P O L A R I S
Images are sometimes not shown due to bandwidth/network limitations. Refreshing the page usually helps.
You are currently reading a thread in /g/ - Technology
You are currently reading a thread in /g/ - Technology
June 1st: Polaris day
http://ir.amd.com/phoenix.zhtml?c=74093&p=irol-newsArticle&ID=2170028
>>54651015
>Only an update.
October delays hahahaha
>>54651028
remember: Pascal "launch" is merely a paper launch
>>54651040
What do you mean, paper launch, you can get a 1080 on the 27th already
>>54651056
I guess we'll have to wait and see, but anyone buying a reference (founder) 1080 for 100$ extra is a fucking retard
>>54651158
They are yeah
AMD has been to quiet about Polaris so far though
>>54651183
too quiet for what? Polaris is not competing for $400+ price segment, so they're fine.
>>54651188
Then what are they competing for, last years performance?
>>54651213
the biggest market segment on PC market is the medium ~$300 and below price segment.
>>54651220
Yeah but those are also people who don't regularly upgrade.
I think most people looking to upgrade each year are people in the higher segment
I am very hyped^tm
>>54651247
This is true, but they are more likely to upgrade if the upgrade is worth it to be honest familia
>>54651015
They better have an itx card I can masturbate to
>>54651015
I'd be interested in $200 970-tier performance if AMD could ever get their linux drivers sorted.
>>54651264
970 and 980 owners have plenty of reason to upgrade with Pascal, 980ti owners, not so much
>Polaris updates
>updates
IT'S OVER, AMD IS FINISHED & BANKRUPT
AMD HAS TO RESPIN POOLARIS AND IT WILL BE DELAYED BY MONTHS
>>54651304
I think they may try to compete with 1080 and 1070 with Polaris instead of going for the mainstream market
Pls no
>>54651304
>poolaris
>>54651370
:(
>>54651304
Fuck sake cant they just release it? i guess ill just keep buying 390's and sending them back
>>54651405
>i guess ill just keep buying 390's and sending them back
??
>>54651405
Polaris is coming out mid-end of June.
>>54651418
Fuck sake i cant wait over a month.
>>54651415
Last two i had shit out on me.
>>54651405
>>54651435
which brand do you keep buying?
>>54651443
Gigabyte gaming last time it was terrible
Going to get a sapphire tri x 390 w/ backplate this time
>>54651459
>Going to get a sapphire tri x 390 w/ backplate this time
good. Sapphire makes the best AMD cards, and after then, XFX.
So AMD is basically go Sapphire/XFX or go home?
>>54651917
>>54651474
I thought MSI/Powercolor were tied for second best with xfx after them?
>>54651015
Will the new APUs be FM2?
>>54652104
Nah, check the slides, FM3
What about Asus AMD cards? Are they bad?
>>54651294
Im keeping my 980 until 1080ti/vega and hbm2.
Im still at 1080p so nothing even stresses my 980, and the 1080 is going for £650 in the UK. £650 for a mid range gpu is ridiculous when the 680/980 launched at around £400.
>>54652133
they've been lacking recently. we have to wait for benchmarks to know if the polaris ones are good
>>54651056
Nope, you can't reserve almost nowhere, and if you do, on May 27, and on June and beyond, you will get a Sold Out in the face. Wafflers are dodgy for Pascal as it has never been for Nvidia, wait and see.
>>54652119
dang, didn't want to upgrade my motherboard.
how much improvement of godavari? top-tier to top-tier
>>54651040
Still Polaris launch on 1st June is still a fucking paper launch. And the performance may not be as good as 1070.
Should I buy 980 today, or wait for 1070?
>>54654126
wait
AMD is hiding something.
Capping their demos at 60 fps, not saying anything at all, waiting for ages to release.
Nobody would be as moronic as to give NVIDIA a free ride on marketing and hype. They have something up their sleeves, trust me.
>>54651015
>sunnyvale
ayyy i hope ricky and the rest of the jackass trio are there.
randy and his fat gut, drunk lahey... shit this gon b good.
is poolaris still GCN?
is this going to be fourth generational rebrand?
