY REDUCING POWER DRAW TO BE BACK IN COMPLIANCE

AYYMDPOORFAGS CONFIRMED ON SUICIDE WATCH

>>55438033
>AYYMD GIMPING PERFORMANCE BY REDUCING POWER DRAW TO BE BACK IN COMPLIANCE
>reduce power draw
>immediately increase performance at the same time
TOPKEKKERINO

File: 1467323624755.jpg (25KB, 377x364px) Image search: [Google]

25KB, 377x364px

>>55437940
>>55437947
Too late! Kiss your shitty Chinese prebuilt motherboard goodbye!

File: 1372498138483.jpg (33KB, 270x237px) Image search: [Google]

33KB, 270x237px

>>55437925
>overload

File: 290x vs 780ti.jpg (736KB, 521x6071px) Image search: [Google]

736KB, 521x6071px

>>55438023

Its fucking hard being an AMD user. I bought a 290x in early 2014 and expected it to fade into obscurity but god fucknig damnit now it competes with a 980. AMD can't do anything properly.

>>55438066

Stay snug, smug.

File: 1460512680972.png (27KB, 181x220px) Image search: [Google]

27KB, 181x220px

>>55438061
>Nvidiot doesn't know what MSRP means

>>55437854
This is why AMD is losing. They don't gimp their cards for mining which means miners buy their cards and gamers don't, because miners buy all of them up. Forever 20% marketshare.

File: poo-in-the-gpu.jpg (293KB, 960x655px) Image search: [Google]

293KB, 960x655px

New driver with 200% more poo.

>>55438066
>should've bought a nchinkia fucboi edition instead

>>55438118
That's hilarious, I always wondered who actually buys these pieces of shit.

>>55438061
>Novidia in charge of protecting GPUs from scalpers

http://www.ebay.com/sch/i.html?clk_rvr_id=1057609652442&mfe=search&LH_BIN=1&_nkw=gtx%201070&LH_Complete=1&LH_Sold=1&rt=nc&_trksid=p2045573.m1684
http://www.ebay.com/sch/i.html?clk_rvr_id=1057607171349&mfe=search&_nkw=gtx%201080&LH_Complete=1&LH_Sold=1&rt=nc&_trksid=p2045573.m1684

>>55437789
So dumb question here, but can anybody actually explain to me WHY this is a thing. Not why AMD fucked up, but why a card that uses like half the power of the previous generation can't find a place to get the juice? Did they f up by using a 6 pin? Could you just buy some adapter that would spread that draw over 2 connections?

>>55438183
It takes too much power from the motherboard. It's basically an engineering mistake, it can't be fixed with any kind of adapter.

>>55437789

AMD Radeon RX 480 in CrossFire - Beating the GeForce GTX 1080 at 4K

http://www.tweaktown.com/articles/7770/amd-radeon-rx-480-crossfire-beating-geforce-gtx-1080-4k/index.html

BTFO NVIDIA!

>>55438179
>>55438183
the only difference between 8 pin and 6 pin is 2 grounds.

6 pins can output up to 250W, so long as the grounds are specced well.

>>55438183
It has nothing to do with the power connector, its internal settings in the GPU with regards to how it regulates TDP, and power draw between the die and the memory. The card can make a whole host of changes to its own settings based on a ton of criteria. Its just a software issue.

>>55438199
>I have no idea what I'm talking about
The issue is only affecting some cards, not all of them.

>>55438222
>The issue is only affecting some cards, not all of them.

Nope, it affects all 480s

>>55438093
so basically you need to wait 3years and 2 generations till you see your card actually outperform similarly priced competition?

It's pretty simple, the card is drawing too much power from the motherboard instead of the 6-pin adapter.

How AMD could have made such a simple and stupid mistake is beyond me.

>>55437789
So not only did AMD give out 4gb vram free they are doing a 3% performance increase a week after launch.
How can GimpVidia even fucking compete.

>>55438274
1060 will cost 180$(msrp for 6gb model), it will also have 20% better performance than RX 480. Nvidia is also supposidly created new async optimization drivers which improve all its card performance by a lot.

File: 6a2.png (47KB, 125x127px) Image search: [Google]

47KB, 125x127px

>>55438296
>3gb 1060
>3gb in 2000 and fucking 16

File: 1467483739528.gif (2MB, 332x215px) Image search: [Google]

2MB, 332x215px

>>55438296
>1060 will cost 180$ MSRP
As much as I would like that to be real, you're fucking delusional.

>>55438306
Nvidia doesnt know how to make a true 4 GB card. The best they can manage is 3.5

>>55438244
>I have 7970GE
>Vega looks like my jam

Sounds about right.

>>55438317
They know how to make faster and better performing cards though unlike AMD

File: 2ec.png (572KB, 600x580px) Image search: [Google]

572KB, 600x580px

>>55438330
Until they gimp them 6 months after launch

>>55438313
Yeah 1060 will be $249, almost guaranteed

Still a much better deal than the 480

>>55438343
Yet they still remain faster than AMD cards.

File: 1467474360169.png (42KB, 653x726px) Image search: [Google]

42KB, 653x726px

>>55438366
DELETE THIS

>>55437820
>Performance figures are not average, may vary from run-to-run.
>Not quoting last line
>Taking ayymd advertisements as facts

DESPERATE

:^)

>>55438313
>>55438354
There is good chance it will be cheaper than RX 480 you dumb amdrones.
Why? Because it doesnt have SLI support. It will be calculated move by nvidia to take control of mid end market, a card which will have extremely good value, while still not exposing its highend market.

File: total warhammer dx12 1080p.jpg (80KB, 548x376px) Image search: [Google]

80KB, 548x376px

>>55438244

The 290x was competitive upon release against the Nvidia cards of the time (primarily the 780 and original titan) - the 780ti was released to be faster than the 290x (and it was).

Now all this time later the 780ti is dying a painful death and the 290x is fighting against a 980. Hell depending on clocks a 290x/390x even competes against a (stock) 980ti in some DX2 games. That is effectively 3 tiers of performance improvement.

Your argument that AMD cards (in this case, the 290x - though the 7970 also counts) aren't competitive upon release is flat out wrong.

>>55438405
If it's $200, it will absolutely destroy the 480, but no way it's going to happen.

If it's $250 it will still destroy the 480, and this is likely.

Either way Nvidia wins.

>>55438432
This whole post is completely wrong

When will the custom 480s start coming out?

>>55438432
>hurdurrdur amd stronk async compute, hurtdurdur

>>55438442

Elaborate.

>>55438183

>>55438199
>>55438219
>>55438222

Ok, I should have elaborated. I've actually READ op's article, where it tells us that the update will let people essentially choose whether to funnel the excessive power draw away from the PCIE and into the 6 pin, or just reduce power consumption altogether. Forcing it through the 6 pin is supposed to go over the official limit, but still be ok for most power supplies. I'm wondering if there's some way to connect that 6pin to a power supply that would split the power draw over 2 connections or something, like to prevent blowing out a cheap PSU

>>55438461
You know he mentioned the 2xx series, right?

>>55438536
>I'm wondering if there's some way to connect that 6pin to a power supply that would split the power draw over 2 connections or something, like to prevent blowing out a cheap PSU

It could definitely have problems with a cheap PSU. I don't think there's any way around that.

According to specs it should only be drawing 75W from the 6 pin, and it always draws more. This new driver just makes it even worse on the 6 pin actually.

