>Pascal is simply Maxwell with higher clocks
>tfw cucked by Jensen once again
Does that eve matter if their cards performs better than AMD's?
>>54735156
It matters to me. Not op, and this may sound stupid, but I was pretty disappointed when I found out my gtx 260 was a cut down gtx280, that's why I always buy flagships like the gtx 680 I'm using now.
>>54735174
But Maxwell was an incredible architecture, incredible performance, incredible frame times and incredible temperatures.
Whats there to complain if the card performs better than AMD, shouldn't you be complaining to AMD instead for not releasing something that can't compete after a complete architecture redesign
>>54735228
>But Maxwell was an incredible architecture, incredible performance, incredible frame times and incredible temperatures.
>let's make a DX11 ASIC
>surprise surprise when it's suddenly great for gaming but shit for everything else
There's a reason why Nvidia's compute cards remained on Kepler for so long.
>>54735134
>mfw 256bit memory bus and gddr5 (x) in current year
>mfw still no async compute scheduler 5 years later
>mfw vega will shit all over it
next year is going to be awesome i cant wait
>>54735134
Don't buy rehashes then? That's how they say: vote with your wallet. I personally upgrade only when something at least two times better than mine comes out.
>>54735363
This.
Also people shouldn't give a shit what company is making what card, if it's a big performance upgrade over you existing hardware and the price to performance is good then get that card
>>54735302
Is it somehow supposed to be great for everything else but shit for gaming? It's a gaming card.
There are compute cards for the same price that suck for gaming and are great for compute.
http://www.newegg.com/Product/Product.aspx?Item=N82E16814132051
>>54735325
>next year is going to be awesome i cant wait
>card released in a year will be better
holy shit, someone call CNN
>>54735325
vega is october, Hynix came through with memory
https://www.youtube.com/watch?v=P7RSszCPt68
How much truth in this thing? Because it might destroy Nvidia in a year if it's marketing flavor at least 35% true.
>>54735325
>mfw 256bit memory bus and gddr5 (x) in current year
Specs don't matter, only the end result does, and currently the 1080 blows everything else away.
Are you insinuating that you know better than the hundred engineers who spent over a year designing this? Do you really believe that if there was a benefit to using a wider bus that Nvidia wouldn't use it?
>mfw still no async compute scheduler 5 years later
All async does is help AMD cards catch up to Nvidia, in the 2-3 games that have it. Nvidia has years to work on it by the time it'll matter to anyone not "playing" Ashes.
>mfw vega will shit all over it
You can see the future? We don't even know for certain how Polaris will pan out.
>>54735504
>you know better than the hundred engineers who spent over a year designing this?
You are correct about one year part.
RIP Kepler
>>54735486
This is no different that was Nvidia has been already doing for years now, downside games that use this will run like trash on Nvidia cards unless the developers works around it.
>>54735486
if it was true it would destroy not only nvidia but also entire console market
>>54735134
>they said that they spent 2 Billion dollars worth of R&D for pascal
>>54735443
Those are not compute cards. Maxwell double FP performance fucking sucks.
http://www.newegg.com/Product/Product.aspx?Item=9SIA1N82856337
>>54735603
it's not some pretty effects that you can slap on any game
this thing is engine defining, at least they sell it as it is
how hard it is to implement will know only people that work with engines, I'd bet on DICE doing it first
>>54735466
ill believe it when i see it
>>54735654
Yes, just like Nvidia specifics shaders don't work well on AMD cards.
>>54735562
KEK sold mine a couple of weeks ago.
was a good card but DX12 performance was unsusable as was vulkan and ogl4.x
for a 680 4gb it went quite well
>>54735644
>Corporation lied
Which is why you buy AMD, so you have less buyer remorse since you paid less but getting about the same performance.
>>54735562
we can agree it was the last great gpu nvidia made to this day?
>>54735644
it could be true, apparently Nvidia is already working on 10nm prototypes, probably why they delayed Volta.
>>54735504
Do you trust engineers responsible for 3.5?
>>54735134
>a new GPU is a bit faster than an older one
Like every GPU ever?
>>54735773
I don't think so. They said that together with those billions, they spent the past 2 years. So they clearly meant that they spent 2 billions on gtx 1080.
>>54735868
it was until last year that Volta was delayed.
>>54735562
gg gtx 650 now it powers my Linux pc to play a quake arena clone
>>54735644
Most of that money was probably targeted to some market outside of gaming with better application and use.