>>54654956
>is this going to be fourth generational rebrand?
lel, Polaris is in 14nm FinFET, how can you rebrand a 28nm GPU to be in 14nm technological process?
>>54654126
Buy a 980ti today before people realize it's more powerful than the 1070 and the price goes back up
>>54652084
Yes, you are correct
>>54653735
>Polaris launch on 1st June
what? are you retarded? June 1st is just an update announcement, nothing else.
>>54654956
>is poolaris still GCN?
>is this going to be fourth generational rebrand?
>>54652084
>>54652133
AMD Tiers, best to worst
>Sapphire
>MSI
>Powercolor
>Gigabyte
>XFX
>ASUS
Have owned cards from all
>>54655413
>Have owned cards from all
yeah but they change from time to time.
>>54651294
I don't think anyone with a 970 or 980 has "plenty" of reasons to upgrade.
I kept a 5870 5 years, a friend of mine kept his 470 for 4, most people don't blow $400-$500 every year for a new gpu.
Maybe if you're a real enthusiast or something but that's not the vast majority of people
I have a GTX 670
Should I Polaris 10? I hate to give my shekels to Nvidia again, and don't want to buy a $600 card either.
>>54654924
Here's to hoping, I'm either going. I want to get a Nano but if Polaris comes through, I'll shell out the extra cash.
>>54655413
XFX are above Gigabyte, who shat out the only voltage-locked 300 series cards on the market.
>>54655672
>voltage-locked
what did he mean by this?
>>54651015
They're allegedly going to release some details on their Raven Ridge APUs as well.
http://phx.corporate-ir.net/External.File?item=UGFyZW50SUQ9MzM5MzMxfENoaWxkSUQ9LTF8VHlwZT0z&t=1&cb=635992964179233292
>>54654993
Nvidia can and did it with Pascal
>>54651015
I can't wait for the crash and burn.
>>54657110
WEW this is great
>>54656343
who gives a shit
>>54655529
I think Polaris may be cheaper than a Nano
>>54654956
You're a fucking idiot, just because it's named GCN doesn't mean it's a fucking rebrand
>>54657239
>responding to a literal retard who thinks 14nm card can be a rebrand of a previous gen card on different die size
>putting your trust in pajeet
I'm surprised /g/ isn't going turbo autistic over that XFX rep. Then again XFX don't make Nvidia cards (anymore) so /g/ has never heard of the company.
>>54651015
hi guys I have GTX560 almost 4 years old card
Will be the polaris at least 2x faster or should i buy 1070 instead?
>>54657721
t. pajeet
>>54657632
Dont question his almighty poo hands. He's actually pretty based and is considered the Jim Keller of Graphics Chips after working for apple
>>54657689
>I'm surprised /g/ isn't going turbo autistic over that XFX rep
What?
>>54657689
Which one was that? The guy who called out Nvidia over their founders edition bullshit?
>>54657689
u wot m8?
>>54657784
Dude posted on a comment on reddit relating to the 490x and waiting 8 days. Nothing concrete in any shape or form but this is suddenly 100% real confirmation polaris is happening in 8 days.
>>54657787
No, but i'm curious to what you are referring to specifically. I know there was some minor drama for 5 minutes over a Sapphire rep effectively saying Nvidia is scared of polaris.
>>54657820
It should have happened today, since AMD said that they were going to do a press day event for Polaris on the 20th of May back in December. I wonder why they pushed it back by over a week
>>54657847
Probably to rejig a few things given its likely Nvidia rushed out their Pascal annoucement just to beat AMD to the punch. At this point another week doesn't mean much.
>>54657847
Computex for Hardwarepartners to show custom cards
https://www.youtube.com/watch?v=Jto3bBwxAtU
>Poolaris will poo on the competition
What did he mean by this?
>>54657878
What's the word on that?
Is XFX no longer producing the same old cooler designs and reference PCBs like they've been doing since the 290X?