>>55438597
>look at me being a tech illiterate retard

6pins deliver the exact same amount of power as 8pins.
Stop regurgitating bullshit about "specs" that you don't understand.

>>55438241
No it doesn't stop bullshitting

>>55438604
Sorry that's wrong

>>55438629
They each use the exact same 3 12v lines. An 8 pin connector only has an additional ground and a useless sensing wire.

They can deliver the exact same amount of power.

>>55437925
6 pin pcie's cana ctually draw 150 watts safely, they just arent rated for it. The extra two in an 8 pin are grounds basicly.
Figuring 10 amps at 12 volts, 120 watts coming through the 12v rail shouldnt be an issue actually given you dont have a shit psu.

So it really isnt damaging your psu, as long as the 12 rail has enough current to supply the gpu and the wire isnt 20 gauge trash chinese shit.

File: fucku.jpg (45KB, 625x499px) Image search: [Google]

45KB, 625x499px

>>55437789

The word you are trying to use is "improving", not gimping.

GTX 1060 days ago was 8% faster than 480 according to leaks. Not is only 5%. By the time is released will be 2%. In one month from them both will be equal.

In a year, GTX980ti == RX480 in directx12 games

>>55438607
Yes it does

>>55438642
You are a perfect example of what an idiot does with a little information.

>>55438655
1060 is 15% faster than 480

It's also the same price.

Kinda a no brainer to get the 1060 at this point

>>55438685
>shit eating tech illiterate retard shill desperately trying to save face so he can keep spreading fear mongering FUD

>>55438689
source?

Does AMD still have shit Linux support or did it get a bit better ?

>>55438655
in 10 years, rx 480 will finally reach its final form and defeat gtx1080ti's
such is pajeetious development cycles

File: killer instinct 1080p.jpg (134KB, 548x629px) Image search: [Google]

134KB, 548x629px

>>55438655

>GTX980ti == RX480 in directx12 games

The nearly 3 year old 290x can do that depending on base clocks.

AYYMD doesn't even support Feature Level 12_1, they're not DX12 cards at all

>>55438734
Shit support even though they're slowly going open source, it'll be another year before they catch up.

>>55438741
Yup, AMD cards are shit, riddled with issues and always cutting corners

>>55438747
I was hoping that vulkan would change this somehow. Kinda sad.

>>55438741
>>55438753


You do realise for some DX12 feature Intel has greater support than both AMD and Nvidia right?

http://www.extremetech.com/extreme/207598-demystifying-directx-12-support-what-amd-intel-and-nvidia-do-and-dont-deliver

>>55438747
- No linux support
- Horrible drivers
- Power issues that will kill your motherboard
- Poor performance
- Inflated prices

What a disaster the 480 has turned out to be

>>55438405
The 6gb 1060 will have SLI, the 3GB (if there is a 3GB and not called the 1050TI) doesn't have SLI.

There is a reason there are two different core versions the /300 and /400

>>55437925
The technical max power handling of four 18-guage wires over a 400-500mm distance is around 10 amps. The contacts are rated for 9 amps. 12V * 9A * 4 = ???

An extra 20 watts won't do anything to a 6pin PCIe connector, the "official spec" numbers are arbitrary. The extra pins in an 8pin connector are ground leads. Both the 6 and 8 pin are physically capable of carrying the same amount of power.

>>55438760
Vulkan exists but it still hasn't gotten off the ground, hence the one year more part.

>>55438806
The only good thing about this is that prices of the r9 300s are going down because they're moving old stock. Good time to buy.

>>55438179
kek, saved

File: 1459956414817.png (157KB, 4500x4334px) Image search: [Google]

157KB, 4500x4334px

>>55438391
>tfw someone saved it

disregarding standards for marketing purposes (look, our card only needs 6-pin connectors because efficiency, next-gen) and failing badly at that is called AMD.

File: dx12.png (67KB, 567x635px) Image search: [Google]

67KB, 567x635px

Muh AMD wins everytime in dx12

>>55437820
Don't worry it's just typical paid nvidia shills

>>55439039
Pretty sad how even an overclocked 480 can't beat a 970

>>55439069
The tiny 480 heatsink makes me laugh everytime

File: 1467480257828.jpg (240KB, 1356x916px) Image search: [Google]

240KB, 1356x916px

>>55439039
>>55439060
It's amazing how DX12 games work better on AMD than Nvidia. Except Rise of the Tomb Raider...Oh Yeaaaa it's a fucking gameworks gameeeeeee.

File: 1467804750.png (82KB, 662x675px) Image search: [Google]

82KB, 662x675px

>>55439039
>amd still sucks at cpu scaling
why would anyone from a poor country buy a card that won't work well with a cheap cpu?

>>55439039
Of course, posting the broken dx12 benchmark in Tomb raider.

>>55439095
But Nvidia cant do dx12 games? hows this work then?

File: XxhXDJKP2x6Z9jPMP7XgjD-650-80-1.png (153KB, 650x388px) Image search: [Google]

153KB, 650x388px

>>55439097

Oh lawdy look at the fury x keeping up with a 1080.

>>55439105

900 series and older cards take a performance hit in DX12 due to lol Nviudia. The 1000 series cards tend to see tiny gains, but no losses compared to DX11.

>>55439069
water boils at 100 degrees celsius (for ameritards). just shy of 4 degrees.
>amd is living on the edge, adventure :^)

File: 1.png (2MB, 1673x945px) Image search: [Google]

2MB, 1673x945px

>>55439101

>>55439126

Stock GTX 480's would hit over 100c - there is a reason why the housefire meme belongs to Nvidia.

>>55439130
amd thought it's a cool thing to join the housefire bandwagon :^)

>>55439095
I thought a big selling point on these cards was the low power draw and efficiency.
Did AMD lie or did they just make one monumental fuck up?

>>55437789
paid shill starts shilling

dont even read his own link
way to go

File: 14610034644x5teWArBn_4_5[1].png (31KB, 623x446px) Image search: [Google]

31KB, 623x446px

>>55439117
i wasn't talking about dx12 (forgot to add that), i was talking about dx9-11, on dx12 cpu scaling is fixed but there aren't many dx12 games out there

http://www.tomshardware.co.uk/nvidia-vive-displayport-incompatible,news-53397.html

>VR ready
>actually, it doesn't work

>>55438689

Actually 1060 is 150% faster.
Source, my ass.

Told be truth, in numbers, the 1060 is a piece of shit. It will be +5% over 480 in old nvidia titles, and only for a while. In new titles and VR, 1060 will be laughable.

>>55438118
you dont really know how marketshare works do you know

>>55439159
He said the 970, and he was referring to my post which showed 970s
There isnt a 980 on there.

>>55439143
They lied and fucked up

480 was supposed to be a 150W card, turns out it actually draws 200+

But not only that, it draws too much of that power from the slot on the motherboard, which can actually damage your pcie slot

>>55439172
1060 is 15% faster than the 480 according to benchmarks

Actually 25% faster in VR

>>55439188
yep I found a nice link to that info
http://www.hardware.fr/articles/951-9/consommation-efficacite-energetique.html

before launch
>polaris, much efficient, such power
after launch
>here's how you flash your bios to draw less power from pci-e slot
>you can have 8gb in 4gb cards, isn't that great? why did you pay for an 8gb anyway?

It's a drama at every turn, it's not funny anymore.

>>55439188
>>55439201
I don't see 200

>>55438918
and you really think paid shill care about technical stuff?

>>55439234
You have to put on gamer glasses to see it.