>>54735134
>Pascal is simply Maxwell with higher clocks
I love this meme, where does it come from?
Last I checked maxwell is not 16nm. Maxwell is not capable of >2000 Mhz core clocks. Maxwell sucks at OpenGL by comparison.
If you own a 980 or 970, you don't need a 1080 or 1070, actual gaming performance isn't the 2x they tote.
>>54735134
>Pascal is simply Maxwell with higher clocks
Not really, otherwise a 1080 would equal a titan X
>>54735649
>Maxwell double FP performance fucking sucks
That was the entire point of maxwell, remove all unneeded compute and make a pure gaming chip
>>54737896
I was the anon who posted >>54735302, I clearly said that.
>>54737868
Pascal is Maxwell + particular VR features and 2ghz clock allowed by 16nm
>>54737857
>>54737868
Honestly reminds me of Intel's tick-tock. We're on a tick generation, new manufacturing process and some small tweaks to the micro architecture to speed it up.
>>54737932
every generation will be tick until market makes them.
Can only hope for Zen and Polaris to be good enough or else the will have no reason to improve.
Why waste money on improvements when you only compete yourself?
>>54737959
>Zen
If the leak from currynigger tech was legit Intel is about to get their ass handed to them.
>>54737925
One architecture is another architecture with just a higher clock speed and magically better support in other things.
You people are fucking morons.
>>54738010
Link?
>>54735644
They did spend 2 billion but it was on marketing.
>>54737857
>>54738015
Probably the better term is there are Pascal and Maxwell IPC are identical. Pascal simply shrinks Maxwell with more subdivided SMs and adds tons of software support. Interested to see if there are real IPC improvements in Polaris though. Definitely there will be new memory compression and prefetch.
>>54737896
>gaming
>>/v/
>>54738053
http://wccftech.com/amd-zen-cpu-performance-double-fx-8350/
Don't get your hopes up.
>>54738015
I bet you're the kind of person that calls all GCN cards a rebrand too. Pascal is nothing more than Maxwell+ on a die shrink. If you look at the die diagrams you could see that.
>>54738091
>nVidia gets slower with DX12 and Fury X matches with 1080
>>54735134
Nvidiafags will defend this bulldozer
>>54735644
Yes. Bribing devs to use goyworks is r&d
>>54738333
The writing has been on the wall for the last two years if you follow game development (especially some of the tricks the console teams are pulling) - pascal (i.e turbo maxwell) is just a stopgap until volta rolls around and maxwell is a piece of shit for the level of parallelism DX12 (and hopefully vulkan) support.
GCN (particularly 1.1 onwards) actually supports some serious fine grain control that is currently also found in the ps4/xbone hardware but it has been crippled by DX11 hanging around for so long.
>>54737896
>the only use of GPUs are gaymez1!1!!
Kill yourself
I was just disappointed in the jump from 28nm to 16nm, just disappointed. It's nice that they released a card that performs better than the entire GPU lineup, but it's not even that faster than the current flagships making it an absolutely retarded buy for anyone running a 390/970 or above. It can barely do 1440p at high refreshes and cannot do UHD at 60fps, why would I buy it then? If Vega and big Pascal don't deliver 1440p at 120fps, then I'm skipping this generation. I want to move to a higher resolution monitor, but the GPUs are not running them as I'd like them to. At least the 1070 is believed to also beat current generation flagships at less than $400. Considering that I bought a 670 for $400 four years ago, it's a nice buy. I'd still be on the lookout for Polaris though, getting almost Fury performance for less than $300 would be a fantastic buy.
>>54735228
>Whats there to complain if the card performs better than AMD
that it doesnt, that it cant overclock like maxwell did, and that it throttles the boost clock after like 3 mins
>pascal is a rebrand meme
yeah, except for the completely redesigned SMs, ALUs, memory controller, new process?
>>54738512
console gpus are all gcn 1.0
>>54737896
fp64 is a useless meme even in the compute world. the only reason it's included on cards these days is because you need it to be spec compliant with opengl 4.x. there isn't a single game or application that makes use of it, outside of maybe some high end video editing software and scientific/simulation type things
yes, pascal is only about 15% - 20% better than maxwell ... so what?
what the fuck did you guys even expect?
of course it doesn't make any sense for anyone having a maxwell card to upgrade. why the fuck do you even want to upgrade your GPU every generation??