Is Sapphire making a Tri-X Vapor for the new cards?
Is ASUS taking the Nvidia cooler design and slapping it on Polaris because fuck it we're lazy?
Is Gigabyte putting more pointless LEDs and less voltage/power delivery for their line of cards?
Is MSI showing that they're the only competent dual-vendor AIB partner like they've always done in the past three years?
What are the chances that AMD will provide something on the level of xx60 Nvidia line for reasonable price, so I wont have to wait whole year for 1060?
>>54657899
Amd will drop the Poo hard on Nvidia next week
>>54657220
I was assuming that the 490 was going to be closer to $500. Maybe I've mistaken though.
>>54657940
Very likely, but don't think that they won't take advantage of that gap in the market by artificially limiting the supply of their Polaris cards to jack up the prices until the 1060 is rushed out. They're desperate enough for cash flows from the Radeon department.
>>54657217
zen based
>>54657935
>Is MSI showing that they're the only competent
The gaming 290x was very, very meh as far as 290x cards went. Not particularly silent, not particularly cool and not particularly highly clocked. The 390x cooler is a beast though (even with its absurd voltage offset - that card must have a 400w bios).
>>54657968
I don't think the 490 was supposed to be based on Polaris 10
>>54657935
While I can't speak for their custom designs, when I got my Fury X, XFX were the cheapest option while also giving the proper faceplate with the card, and not the shitty 'glossy plastic' thing that some others were.
Also, the XFX logo looks the best on the fan.
>>54657991
Compared to ASUS and Gigabyte, they are the most competent. Sapphire blows all of them out of the water, with XFX being the hipster choice and Powercolor being the budget choice.
>>54657820
>490x
What the fuck? Fuck off, Polaris 10 was supposed to be mainstream
I fucking swear, if AMD fucks this up I'm never buying their shit again
>>54658004
Just look at the 390X and the 290X from XFX. It's literally the same cooler designs.
And they even managed to reuse those terrible AMD reference blowers from the 290X for their 390 cards.
>>54657721
wait on polaris announcement and benches. shit will be good, but price and what it preforms like are still unknown, we only know one game but its very promising.
>>54658011
XFX made some of the smallest (if not the smallest) hawaii basrd cards out of any AIB. The cards are shit for overclocking but great for stock clocks.
>>54658028
>Just look at the 390X and the 290X from XFX. It's literally the same cooler designs.
Its not - they (slightly) redesigned the heatsink.
>>54657970
It is just that I planned to buy new PC around mid/end July and originally planned to get GTX 960, but since both Nvidia and AMD announced new lines, I got confused.
So far, Nvidia only presented 1070 and 1080 - I never buy 70s or 80s due to their very high price, always buying 60s ones. If AMD would provide something for less than 220 EUR but what would not be worse too much than 1060 (that Nvidia will probably provide much later this year), I could actually buy Radeon for the first time in many years (I only had Radeon once, after my ancient Geforce 2 card died).
My 290x is starting to reach 90 degrees I need an energy efficient poolaris NOW before my cuck shed burns down.
>>54658071
Apply new thermalpaste
>>54658071
>TFW my Fury X reached a record high of 65c while playing Ashes of the Singularity
Fuck me did I feel the heat. My card normally tops out at 55c but yesterday it hit 65c during my 2.5h rumble.
Can't believe how much of an effect it had on the room, never really understood why people moan so much until then.
Any other 380 owners here? Just wondering what a safe OC looks like.
>>54654924
>Capping their demos at 60 fps, not saying anything at all, waiting for ages to release.
I think that's because AMD doesn't have much else to show about Polaris other than "better pixels" and "it's not Nvidia".
>They have something up their sleeves, trust me.
If they did, don't you think they would have said or leaked something about Polaris immediately after the Nvidia event?
>>54658060
That looks very deliberate, more like a marketing move to drive up hype for a brand that's been in the backwaters of attention for so long.
Maybe the reason for the lack of leaks is because there aren't that many places that would want to hear leaks about AMD products these days if it's not Zen.