>>55438033
card has more perfomance with less voltage
>AMD GIMPING PERFOMANCE

>>55439195
Only in Nvidia's intentionally leaked cherrypicked benchmarks. Nvidia's VR benchmarks are done with a yet unreleased game called Barbarian so they are even less representative.

Actual leaked benchmarks put it at less than 10% faster than a 480. That's pathetic considering the 6GB variant will be around $300.

>>55439201
AMD RX480 Lowest power consumption out of the entire AMD line as they intended. Thank you AMD!

>>55439256
You mean they may have fixed it then. What power does it now pull at peak

>>55439197
>Can pay $800+ for a GTX1080
>Can pay $800 for a Vive
>Can't afford a $20 DP to HDMI adapter.


Even if it couldn't be fixed by driver (which it could unless the hardware is different which then it would failed DP1.3 compliance not to mention the same has happened before with both AMD and Nvidia where bios updates were released to improve compatibility with some monitors)

You're could still workout the issue pretty easily.

>>55439272
The leaked benchmarks put the 1060 at 15% faster than the 480

Considering they're the same price, the 1060 is a much better deal

>>55439253
>>55439201
>>55439188
The 970 is only supposed to draw 145 watts according to nvidia. We all know that's a lie, they fucked up big time with the power consumption and the vram.

>>55437832
The only difference between 6 and 8 pin plugs is the 8 pin has 2 additional grounds. Specs on those cables are over half what they're actually capable of. Kind of like how pcie 16x used to have a rated 25 watt limit.

>>55439188
show me a gaming loop bench that the card actually reached 200 watts paid shill
the maximun they found was 164watt

>>55439284
fuck off shill. The only leaked benchmark that says 15% is Nvidia's own bullshit graph. All other leaks say 10% or less.

>>55438118
>This damage control.
Stay buttravaged shill.

>>55439284
what leaked bench?
the one graph of nvidia?
or the firestrike result of forbes that showed the card running on 1506mhz? lol
the card has less hardware less compute power less memory bandwidth the only scenario i can see it being faster is being overclocked like the one we saw and on certain gameworks games
and im not even gonna mention dx12 with such a cut down version gonna be so far behind hubble wont be able to see it

>>55439306
Tom's got the 480 to over 200W on the pcie slot alone.


The 480 is a disaster power wise.

>>55439342
>or the firestrike result of forbes that showed the card running on 1506mhz? lol


Why do you say as if thats something surprising? we saw how high does Pascal clocks, Maxwell OC average headroom was 1500mhz and Pascal blows that out of the way on stock.

>>55439347
>Got
Fuck off nigger

>>55439347
dont be a liar the card on gaming loop as per toms is 164 watts
on torture loop 225watt
unless ofc you are suggesting that the torture loop is the normal condition then you really need to stop shilling

>>55439391
>225watt

Yup the 480 uses a ridiculous amount of power, it's incredibly inefficient.

>>55437832
they'll roll a "compatibility" mode that complies with both connector and slot specifications, along with shifted power distribution which will be still off spec for 6 pin connectors.

reference card owners will literally run their cards off-spec if they want on par performance with shifted power cards, as amd already said it'll have a performance impact (whatever "minimum" that is)

>>55439283
>I know nothing about DP and HDMI Technology because I think its just different cables.

Do Nvidia cunts accept faulty hardware all the times?

>>55439405
>>they'll roll a "compatibility" mode that complies with both connector and slot specifications, along with shifted power distribution which will be still off spec for 6 pin connectors.
It's a toggle though

>>55439359
i will say why

all the cards we have seen so far with the shitty nvidia cooler cant sustain the boost clocks (that are 100mhz above the base on both 1070 and 1080)for more than 10 mins
and somehow forbes shows on STOCK cooler that the 1060 not only delivered 1506(which neither 70/80 could do even for 2 mins on stock cooler)mhz and then calls it a base clock
i mean literally is someone here stupid enough to think nvidia somehow will launch a card that has 400 mhz more as a base clock than 1070/80
>>55439403
literally you are not even able to shill correctly stop posting and go read the nvidia "how to shill for dummies" book

>>55439415
a performance lowering but standard complying (as written on the box - pci-e compliant) toggle, yes.

>>55439403
>480 uses a ridiculous amount of power
How much does it draw? and post link.

File: 103-Overview-Power-Consumption-Torture.png (95KB, 601x2156px) Image search: [Google]

95KB, 601x2156px

>>55439403
970 will pull 240 with the same test.
They advertised 145 watt tdp
They advertised 4 GB of RAM

nvidia lied

File: 1457034882432.gif (2MB, 504x274px) Image search: [Google]

2MB, 504x274px

>>55439342
AMDcucks are hilarious

>>55439442
First the shills on here say you need to talk about average power consumption but if its the 480 then its the spike consumption that counts.

File: Untitled.jpg (81KB, 644x559px) Image search: [Google]

81KB, 644x559px

>>55439425
>nvidia somehow will launch a card that has 400 mhz more as a base clock than 1070/80
What the fuck are you talking about you stupid nigger?

File: 02-Overview-Gaming_w_600.png (21KB, 600x450px) Image search: [Google]

21KB, 600x450px

>>55439433
Ignore shills.
The RX 480 cards affected by the aberrant power draw issues are only pulling 164w~ peak. TPU measured 163.
Slightly higher than the rated 150w TDP, but lowering the top pstates voltage also addresses this.
This board is rapidly becoming a place where all logic and sanity goes to die.

Jesus fucking Christ AMD is so bad they can't even gimp a card properly.

>>55439525
>>55439507


Tdp of a 970 is 145 watts
Advertised ram is 4gb

It's a joke of a card I don't know how they could fuck up so much

>>55439403

>t. Paid and poorly trained NoVidya shill

File: NsHILLed-1.png (72KB, 1030x600px) Image search: [Google]

72KB, 1030x600px

>>55439403
Here fagget hold this.

>>55439507
do you actually believe that paid shills care about logic?

they are following the same pattern as the ones on ocn...
literally accounts that never was used for 2-3 years where activated on release day and start shilling
ocn banned 23 of those accounts...

>>55439545
Yet they sold and still sell so well.
Answer is theyre a good card. I mean its taken 2 years for AMD to bring out the 480 which is similar.

>>55437899

lmao, is that real?

>>55439557
Uses more than a 1070. what does it do with all the fucking power it sucks up?

>>55439557
They are rapidly loosing any rational talking points. Powergate will soon be fixed with a driver update. Aib's will have 8 pin, and more oc headroom. 8gb for the price of 4? Based AMD, thank you.
Criticism left: it's not as good as a $500 card.

>>55439585
Melts PCIe slots, apparently.

>>55439581
ofc it is would be stupid to have 2 production lines so early amd just bios locked down 4gb

>nvidia 4=3.5
>amd 4=8

File: amdealer.jpg (62KB, 533x800px) Image search: [Google]

62KB, 533x800px

>>55438918
>psst, official spec numbers are for noobs
>we're all overcl-.. ahem.. end users here, trust me

>>55439602

yeah i know about the 4=8

but is the sticker real?

>>55439576
The answer is obvious: nvidiots are cucks.
And the 390 has been a thing for years now.

>>55439617
yes

>>55439469

Average is the only number that matters regardless of gpu - momentary spikes last for milliseconds and anything but the shittiest of psus will handle that.