CPU's are stagnating for a long time by now and GPU's are simply starting to do the same. In fact - even a GTX 580 can still run every modern game @ 1080p (high settings) without too much of a problem.
pascals aim is to reduce power consumption and make slightly better performing cards much more affordable (to allow for mainstream VR) and it seems to do that quite well. (1070 will be quite cheap and they all only have one 8pin connector)
>>54735134
>Maxwell was brilliant
>we now have Maxwell with much higher clocks and lower power consumption
>this is a bad thing
>>54739056
>yes, pascal is only about 15% - 20% better than maxwell ... so what?
Only if you compare the 1080 to the 980ti which isn't it's predecessor.
You should compare it to the 980 since its a midrange chip, the GP 104
>>54739056
it's 20% faster than the last gen flagship that was double the die size, had more cores/rops/tmus and shits out twice the heat. it's 2x faster than it's last gen equivalent as well. people were just hyping up the die shrink to be more than it is, especially since tsmc 16nm and samsung 14nm are essentially the same 20nm process with finfet added into the mix.
>>54735156
It's not about competition stupid, it's about the false advertisement
>>54738581
That isn't even close to what he said, please gain some reading comprehension and come back when you are 18+ years of age.
>>54739125
How did they falsely advertise about Pascal?
>>54739300
Fake benchmarks, the temperature the card will run at under load, and max clock speed. It's a good card though.
>>54735156
>Falling for the nvidia is better then amd
>mfw going from a 770 to a 1080
>>54739361
>buying a dx11 card in 2016
>>54739440
because async compute is totally 100% of dx12 right lol
even taking that into account I get gameworks in the tradeoff which is naisu
>>54739483
Ask Kepler owners how gamesworks is treating them.
>>54739440
>async meme
t. indian AMD shill
>>54739440
>not realizing we'll all be upgrading before DX12 is mainstream
ok
>>54739495
My 780 is fine for now, I only want a new GPU for use in MATLAB. Granted, I don't play gaymewerks games. I imagine Kepler gets shit on in those titles.
>>54739495
I already said I have a 770 so I don't have to. It generally works fine excluding brand new games, but who the fuck cares about shit like the division anyway.
Daily reminder.
>>54739328
What fake benchmarks?
>>54739505
t. nvidia goy
>>54739573
What did you have to do to make AMD cards run better again? Oh yeah turn off tesselation
>>54735504
You seem to forget that Nvidia's goal isn't to make the most powerful GPU, it's to make the biggest amount of money.
Those hundreds of engineers aren't there to do anything else than generate profit.
>>54739582
>dx12
>windows
>amd sponsored
the trifecta of poor indian software
>>54739610
They spent more time optimizing with Nvidia senpai
>>54739641
>They spent more time optimizing with Nvidia senpai
nvidia was invited to optimize ashes at all considering the renderer was originally written against mantle and later ported to dx12
>>54739663
>was
wasn't
>>54735814
no, typically from the same company the new top end kicks the shit out of the last get to the point the last gen doesn't even look like an option, even when last gen was oc'ed, this time... 2% over last gen...
>>54739687
are you sure about that? because that hasn't been the case for a long time, we haven't seen exponential growth in gpu performance ever since we reached peak die size and node shrinks stopped coming every year.
>>54738172
my hopes are sandbridge 8 core at 500$
i think these are modest hopes given that sandybridge is bare minimum where amd will shit itself out at.
anything more (or less) is a happy surprise
>>54739591
>enjoys subpixel tessellation
No thank you.
>>54738015
the vr shit is all software side in how it renders though, there is no hardware that enables it, its software telling a gpgpu how to deal.
>>54737925
>2Ghz
Can't even hold the boost clock without throttling back to 1600mhz.
>>54739079
2% better then a 980ti for 150$ more.
yea, so worth it
the card, given the cpu cooler benchmark, will only get it to 12-15% better so there is your aftermarket cooler what you can expect for the oc on a 1080, still, even at msrp, it will be minimum 50$ more then the 980ti right now, even bigger gap when the better coolers come along.
>>54739782
>2%
in what world
>>54739782
The 1080 reference pcb (sorry, founder's edition) is a piece of shit. At least AMD (for hawaii and fiji) build their pcb's to withstand enormous punishment.