>>54657940
That's the sector Polaris is supposed to target so I'm going to say they have a high chance of shitting out a good mid range card
>>54658105
Or, much more likely, they didn't want Nvidia getting the upper hand by releasing specs first.
I know that sounds odd, since people would love to know specs and an architecture redesign would be well out of the bounds of reality - but forcing higher stock clocks or rehashing stuff to try and improve in other ways, even if it hurts yields, is a definite possibility.
Can't wait for AM4 DDR4 APUs
>>54658114
According to >>54658060 I think they can speculate they are going to try to pass off Polaris 10 as high-end cards, not mid-range, as they initially said
>>54658092
yea that shit can be painful over time, especially if you don't have ac on, or if you don't have a way to cool the room.
50-55c cpu
55c motherboard
80c gpu,
all hotboxing itself
need a new cpu but holding off on zen.
>>54655496
You have many if you play total war or any game that has to do with warfare simulation for that matter.
>>54658127
>but forcing higher stock clocks or rehashing stuff to try and improve in other ways, even if it hurts yields, is a definite possibility.
That's something AMD has been used to since forever. It's a well-known fact that all Maxwell silicon comes far too underclocked, whereas the AMD stuff have been pushed as far as they can go from the factory.
>>54658027
With AMD's new naming conventions 390 and 390x are mainstream products.
Vega will probably be called Fury or Rage. Do they have any other old brand names to dig up?
>>54658135
Don't get your hopes up for Bristol Ridge.
Carrizo has the same cheap 3rd party IMC that Kaveri had. It can't support memory frequencies higher than 2400mhz without adjusting base clock. The highest stable memory OC is something like 2600mhz~
So even though Bristol Ridge will use DDR4 on socket AM4, its all but useless since you can't drop in a 3400mhz CL14 kit and enjoy the increased bandwidth. It'll drop down to 2400mhz and go no higher.
Zen likely has an internally developed IMC, and that supports at least 2800mhz DIMMs, so the higher frequency stuff shouldn't be an issue. Then again Raven Ridge chips won't be on the market until some time in 2017.
>>54658158
No love for people that dont want to burn 300 EUR into GPU?
>>54658187
$400 isn't very mainstream tbqfh, 380/380x is mainstream
>>54658187
Mach
>>54658092
Ashes is a GCN killer. My monstrously overclocked 290x normally sits around 70c core and 80c vrm1 at a monstrous 1200mhz core - ashes pushes the core to over 80c and vrm1 to a peak of 105c.
>>54658187
That's fucking retarded. It was a lot easier when it was just 10 above the competing Nvidia cards, especially for first-time builders.
I still have people asking me if a 380X is the same as the 980 because of the numbering scheme.
A 490X should compete with a 1080.
A 480X should compete with a 1070.
That is a lot easier than 490 = 1070
>>54658215
Yeah, I'd definitely noticed that. Doubled up for the fact I'm playing it on DX12 and thus its enabling as much of it as the driver will allow, rather than having some bits of the card be idle.
I mean, the difference probably isn't huge, but its still making some kind of difference I'd bet.
There are two confirmed Vega chips so it will probably be something like this.
>>54658216
490 = 1070-
490X = 1070+
Vega 10 = 1080
Vega 11 = 1080ti
>>54658212
280x and 7850/7870 used to be considered mainstream and those were $300-$350 cards. 390 is filling those price brackets now.
Mainstream just means "not $500"
>>54658266
That's still fucking retarded. AMD should shoot the man who thought this numbering scheme was good (the 300 series had to be done because all the cards were rebrands).
I hope Polaris 11 is faster than a 7870. I want a small and quiet machine that sips power.
>>54658316
I can 100% promise you it will be.
VICTORY IS UPON US
>>54658316
Its probably somewhere around there.