>>55439601

The miner that melted his motherboard 24 pin connector was hilarious using an unpwoered riser to run 3 480's off a board with only two pcie slots. Naturally running more cards than the board supports without additional power support (hence why powered risers exist) is going to put strain on components.

>>55439624

thats like the most awesome win ever.

>>55439602
https://youtu.be/HYb7FYfIc2g

The 8gb card you can change to 4gb but it on the 4gb card memory is 7gigabits per second whereas the 8gb card dumbed down to 4gb runs a 8gigabits per second

>>55437899
Thanks AMD.

File: 45498469.jpg (239KB, 1024x1149px) Image search: [Google]

239KB, 1024x1149px

>>55439635

>>55439585
Yes it uses more than the 1070 I'm happy all of a sudden you can read. Nvidia is a head of AMD in efficiency. This is AMDs first efficient card in and of it self. You won't find an argument about that from me.

nvidia the way its made to stutter ™

https://www.youtube.com/watch?v=uxnFv8cnjZA

File: rx480-13b.jpg (157KB, 1920x1080px) Image search: [Google]

157KB, 1920x1080px

>>55439585
Its a factory overclocked card, it doesn't have that peak 1266mhz clock because of any arch changes. 4th gen GCN has the same 8 stage pipeline that all prior implementations also had.

Look at Polaris 10 in the RX 480 vs Hawaii/Grenade used in the R9 290/x and R9 390/x.
Polaris comes in at 36CU vs Hawaii/Grenada's 44CU, 40CU for the binned chips.
Polaris has 5.7 billion transistors vs Hawaii/Grenada's 6.2 billion
Polaris has 32 ROPs vs Hawaii/Grenada's 64

The 4th gen GCN arch used in Polaris has higher performance per clock inside the CU than prior GCN arch, but when it comes to gaming performance you don't see it. Why? The limited pixel throughput from having half the ROPs. This is why the RX 480 has a higher stock clock than any other GCN card to date. AMD opted to design a die with a simple layout and the least transistors possible to maximize yields and profit margins. Clocks were pushed as high as possible to try and make up for some of that lowered pixel throughput, and the cost of this was increased power consumption.
Creating a 64 ROP die would have demanded a much more complex layout, with more transistors, and lower yields. They would have been able to clock it much lower, and it would use less power over all, but it wouldn't have been as economical.

The other massive part of it is the memory. Out of the rated 150w TDP the die itself is limited to 110w, and 40w is allocated to the 8GB if 8Gbps VRAM. A wider memory PHY with slower memory, or GDDR5x would lower power consumption significantly, but again wouldn't be as economical.

Thats why the RX 480 draws more power than the GTX 1070. Its a factory OC card that sacrificed power consumption to ensure it would be profitable at a $200 price point.

>>55439634
Several people have had PCIe-frying experiences, not just the miner. I'm not going to get into it though. It DOES happen, even to people using their cards normally, but it is getting fixed and the update should be out soon, so no worries.

Are the people with fired slots able to get like reimbursement from AMD, or nah?

>>55439670
nvidia was never ahead
you cant compere a card that boast with hardware for all its features with one that miss most of them and is relying on the OS to do the work for them..
hence the awesome gaming experience you get on dx12....
see>>55439676

>>55439676
The 480 stutters see below 9m40seconds in
https://youtu.be/HYb7FYfIc2g

>>55439680
i dont know 161 watts on 1070 and 164 watts on 480 isnt really anything big to even go crazy about...

>>55439684
I'm going to withhold my judgement about boards frying until I see concrete, independent confirmation. An anon post on the internet is just not enough for me.

Do I think it's possible? Sure. It could happen.

>>55439699
i dont see anything on that time you say...

>>55439684
>>Are the people with fired slots able to get like reimbursement from AMD
Well since there weren't any people with fried slots I can't say any have

File: giphy.gif (1MB, 699x539px) Image search: [Google]

1MB, 699x539px

>>55439692
Higher frame rate more stutter. This makes sense now. I was reading an article about AMD is not using the newer DX12 API/driver(Some thing) because the fixes in the new DX12 where for software based Async which AMD does on hardware and Nvidia pushed on to the GPU. So the AMD cards would use more energy where the nvidia card would send all the work to the CPU? Do I have that right?

>>55439718
Google it. If you don't mind Leddit, there's a massive thread ongoing on there, with like 25+ updates with legit reviewers weighing in as well. Ditto for the AMD forums. Just Google it and things will crop up.

I'm not saying it's common, because it's not, but yeah, it's definintely happening.

>>55439721
wrong one apologies. I linked the incorrect video

This one.
https://youtu.be/2kjIHgq2zBU

>>55439684
several?
you mean the idiot with the asrock motherboard or the miner that actually fried his 24 pin not the pcie because he installed 4 rx480 with a 450 watts psu...

>>55439707
Its not a big deal, 150w~ for a mid range GPU is perfectly acceptable. Perf/watt is much more critical on the low end for entry level cards, mobile GPUs, and for the highest end enthusiasts parts.

We've already seen that Polaris 11 has markedly higher perf/watt than Polaris 10, and its because its clocked right in the sweet spot for GCN. 850mhz with a 0.8375v vcore. The card AMD demoed vs the GTX 950 was only pulling 50w. Thats the level of perf/watt we'll see from Vega, only with HBM on an interposer which uses far less power than any other type of memory.

>>55439676
Wonder how itd do at a lower fps. Id like to have seen the results from it when they all ran the same framerate.

File: xBgSksM.jpg (18KB, 340x260px) Image search: [Google]

18KB, 340x260px

>>55439736

No.

Hardware is ALWAYS better than software for a given task. GCN based cards (especially hawaii and newer cards) pull more energy when async compute is leveraged because the silicon that allows for it is no logner sitting idle and thus requires power.

AoTS is a GCN killer in that regard - on the higher clocked models you can see some truly ludicrous pwoer draw (and equally insane performance - a 2/390x rivalling a 980ti for example).

Pic unrelated, its the sort of shit you expect to see on /g/.

File: 20160701_144937.jpg (2MB, 2000x1240px) Image search: [Google]

2MB, 2000x1240px

>>55439743
>several
Here is the one board note the board, case and condition.

daily reminder PCI-E 3 provides up to 300w
if you own a shit mobo and psu it's your fault

>>55439781
He shouldnt have dropped that turd in the bottom of the case.

File: c2b856ed_58dc_48ee_abb0_1a51318a2475.gif (2MB, 344x338px) Image search: [Google]

2MB, 344x338px

>>55439770
>Dx12 is not multithreaded

>>55439781
That's disgusting.

File: 1461002493442.jpg (842KB, 2048x1365px) Image search: [Google]

842KB, 2048x1365px

>>55439770
Boom! Got it, thanks for the explanation anon!

>>55439740
I think they're fixing it in 16.7.1.

Still, the 1070 1080 Fury comparison video is bullshit.

http://www.guru3d.com/articles_pages/total_war_warhammer_directx_12_pc_graphics_performance_benchmark_review,6.html

>>55439736
those fixes are since early 2016 on dx12 to accomodate nvidia
they are using the os and cpu to do async while the rest are on the card hence the stutter you see
nvidia never actually pushed it nvidia since 2014 was saying they were ready and not only that they said they were ahead of amd http://www.legitreviews.com/nvidia-highlights-directx-12-strengths-amd_138178

but the truth is after august 2015 when the guys on beyond3d started to test the maxwell they found out that it doesnt have any sort of hardware sc..only a small arm chip between the driver and the gpu to controll the workload
their cores cant flip or jump or pause midcycle
its literally a card for dx11 and nothing more
>>55439740
yes we know 480 is stuttering on gtav days now but this has nothing to do with what i showed tho
my video is the prime example that they are lying once more
>>55439747
im sorry 1070 is rated on 150 watts yet it goes on 161 on heavy gaming loop
480 is rated 150 watts and goes on gaming loop at 164 watts
>>55439781
yes the same idiot from amd forums
he literally said that he was playing for 7 hours and benching the shit
what he didnt said is that the d6 code is purely vga problem and error 99 is purely pcie problems(those are taken from asrock site btw )
when people called him out he just went away..