The 1080 as of right now is a terrible card if you plan to overclock in any shape or form as it simply cannot deliver the power to the core required to keep it at 2ghz+ - even if you use some custom cooling. The annoying part of it all is Nvidia hyped it as a premium card when it is nothing of the sort. If the founders edition did have a good pcb then sure you might be able to justify the extra cost but the flimsy piece of shit they delivered is not worth anyone's money unless you are a consumer whore.
Plus as its a midrange chip there won't be a 1080 lightning which would really unleash the silicon.
>>54739865
>At least AMD (for hawaii and fiji) build their pcb's to withstand enormous punishment.
the only good pcbs amd has released in recent history was the 7990/295x2/pro duo (they're all variations of the same pcb design) pcb and the fury x pcb. the 290/290x pcb had serious flaws, the reference models had over 30% failure rate because of it.
>>54739865
>The 1080 as of right now is a terrible card if you plan to overclock in any shape or form as it simply cannot deliver the power to the core required to keep it at 2ghz+
this should change one AIBs add in the missing power phase, power connector and give it a better cooler. it seems like nvidia's biggest failing was the reference cooler this time around, since now you need to pay $100 extra for it.
>>54739782
>2% better
top kek, AMD fags and their lies
>>54739908
>the reference models had over 30% failure rate because of it.
Going to need some sauce on that claim anon.
>>54739932
That is the whole point - its an overpriced card. What is worse despite the so-called MSRP I sverely doubt any AIB card is going to be cheaper than the FE outside of p[ressure from AMD causing Nvidia to lower prices.
>>54740273
not sure where he got the 30% from but it's been tested & is well known that hawaii has absurd failure rates, https://www.pugetsystems.com/labs/articles/Video-Card-Failure-Rates-by-Generation-563/
Polaris is worse than current gen. All those shills are funny as hell to me. Vega probably won't be even in 1080 range. Kek.
>>54740341
depends on yields. there have already been rumors about nvidia only shipping the 450mm2 gp102 as this generations flagship, i can't imagine it will be any better for amd since they're already stuck shipping a die smaller than gp104 and are using a process known to be inferior.
>>54740313
>puget systems
Christ might as well be backblaze of the gpu world.
Also note their data (as presented) doesn't explicitly highlight hawaii given the chart also covers R7 cards as well. Plus the asus 290x is known to be the worst of all 290x models.
>>54740382
The sole reason founder's edition exists is to hide low yields behind marketing dribble.
>>54739865
>The 1080 as of right now is a terrible card if you plan to overclock in any shape or form
https://www.youtube.com/watch?v=ZO_0SPvy_4o
note that he kept this oc stable after 2 hours of gaming.
AMD shills lying and spreading misinformation as usual
>>54740421
>The sole reason founder's edition exists is to hide low yields behind marketing dribble.
founders edition has nothing to do with yields, they're selling reference pcb/coolers for $100 more because of OEMs and watercooling autists.
the yields for gp104 are already good, tsmc 16nm has been maturing for 2 years and isn't derived from stolen technology like samsung's shoddy clone of it is.
>>54739717
look at the 7970/280x to 290x, that's a significant jump.
i guess nvidia cards are alot closer due to their stop gap end of life ti versions. even then, id want to see more recent benchmarks without gameworks cancer to see the difference.
>>54739738
16x is the last level you see a difference at, 8x is the level most people would chose just because performance over minor visuals being better.
>>54740565
the 290x shipped at the same clocks as an overclocked 7970 and was a 100mm2 bigger die, the 980 and 1080 are the same die size, as was the 680
>>54740313
from looking at that site, it seems the cooler they chose to test did not over other components outside of the gpu silicon for amd, so there is a reason those failure rates are abnormally high, they used a shit third party cooler.
>>54738091
Can't wait for Polaris almost matching the 1080 for 1/3 the price.
>>54740491
yields on the chips themselves may be ok, but not on gddr5x yet, that's the bottleneck.
>>54739575
Not him but he probably means the 1080 demo running at 2114 mhz on the core and was 67 degrees C during some bullshit rendering, when in real life it overheats and downclocks like a bitch.
>>54740727
>Edit 5/5/2014: We've received some questions about what brands and models of card we used. We primarily use Asus DirectCU cards whenever possible, and the basic Asus models when there are not DirectCU versions available. It is possible that the high failure rates are limited to Asus cards, but we have used Asus as our primary supplier for video cards for a long time now. This includes NVIDIA cards as well as the Radeon HD 7xxx and Radeon HD 6xxx cards which have a much lower failure rate than the Radeon R7/R9 cards. This is a clear indication to us that Asus is not the problem, but rather something fundamental to the R7/R9 cards themselves.