The early silicon they showed was competing with the GTX 950 while drawing a fraction of the power. I think a Radeon 7870 is a bit faster than a stock 950, it'll probably be about on par with the Tonga R9 285.
http://www.anandtech.com/bench/product/1665?vs=1666
>>54658339
>Inb4 someone asks how I can promise such a thing
HD7950 is the absolute bottom line for VR gaming. Nvidia's 970 is the bottom recommended card for VR gaming (though the 3.5 KILLS any potential it has for VR, kek). If Polaris is to be relevant, Polaris 11 has to be cheaper than a 970 while also providing as much performance, minimum. Not just 'sometimes' but all the time.
On top of it, it'll need VR focused improvements over a 970 to make it more than just a level playing ground - but a champion over the 970.
Even then, I don't think AMD are happy to reside in the shadow on the mainstream crowd. They don't want to settle for 'okay'.
Maybe they'll let Nvidia waste all of their money on stupidly low yield incredibly expensive to develop/manufacture and high price GPU's, we have yet to see on that, but no way will they want to be a generation behind in performance for where it really counts in the market.
Oh shit niggas.
It's happening.
>>54658060
So if they are showing the cards at the Macau event and the Computex "Updates about Polaris" event is just a ruse?
>>54658433
that makes no sense. They've already demoed Polaris 10 performing better than a 390.
>>54658433
>390 levels of performance
But I've already had that for a long time
>>54655413
I have an Asus 280x it's super quiet overclocks decently good temps. I don't know why it's so low
>>54658440
I'd wager the Macau thing is to sweeten up reviewers and AIB's while computex is to announce the cards to us plebians.
>>54658454
Polaris 10 probably draws 100w~ and its only mid range.
Polaris 11 is only entry level.
>>54658478
I'm not going to upgrade just for lower watts
>>54658433
>>54658454
it'll be 390 perf for the budget polaris
Why is AMD so slow? They're always getting beaten to the punch by Nvidia.
>>54658508
So I shouldn't buy a 390 right now...?
>>54658478
Compared to the 8gb 390X drawing 295w~ average.
>>54658478
>whining about not being able to upgrade to a mid range card
AMD never billed Polaris as the top end enthusiast GPUs. Not once.
Only Fury and Vega appeal to you.
>>54658523
>buying anything right before a new generation comes out
>>54658549
The msi 390x is not your typical 390x - most 390x models have significantly lower voltage. msi merely jacked up the voltage to get 1100mhz core clocks without doing any real binning.
>>54658204
I'd imagine they'd be smart enough to tweak that you can run higher clock speed RAM with the APUs.
>>54658586
is that a bad thing?
>>54658715
it's the definition of being a retarded faggot
>>54658522
You could say they lack drive.
>>54658737
could you explain?
>>54658757
the current generation won't go down in price until the next is on the market so no matter which gen you plan on buying, it's always better to wait.
>>54658757
>buying something just before its price drops heavily
>buying something just before it gets BTFO by a new product (in terms of raw performance, power usage and price/performance ratio
>>54658715
You're buying a thing right before it becomes obsolete and you're not even waiting for it to get a discount in price for being obsolete.
How are the beta drivers for amd?
280x here
>>54658071
Yup, looking forward to the Polaris 11 here (the lower end one). Quiet and cool hopefully, coming from a gtx 650
>>54660156
I'm waiting to see what Raven Ridge has to offer.
If the IGP is comparable to the entry level Polaris and has HBM on package then I'll pick one of those up.
Somehow I thought this thread was about PolarSSL.
>>54651015
I was really looking forward to getting a Polaris since Nvidia is just going full retard with the Fuckboi Edition bullshit but this waiting is getting ridiculous. AMD had a real chance to regain marketshare and start making money again, but they're pissing it away by not having a product on the shelves. At the rate it's going, Nvidia will have their mainstream cards out soon after te Polaris launch. Hell, with the way it's looking, Nvidia might even beat AMD to the market. They'd better release some sort of information at least, or all the people who have been putting off upgrades because of the imminent node change are just going to buy the 1070.
>>54660740
>this waiting is getting ridiculous
not really though, just be patient
>>54660750
It really is. They put out Polaris teasers like half a year ago.
>only a few days from Polaris launch
>still barely any info leaked about it
It's going to be a paper launch with a 2 months delay until availability, isn't it? Fuck AMD.