>>55439770
Nano would be a beast at 14nm then with its low clocks?

>>55439829
If the Fury ASIC were ported to 14nm and had updated PowerTune IP it would probably average 100w or less at 800mhz. Thats just too expensive to produce for the level of performance provided, even though perf/watt would be exceptional.

>>55439829

If you simply scaled it down then yes, it would be a beast. The fiji die that powers the fury line on 28nm is absolutely enormous (one of the biggest dies ever created - only the titan x and 980ti are bigger). Plus AMD's chips tend to be considerably denser than Nvidia's so thats a factor into heat dissipation.

>>55439847

To be fair GCN's sweet spot is 800-900mhz. 1ghz+ only exist for AMD because they had to jack clocks up to get similar performance. An underclocked 290x is surprisingly power effecient once you start going below 900mhz.

>>55439847
if? we are expecting them with hbm2

>>55439635
it's not a win, it's literally a crippled product.
You HAVE 8gb, but AMD locked it to 4gb. Awesome 'win'.

>>55439883
I'd bet on Vega filling the same TDPs as existing Fury cards, nothing small enough to hit 100w.

File: you.jpg (7KB, 197x256px) Image search: [Google]

7KB, 197x256px

Wait for the 490, its the 1080 killer.

>>55439893
Reasoning such as this is why I put no validity into nvidia shill talking points.

Upgrading from a 7970Ghz what Is the most reasonable card to get In the same tier, 1070, 980 ti, fury x, fury or 390x? Unfortunately waiting for top end 400 series Isn't an option since I'm getting the card while on a trip to the US for half of what I would pay on this shithole of a country I live In.

>>55439893
lol there is a bios out there already

tell me can you unlock the 0.5 to work as intended with the 3.5 ?
>>55439907
whoever counts amd to just have such a small tdp given the amount of hardware they have on the cards is delusional

>>55439893
>You HAVE 8gb, but AMD locked it to 4gb. Awesome 'win'.
And charged $40 less
And you can get the extra 4GB by flashing the bios, so what you're actually doing is rewording "You get 8GB for the price of 4GB" to make it sound like something bad happened.

>>55439893
>bought a 4GB card and got an additional 4GB for no extra charge
Looks like a good deal to me.>>55439893

>>55439917
Mail the box back and put card in luggage. Might save some import tariffs. Do not keep your receipt on you.

>>55439893

If its a simple bios lock then fixing it is easy. Plus the most likely reason AMD did it is the 4gb chips they were using either went EOL or simply couldn't meet demand so using the 8gb setup and bios locking it is the simpler way to meet demand for the card.

After all at this point GDDR5 cost pennies to bolt to a chip until you hit mega densities.

>>55439915

If AMD went mental and gave it 128 ROPS a theoretical 490/ fury x2/whatever would not only be enormous but have unrivalled performance.

>>55439930
tell that to 8 gb customers

good deal indeed

>>55439930
The memory runs slower on the 4gb cards

>>55439942
$240 for $350-450 card 3 weeks ago? Sure, it's a great deal.

>>55439939
>If AMD went mental and gave it 128 ROPS a theoretical 490/ fury x2/whatever would not only be enormous but have unrivalled performance.

10+ billion transistors and one hell of a crazy die layout

>>55439942
paid shill dont even know amd gimped by 1gbit the memory
paid shill doesnt realise he just helping the situation
paid shill needs to be fired

>>55439952
http://www.amd.com/en-us/products/graphics/radeon-rx-series/radeon-rx-480

Does it, though?

>>55439942
It's not like this kind of stuff is new. Softmods have been around forever.

>>55439984
>Memory Speed (Effective) 7 Gbps or higher
Yes, it does.

>>55439975
vega cards are from 15 to 18 billions as per amd
if someone thinks amd will give a 10 billion chip on polaris price must be crazy

>>55439984
according to this guy yes. He tested the 4gb to the 8gb
https://youtu.be/HYb7FYfIc2g

>>55439938

Ill just bury the card In loosely crumpled clothes which normally does the trick well enough when used with the big ass suitcase I have and I don't really care enough about the box to go through the hassle of mailing It back, still a nice tip, hadn't thought about that.

>>55439977
oh, amd also said its 8 gb versions may have gimped memories too, so there goes your arguement.

>>55439952
>literally flash the BIOS to the 8GB bios
>VRAM runs at the clock speed as any other 8Gb card
Not seeing a downside.

>>55440013
watch the video posted >>55440009
The 4gb is 7 per sec
The 8gb dumbed to 4gb is 8 per sec.
He points this out

>>55440029
Is 1GB/sec a major difference?

Is the extra 4GB of RAM for free worth the 1GB/sec hit?

>>55440029
in real life this is 3-5 fps down
literally its like what 3%? slower?wow i guess its the same as having 3.5gb

>>55440010
They will search everything, including your phone, if they think you are trying to get around taxes.

>>55440045
not really as you are not using the max 3.5gb when gaming all the time.
Its a piss poor comparison, Also why compare it to that?
Im not comparing amd v nvidia im pointing out the fact the 8gb card has the faster memory

>>55440029
And 4GB flashed to the 8GB bios runs at 8gb/sec as has been confirmed by techpowerup.

>>55439922
>>55439923
>>55439939
Yes, flash the bios and void your warranty.

>>55439303
Well yes, the problem with the 6-pin connector is that it has fewer ground wires than supply wires.

Id rather buy a card, plug it in and it works.
All this fucking about and flashing, overclocking etc.
Just buy a decent card

>>55440077
being this stupid
>>55440074
it was a tard comparison lol i know but the fact remains in real life condition that 1gb/s doesnt have the impact the idiots keep saying here
also the memory modules are identical to the 8gb ones so that gimp must be bios locked too

>>55440095
> not wanting higher performance for free
k

>>55440107
You obviously only have ever bought cheap tat then

the shitposting is real in this thread

>>55440121
It doesn't matter what fucking level you buy. You can buy a 1080 and still overclock it and eke more power out of it. It's irrelevant what level of product you buy, considering you can always make it better via overclocking.

>>55440133
>amd logo
>expects no shit posting from paid shills
>lel

>>55440135
the majority of those who buys ever 5 -6 years cards dont care about overclockability or power issues or 3.5gb of memory or whatever for them if the card is working when you put the connector its fine

>>55440135
It does. Problem is you dont know because like I say you are one of the budget buyers and thus havent experienced the other

>>55440051

See, that's one of the few "good" aspects of shit holes, the same way crime Is rampant, tax evasion Is just as common since It's just too troublesome the Inspect everyone that arrives from abroad so you either have to be a very frequent flyer which makes customs officers suspect that you're making a business of contrabanding or just be exceptionally unlucky to get randomly selected by customs for Inspection.