>>54740734
>delusional : the post
>>54740801
We already have benchmarks of the C7 polaris Capsacin Fire edition outperforming the Titan X
>>54740801
nvidiot
>>54740448
how are they overclocking it because everyone has shown that they throttle like a motherfucker and can't even keep the boost.
i'm willing to bet money that the people who buy gpus and watercool their gpus are almost a statistical anomaly. cant see many people willing to toss 600... scratch that 700$ at a computer part and then risk it by tossing the warranty away altogether.
granted love the bottom where they are saying only 5-6% better than 980ti
>>54740773
That still doesn't exclude the possibility that asus cards became shit for AMD in 2013. This is a more likely possibility given we KNOW the DCU cooler on hawaii flatout did not fit properly.
Without data from another source to compare against puget's data is worthless.
>>54740448
So exactly where does he say that he tested the games at 2hours ?. tried finding it but only skimmed results and notes
>>54740801
200 dollars and its as fast as a 1070
BTFO
>>54740927
>crossfire a $200 gpu and get the same performance as a card that is cheaper than the two $200 GPUs
who are you trying to fool, raja?
>>54740927
>$200
try $350
>>54740927
cool meme. Almost tricked me with that crossfire.
>>54740927
He thinks he will pay 200$ for a 480X.
medium kek
>>54740948
Try 175.59
>>54739102
in what fucking world is a 500+$ chip mid range?
>>54741189
When $1000 is the top end
>>54739663
mantles end game was to open it to everyone but got canned and moved off to vulcan before that happened.
>>54741096
>>54739575
>>54739328
I don't believe those were "fake" so much as "optimistic".
Perhaps after further testing they realized they couldn't ship it in that state and have it be stable in all instances.
>>54740341
polaris is the 280/380/x successor,
polaris is performing at around a fury level in the leak, possibly without proper drivers and definitely with proper cf profiles.
happy with the massive ass jump in performance, all thats left is price which is anticipated at 200-300$ due to the other 280-380x prices.
>>54741290
or they watercooled the card, and showed it off that way, so they weren't technically lying,
remember that kid in school who would constantly lie while not technically lying?
in my school they got their shit kicked in till they stopped, nvidia has been doing this for almost a decade now.
>>54741245
it wasn't. mantle had a ton of amd specific features that never got brought to vulkan/dx12. the reality is that it just didn't catch on and the only developers that ever implemented it did so because AMD paid them to.
>>54741376
Fun fact: Mantle is the base technology powering AMD's LiquidVR.
>>54741305
>happy with the massive ass jump in performance
fucking what
>>54741448
I sure am!
>>54741447
and? do you actually think people are going to use AMD's low end GPUs for VR? VR is extremely demanding.
>>54741486
Given one major goal of Polaris is to reduce the entry cost for VR, yes.
>>54741498
VR isn't going to have playable performance on any of the poolaris GPUs.
>>54741482
Does he really have that tattoo or is it a shoop?
>>54741482
No wonder AMD has such drive to focus more on mid range. They all fucking imbeciles who don't see any difference.
>>54741512
>>54741588
Nvidia is still well behind AMD on the software front when it comes to VR and we all know how they are on the hardware.
>>54741513
Dunno - I grabbed it from google.
>>54741512
>>54741719
from what i can tell, nvidia jumped ahead in vr due to rendering out an entire scene, and haveing 2 camera angles, it cuts the need to render the entire scene twice.
its a thing that blender does, the video isnt loading but i think its soemthing like this
https://www.youtube.com/watch?v=sB09T--_ZvU
they render out the scene/bake/move the camera a bit, in the baked shit, and its able to render it faster this way then render 2 entirely different viewpoints at the same time.
could be wrong with this, but its an idea how the new way to render vr is being dealt with.
>>54741588
difference in what?
>>54740994
nope, but it's pretty fucking obvious that the 480x will be sub 300 and the 480 will be <= 250$
the 380x was 250$ at launch. Jesus you nvidiots are dense.
>>54735156
Not atm; it's the same problem on the CPU side of things: there isn't real competition at the high end, so Nvidia/Intel can screw with the "new" product releases all they want.