>>54660809
Polaris likely isn't launching at Computex.
AMD only mentioned "updates" on the cards. Bristol Ridge is the only thing they said was launching.
>>54660832
it's IS launching VERY soon. No later than mid june
>>54660832
K 3 K
Nvidia will release the 1060 Ti based on a further cut down GP104 before Polaris 10/11 can even make it to market. Polaris will be DOA.
Goodbye AMD.
>>54660832
That's because the Polaris launch event is before Computex. Literally next week.
>>54660898
who /watchinglivestream/ here?
>>54660944
Link and time?
>>54660809
My guess is that GloFo is fucking them over yet again, by not being able to deliver on the manufacturing of the chips. If AMD had a reliable fab partner, they should be wiping the floor with Nvidia because of the fact that Nvidia has had to scrap 2+ years of R&D after being blindsided by DX12 showing that their existing arch is a steaming pile of shit. If you look closely, it's fairly obvious that Pascal (the chip used in the 1080 and 1070) is nothing more than a refined Maxwell (the chip used in the last generation GTX 9 series cards) with a couple of minor additions. And although the 1080 has managed to take the lead by brute forcing it with a clockspeed increase, what nobody mentions is that a purely die shrunk 390X would be kicking that 1080's ass.
>>54661018
Samsung will be manufacturing polaris as well as GloFo.
>>54661018
>If you look closely, it's fairly obvious that Pascal (the chip used in the 1080 and 1070) is nothing more than a refined Maxwell
Do you have any source or information to back this up? I've seen it mentioned a lot.
>>54661050
Do you have a source for that? Everything I've seen is that GloFo is doing all of the manufacturing of the chips, and that Samsung is only involved in that GloFo is using their 14nm process tech.
>>54661018
Your guess is pretty dumb.
Global Foundries fab #8 has parity with Samsung's 14nm lines. They have the exact same tooling, set up the exact same way, they constantly run test wafers on the lines to ensure that yields remain the same at any 14nm FinFET line across all 4 facilities around the world.
>>54661050
According to who? Some stupid shit rumors started by people who were to dimwitted to understand the IP licensing between the two companies?
Samsung only steps in if a GloFo customer is exceeding their volume capacity. The only instance in which this would happen is running a smartphone SoC like Apple's A9, and this deal really only benefits Samsung. Samsung gets a 4th line if they need it where as GloFo never runs any chips with volumes high enough to be capacity limited.
https://youtu.be/eXCXJoRsgJc
>>54661073
I confused zen and polaris. Samsung will be manufacturing zen alongside glofo . not sure about polaris
>>54661096
Stop regurgitating wccftech bullshit as if its factual.
>>54661060
I don't think that Nvidia has actually confirmed it, and frankly, they'd be idiotic to do so, but it's a fairly safe logical deduction when you look at the die layouts and compare the specs.
>>54660960
http://ir.amd.com/phoenix.zhtml?c=74093&p=irol-newsArticle&ID=2170028
>>54661106
>>54661092
https://www.techpowerup.com/218578/samsung-to-fab-amd-zen-and-arctic-islands-on-its-14-nm-finfet-node
http://wccftech.com/samsung-14nm-lpp-finfet-amd-x86-zen/
http://www.engadget.com/2015/12/22/samsung-building-chips-for-amd/
Register here for ComputeX Live Stream:
http://experience.amd.com/NotifyMe/computex-live-stream/
>>54661120
>literally posting a wccftech rumor
Good job, you subhuman retard.
Electronic Times reports on erroneous forum rumors constantly. They are not a reliable source of information, and AMD themselves have never said a single thing about having parts run at Samsung. Or anywhere for that matter, they have no reason to discuss internal business decisions with shit eating children like you.
But hey, if you're fucking stupid enough to believe that I can show you an article claiming Kaveri will have 4 way SMT and 12 CU on die.
>>54661092
GloFo has somehow managed to bork every single AMD product launch since they were spun off, so why in the world would you be inclined to suddenly trust them to be reliable?