People don't really give a fuck about customs In airports here, traveling abroad with at least two bags to carry back all the shit you buy Is the norm and a few people just don't give a fuck and even bring whatever they bought In the big ass retail boxes. I've personally seen full sized bike and home theater boxes which passed right In front of the customs officer without a hitch.

>>55440107
Overclocking always introduces more instability. It's never a good idea if you don't actually want your shit to crash.

>>55440170
> I bought a card for $X
> I get X speed
> I spent the exact same amount of money
> I get a faster card
Why is this bad?

>>55440195
Then do it safely and you're completely fine.

>>55440146
I wonder how many Chinese slaves AMD pay to shillpost?

No drivers. Only poorfag cards and 20% of the marked. When you look on /g/ it's like they got 90% of the graphics marked.

>>55439795
No it doesn't, it only goes to 75W.

That's why the 480 is causing so many problems on motherboards.

>>55438642
>only has an additional ground
You can't ignore that, though. The 6-pin connector has three supply wires but only two ground wires, so the ground is effectively its bottleneck.

>>55440195
Not always, it depends on how high you overclock and the quality of the silicon.

>>55439557
Even looking at your own chart, you can see how inefficient the 480 is.

The 1070 is both faster and using less power than the 480.

The 1060 will be even better at the same price, really there is no reason to even consider the 480 anymore

AMD can't get anything right anymore, it's no wonder Nvidia has 80% of the market now.

>>55440195

Yeah, tried overclocking my 2500K back when I first got It and It ran just fine for a few days but after a week or so of having my computer on continuously It would blue screen so I just gave up on overclocking anything, I'd much rather my computer on for a whole month(only ever gets restarted to Install updates) than a bit more performance which I wasn't lacking at the the time and am still not lacking today, five years later.

>>55440205
so /g/ is basically rich vs poor?

Sad that even Logical Increments are shilling this trainwreck now

https://twitter.com/logicalinc/status/750394569015980032

>>55440252
It always has been

>>55440273

Does that have any affiliation with the falcon?

File: trip.png (46KB, 168x244px) Image search: [Google]

46KB, 168x244px

>>55440307

nvm

File: jflC60V.jpg (968KB, 3187x1793px) Image search: [Google]

968KB, 3187x1793px

>>55439684
>>55439718
These are the kinds of boards that fry.

Fuck AMD those scamming fucks

File: DqhBcQ4.jpg (69KB, 900x598px) Image search: [Google]

69KB, 900x598px

>Nvidia says you have 4gb when you have 3.5gb
>AMD says you have 4gb when you have 8gb

>Nvidia gimps cards through drivers
>AMD gives a 3% performance boost within the first 2 weeks

Apoologize. Right now.

>>55439130
Nvidia literally said their operating temperatures were 105 °C

>>55437789
lol this nvidiots are panicking. AMD acknowledged this issue, and had a driver fix within a week, meanwhile it took nvidia AGES to even acknowledge the 3.5GB vram in the 970.

I can't believe you fuckers are defending such a shitty company.

>>55440205
I guarantee you AMD couldn't afford to shill on /g/ of all places. Nvidia's shilling is much more prevalent and widespread and you can easily tell if you spend some time on other forums across the web. If AMD has shills there aren't nearly as many as Nvidia shills.

Case in point, this thread.

>>55439095
>gameworks meme

why can't amd optimize standard d3d code?

I will buy nvidia card next, just because of all these pajeet cucks spamming and praysing async compute! thinking it provides some unlimited power, the same shit sony faggots where spamming about cell cpu.

>>55439613
What the fuck are you even trying to say?

>>55442722
yes not the fact that he doesnt have a single fan on the case lol..

>>55437968
AMD barely makes any profits on the cards. They are literally selling 8GB model as 4GB (a loss). If they don't outsell the 1080 5 to one it's a huge loss for AMD. That's like saying AMD is in good financial standing because of consoles, pre-built, and budget builds.

>>55439060
Where in that chart does it show overclocked cards? They were testing different CPUs not different GPU settings.

>>55438211
> dx12
> ashes of singularity
> no games have good crossfire support
Face it Ashes of Singularity is literally designed to be a AMD marketing tool. They have ties to AMD.

>>55438343
They don't gimp them themselves. They use game works which gizmos both them and amd.

>>55444845
you really dont know how market works do you know....

>>55437789
nvidia damage control campaign still going strong I see

I guess the RX 480 is just that good

better put out that 1060 real soon you jew fucks

>>55440224
I'm not willing to spend $400 on a GPU and if that difference in price means the same (OR GREATER) power use for less performance so fucking be it,
>164 - 147 = 17w
>.017KwH * 24h = .408KwH
>.408KwH * 365.25(1 year) = 149.022KwH
>149.022 * .135(average KwH cost in USA) = $20

>370 - 230 = $140
>$140 / 20 = 7 years
And that is assuming constant full load use 24 hours a day every day for seven years.
You can kiss my ass.

>>55444910
so hitman
and quantum brake
and catalyst
and total warhammer
and deus ex
and star citizen
literally every single dx12 has ties with amd its something we know dx12 is desinged around gcn thats why pascal its like a gcn 1.0 like it or not every single dx12 will be favouring amd and already is

>>55444946
AMD makes no money and is literally on the corner of bankruptcy
>le you don't understand the market taking losses on hardware clearly helps them just.look at how g-g-gre.. oh fuck it.
The reality is simple they are literally losing thousands of dollars selling 8GB models as 4GB models just so they can avoid the "paper launch" meme. The people buying stock 480s are AMDs biggest fucking fanboys. It's worth the wait to say you're manufacturing more 4GB models and you don't have any. Winning by losing is not a strategy and AMD is proving just that. I can't wait for them to sell ore hardware in consoles or pre-built at a loss it clearly works so well for them.

File: images.jpg (7KB, 284x177px) Image search: [Google]

7KB, 284x177px

>>55444910
So it has been, so shall it always be

>>55445114
>makes no money
>exclusive deal with china for literally everything since u.s blocked intel to sell them
>100% locked console ecosystem
>ramping up their hpc department
>about to unleash their new cpu
>still having licenging deal for the adreno on almost 80% of the smartphones
>having locked the future api's
>stock market is growing rapidly
>literally making millions per day

>AMD IS LITERALLY ON THE CORNER OF BANKRUPTCY

>>55445057
Not a single one of those games has AMD 480x crossfire beating the 1080 or getting even close. AMD better its entire company on async compute and it was a huge mistake. Even if they could level performance with it there cards are still inferior OC wise or power efficiency or performance for watt. In order for AMD to make a card that performs the same as Nvidia counterparts on the enthusiast market they had to have it cost more and now they can simply match the price. Wow, what a huge revelation. AMD can actually compete with Nvidia on 1 API with a handful of games. That is just pathetic. Superior engineering > Async compute. Forever and always. You can't remedy hardware inferiority with software.

>>55445157
Have you looked at their balance sheet? They are losing money every quarter.

The funny thing is their desktop GPU & CPU business only makes them lose money.

The only thing even makes them any money at all is their console chips.

>>55444986
Just get a 1060, it will be out soon. It's the same price as the 480 with everything better - better performance, better power and no AMD motherboard destroying issues.

>>55444845
>AMD barely makes any profits on the cards.

They actually make no profit. Their desktop GPUs and CPUs are losing them money. I would not be surprised if they drop out of the desktop business altogether soon and only make chips for consoles.

>>55445157
Wow, and they make no money off of everything listed. A great future indeed. Maybe when AMD finally goes under after Zen you'll realize you're just a huge fanboys. Simply put there is next-to-zwro money to be had in consumer grade hardware whether it's pre-built or consoles or budget builders. Their command of that market means absolutely nothing. AMD has already had bankruptcy scares last year. Polaris and Zen were the last bone to throw to investors. Once Zen fails it is over for AMD.

>>55445191
its funny since async was born because the devs actually wanted to..

you dont even know why nvidia has such a big difference on power eff in the first place why do you even speak? when volta comes and nvidia will be forced to have a hardware sc in order to compete with amd on dx12 and the power skyrockets because of it you guys wont even have a single excuse left

also in almost all cases cf 480 on dx12 is AHEAD of a 1080...sorry to burst your bubble...
enjoy your stuttering while nvidia is emulating async on os and on cpu
https://www.youtube.com/watch?v=uxnFv8cnjZA

>>55445246
do you even know the growth market between gpus and cpus? its 1 to 6 you think amd got keller
literally the only person on this planet that every single big company begs to work for them for shit and jiggles?
keller was behind k7 k8 the good athlons the whole lineup of intels i series on polaris/vega and now zen
so please instead of you actually say anything more stupid at least get your facts straight

>>55445283
>power skyrockets

doubt it would have the same benefit for nvidia since they are not as geometry limited, and tend to have more fill rate as well.

asynchonous compute is only helpful to the same degree that your SPs or cuda cores are being stalled by other parts of the hardware.

>>55445157
ye but they are losing 100's milions every year, while nvidia is making of 100's milions every year. Only reason amd is even selling there hardware for consoles is because they are basically earning penies on each apu, nvidia declined this deal, since they would actually be losing money in the long run just like AMD is now. Just because you flood the market with cheap shit, doesnt mean you will make money at the end of the day (see: Intel, Nvidia premium parts and then see AMD cheap parts).
>about to unleash their new cpu
Zen is just another amd aka "another major disappointment".
>literally making millions per day
No they are literally losing millions every day for years now
>having locked the future api's
yet the best and most powerfull gpus are still Nvidia based and it seems like it will happen again with volta, while pascal is previous generation shrink and upgrade, volta is complety new architecture, with stacked ram, meaning it will be true generation which actually runs 4k decently.

>>55445386
stacked dram means it will be actually be located on gpu die itself.

>>55438219
2 grounds and 1 additional 12v

>Intel's CPUs are overpriced into high heavens, to the point where you have to save 20 years to buy a 3-gen-old used i5
>The possibility that AMD may go bankrupt is real
No. God no. Please, anything but that.

File: 1448917281249.jpg (39KB, 297x296px) Image search: [Google]

39KB, 297x296px

>>55445386

>losing millions every day

>>55445381
why you think amd has such a tdp on dx11 and just goes bang on dx12? literally using 50 more watts when async is on...
currently nvidia lacks such technology and instead they are using a small arm chip to handle the workload between the driver and the gpu
the last card nvidia had a hardware sc hardcoded was the 780ti thats why it can go head to head with 980 on aots..well its 2k since it doesnt have the memory for anything bigger

also no tiago from epic said with async they managed to decrease their latency from 6 ms to 3 ms on 1080p and from 10 to 4 on 4k using async..that is not "some degree" that is a huge boost but sadly he was talking only about consoles..i doubt we will see anything from them on pc..

>>55445386
i guess if i ask for any proof that you have insight on amd plan you will provide it right? or you just base your though on literally nothing?

>>55437899
>>55439668
Why do people care about the stickers on the boxes? Isn't that a common thing companies do? It's just a box

>All those uneducated post about DX12.

nVidia’s implementation has less granularity than AMD’s, and nVidia also relies on the CPU/driver to handle some of the scheduling work.

nVidia’s and AMD’s approach cannot be compared directly. nVidia does indeed support async compute shaders on Maxwell v2, and indeed, there are workloads where nVidia is faster than AMD, and workloads where AMD is faster than nVidia.


The problem with DX12 is about managing resources and pipeline states in order to reduce CPU-overhead on the API/driver side and nothing to do with either lack of async, now if the game can get a certain level of performance in DX11, clearly that is the baseline of performance that you should also get in DX12, because that is simply what the hardware is capable of, using only DX11-level features. Using the newer API, and optionally using new features should only make things faster, never slower. That’s just common sense.

If you see a game where the DX12 rendered performs worse than the DX11 path without increasing the graphics fidelity you can immediately tell they are incompetent developers.

>>55445538

>If you see a game where the DX12 rendered performs worse than the DX11 path without increasing the graphics fidelity you can immediately tell they are incompetent developers.

I, too, have played tomb raider. Funnily enough (whether by design or not) its the only game with gamesworks that also happens to run worse in DX12 than DX11.

>>55445561
Last time I checked GameWorks was only available for DX11.3

File: 1466798801355.jpg (28KB, 640x339px) Image search: [Google]

28KB, 640x339px

>>55439039
>difference of ~4fps

>>55445577
Nevermind that was just VXAO.

>>55445501
asynchronous compute can only improve performance is hardware that was sitting idle is now doing work.

even oxide has said they think that fiji is held back by geometry performance. this is what gives asynchonous compute an opportunity to make such a large distance

tiago didn't say anything that contradicts with me

here is a well known console graphics programmer on the subject

https://forum.beyond3d.com/threads/asynchronous-compute-what-are-the-benefits.54891/page-14#post-1835425

>Asynchronous compute is useful in similar cases where hyperthreading is useful.

>The first use case are stalls. A queue needs to wait for something (for example end of the raster operations and ROP cache flush to start sampling that render target as a texture in the following post process shader)

>The second use case is workload that is bound by resource limits or fixed function hardware. On GPU, there is a fixed maximum primitive setup rate, maximum fill rate, maximum texture filtering rate, etc. Bandwidth and LDS work memory is also limited. When any of these things are the bottleneck, the shader cannot run at maximum speed, meaning that some instructions slots are unused. Asynchronous compute can use these instruction slots (if that shader has different bottlenecks).

>>55445520
>Net income 2016 -$109 million reported by AMD themself
>http://www.amd.com/en-us/press-releases/Pages/press-release-2016apr21.aspx

File: image.jpg (24KB, 400x300px) Image search: [Google]

24KB, 400x300px

>>55445442
You laugh, but any industry without competition suffers. AMD keeps Intel and Nvidia on their toes, and keeps prices from shooting through the roof on even entry level hardware. You think Nvidia is gonna keep pouring boatloads of cash into integrating their features into every game when it doesn't help their bottom line?

>>55445520
>nvidia net income US$ 614 million

>>55445489
they only lost 109 milions in first quarter.

>>55445645
>How many days are there in a quarter of a year

>>55445620
thats a cute doggo

I have a 390, when do I upgrade? I'm waiting for AMD's answer to 1070/1080 since it's not like I'm starving for framerate at the moment.

>>55445314
Keep on holding on to your dreams. Reality is only a few months away. See you then. There is nothing that indicates any change on AMD and nor is there anything they can do. As I've already said they've betted fully on Async compute at the cost of power efficiency, card-design/engineering, etc. They're relying on software to fix hardware. Despite having first rights to HBM and other opportunities they've fucked up every single one. AMD has been designing their cards for future theoretical software because they lacked engineering prowess. The gamble is finally going to come to an end soon.

>>55445596
Hey maybe you can help. I remember reading something a while back about Nvidia requiring less of the processor than amd, so green would be less restricted by low end CPUs. I was trying to figure out if that was true and if it would be changing in the new generation.

>>55445442
Anti-trust laws exist for a reason. It would be a Chapter 11 bankruptcy and someone more competent would acquire AMD. The fact is that AMD isn't competition. They haven't competed on the enthusiast grade level in years.

>>55445673
365 days in a year-4quarters in a year so divide 365 by 4=91,91,91,92 days

So they lost 109 milions in 91days, meaning they lost millions each days, instead of earning milions

>>55445620
I fail how to see how AMD dying will stop Nvidia from investing into new features.

Nvidia invest millions in research every year https://research.nvidia.com/ features that AMD don't even considers at all or that might no have use until hardware matures which could take generations.

File: NVIDIA-GeForce-GTX-1060-Fire-Strike-Performance-900x635[2].png (405KB, 900x635px) Image search: [Google]

405KB, 900x635px

First 1060 benchmarks in

>>55445596
this is quite outdated given what we know based on the behavior of both companies cards not that its not true but given how complex are the cores on amd cards we know they arent really there for only to sit and wait if and when a workload just comes in (power gating)

>>55445709
yeah, it's true, but probably not in every game. if you're asking if dx12/vulkan will help, then yeah. if you mean the new hardware generation? maybe i guess, but that would require amd to optimize their drivers more.

File: NVIDIA-GeForce-GTX-1060-Fire-Strike-Ultra-1-900x545[1].png (403KB, 900x545px) Image search: [Google]

403KB, 900x545px

>>55445752

>>55445620
How is AMD competing? Have they made a card on the enthusiast grade level that people actually buy for its metrics rather than it being AMD? They aren't competing at all. The market share Intel and Nvidia has on enthusiast grade hardware is daunting. Under anti-trust laws Nvidia wouldn't be allowed to exist. Also, Chapter 11 =/= Chapter 7. Someone will definitely acquire AMD.

File: jd0SyfF[1].png (467KB, 595x593px) Image search: [Google]

467KB, 595x593px

>>55445752
>>55445768
1060 a shit

>>55445758
the fact that they power down doesn't conflict with what i posted. you even said that yourself
>not that its not true

what is your point?

>>55445705
obviously it was a gamble...not the fact that they control all the consoles
and every game that is being ported from now on on dx12 will be naturally favouring amd noooo..
its a gamble..
thinking amd is dumb enough to just "gamble"
http://translate.google.com/translate?depth=2&hl=fr&ie=UTF8&prev=_t&rurl=translate.google.com&sl=fr&tl=en&u=http://www.hardware.fr/focus/89/amd-mantle-interview-raja-koduri.html
read what raja was saying back in 2013..

>>55445795
the fact that they NOW power down is what is going on
http://www.freepatentsonline.com/20160085551.pd

AMD as we know it will die soon, but they will sell their GPU business to Intel, so hopefully Intel will be able to do what AMD never could.

>>55445853
still not seeing your point

Nvidia GTX 1060 cannot beat a 480. See above scores.

>>55445895
1060 easily beats a 480, it's 15% to 25% faster

>>55445739
>>55445769
I really doubt anti-trust is going to help if it comes to it. You know how hard it is to split up a tech company in any meaningful way? Plus there are all kinds of loopholes. Look at microsoft. All they have to do is claim that their "divisions" work independently and they get away with tons of shit they shouldn't. And amd does contribute. Half the time when Nvidia or intel come out with a new feature, its a modified version of something AMD started. And look at how every generation Nvidia has to keep dropping prices on cards to match AMDs deals, and if you want something i3 or below in cost you strait up go AMD. Believe me, things can definitely get worse.

>>55445917
Yes it can get worse, all you have to do is buy an AMD card.

>>55445883
async relies heavly on the cores capabilties to deliver
currently we know that on amd they can flip jump or pause midcycle and given that the new cores along with the new programmable ace engines(HWS) are being able to just shut down or having a short burst up to 1.8ghz(or 2.0 i cant remeber the threshold ) makes async a lot more relevant to provide a significant boost

>>55445914
nvidia released a shitty graph showing 15% 2 days ago
4chan increased it to 25% literally 1060 is beating 1070 now

people are so fucked up

>>55445945

Nvidia's architecture really hates context switching due to the enormous latency penalties it brings. GCN simply doesn't give a fuck.

>>55445985
and that brings us to my original point..
pascal is gcn like more than people want to think
they just dont have any of the cores capabilities or the parallerism gcn provides..and slap it on a static software sc they have it makes everything go well...
i guess thats why we never saw any real video of the pascal cards running async till total warhammer..or to show what will happen if you use any cpu lower than i7...(and that is why the gcn 1.0 cards cant play total warhammer too..)

>>55445914
The score says otherwise. RX 480 for comparison.

http://www.3dmark.com/fs/9145527

>>55445811
it can favour AMD, but NVIDIA cards still outperform AMD parts. Guess what? Nvidia will continue to shit up AMD, just like Intel is shitting up AMD right now. Most powerfull AMD new generation gpu competing in low end with 1060. Vega will be out untill 2017 1Q, and if it actually is decent guess what? Nvidia just slashes prices and in 2018 releases volta.
you seem to hope for nvidia to fail and be forced out of market, but it will never ever happen.

>>55445811
I'm not saying it wasn't AMD's smartest move. The fact is this probably has been coming for a long, long time. You cannot compete without superior engineers. AMD still isn't competitive. They can now make cards that match Nvidia counterparts for the same price of course if they rely on DX 12, but now they're inferior OC wise, power efficiency wise, etc. Nvidia could always drop the prices too they have plenty. AMD is truly and duly finished no matter what happens.

>>55446156
refrence to refrence all the repsective counterparts on dx12 is behind and im talking on aots that has a low asyn usage... im not gonna even mention how badly maxwell and paxwell is already doing on total warhammer...
or how bad they gonna be on deus ex that will use way lot more of its thing on the compute pipeline..

you can shill all you want the fact is your company kept lying since 2014 about their capabilities in general
http://www.legitreviews.com/nvidia-highlights-directx-12-strengths-amd_138178

>>55446250
you are the one that shilling about async, while amd is nowhere near nvidia cards performance, keep shilling retard.
>MAH ASYNC
>MAH DICK

AMD is shit at API Overhead

>>55446240
in dx11? they are in neutral games.. they are doing FINE
the problem is and always was the gameworks games and the over tessellation that nvidia used

also
1)you cant compare to difference uarch on oc it doesnt work like that..for example nvidia in relation to their oc they dont get the same amount of perf increase as amd gets
2)power eff as i said earlier is something that dx12 is purely makes us aware... nvidia is lacking quite a lot of hardware in order to use it hence all the terrible videos that are getting on the surface with horrible stuttering on async or when the games use casting..

no nvidia cant drop the price and that is purely their own fault they tried so hard to be "premium" that they are locked on what we call "escallation of commitment" go read it on google
if they drop prices their investors will shit on jackie chan its not like someone of the board will like that move

>>55446155
Nope, look at reviews, the 480 gets like 12000 on Fire Strike while the 1060 gets like 13500

The 1060 is faster than the 480 and the same price, it's not even fair honestly.

>>55446291
>nvidia much better
>nvidia much better card
>nvidia best
>hurr durr
https://www.youtube.com/watch?v=uxnFv8cnjZA
>oh wait
>nvidia stutters
>literally killing the cpu
>hurr durr

>>55445535
It's to save costs, not sure why people are losing their minds.