>>54661120
>an unnamed aquaintance of a random forum poster disclosed that a pigeon whispered in his ear that such and such is happening.
>>54661191
Because they aren't using internally developed process IP, thats why.
Their issues have always been related to R&D. Not developing competitive processes, not getting them ramped up to volume readiness on time. Thats an R&D issue.
They dropped their own internal 14XM, and licensed Samsung's 14nm FinFET process. In addition to this they've heavily been licensing IP from STMicro, and they now own IBM's foundry business including a ton of staff and their whole portfolio of IP.
Global Foundries will operate issue free for the next 3-4 years until they start having to scrape a new node together on their own. They already have a functional 7nm process thanks to IBM. All of the grunt work for 10nm and beyond is finished.
>>54651294
>970 and 980 owners have plenty of reason to upgrade with Pascal
if you think sound minded 970 and 980 want to "upgrade" to nvidia you're deluded. i'm already aware that pascal numbers will be inflated in upcoming games because they run gameworks 2.0 features (VXGI, VXAO) better than maxwell can. they did this to kepler based cards with heavy tesselations in games which made maxwell cards always have the lead in performance.
you're literally a brain dead consumerist if you're still buying nvidia in 2016.
this is the time to strike
>>54661257
>If you turn up the settings the game runs more slowly.
How you jump from that to buy AMD just blows my mind. AMD sucks lol.
>>54658206
I was led to believe that there was going to be a $300 gpu
>>54658478
>mid range
Ah, the good old bait and switch
>>54658463
run bf4 with mantle and 1440p max settings and tell me if you don't get artifacts
>>54660740
>Nvidia will have their mainstream cards out soon after te Polaris launch
They won't, or only in very limited quantities.
Polaris has been in mass production for half a year already.
Would be absolutely amazing if Polaris10 comes with GDDR5X at a lower price point than the 1070 just to spite Novidya
>Alleged 2304sp P10:
175w TDP
1.45Ghz stock (1600Mhz boost clock?)
2304/2048 = 1.125 (to normalize for shader/cluster counts)
P10 is allegedly 25-40% faster per-core than what it's comparison may be.
1.125*1.25*1.45 = 2.04
*1.6 = 2.25
*1.35*1.45 = 2.2 <<< 35% IPC compared to Tahiti/Pitcairn?
*1.35*1.6 = 2.43
*1.81 = 2.75 <<< 25% overclock (1800Mhz), 200w power target
So a P10 appears to be at least 2.0-2.25x as capable as full Tonga
>3.973 TFLOP/s * 2.04 = 7.946
>3.973 * 1.125*1.25*1.45 = 8.1
>Tonga compared to a 14nm 290X:
190w TDP > 175w
1.0Ghz > 1.45Ghz
2048/2816 = 0.727
1.0/1.45 = .69
Let's assume shader efficiency is the same
.727*.69 = 0.5
Thus,
Tonga is about half as capable as a 290X
Finally,
P10 (reference) is either roughly equivalent to a 290x, or capable of being 30% faster, in raw flops alone
> 2816sp * 2 * 1 = 5.63 <<< stock 290x
> 2304sp * 2 * 1.45 = 6.68 TFLOP/s <<< 18.6% more FLOPs
> 2304 * 2 * 1.6 = 7.37 TFLOP/s <<< 30% more FLOPs
> P10 @ 1.85Ghz
> 8.56 TFLOP. FuryX = 8.6 TFLOP
These are all very rough numbers because we simply don't have any facts. The variations between card comparisons also take a slight bit of leeway (graphics rendering IPC used to normalize "Effective TFLOPs", unknown rendering IPC improvements between different arches, unknown model used to compare P10)
>>54651264
soooo they will not be buying an amd card then
>>54665915
What happened to GDDR6?
Couldnt give a shit about HMB2 until its ready anyway also its expensive as all fuck atm so i doubbt mid tier cards will get it.
>>54661018
>die shrunk 390X would be kicking that 1080's ass.
Add GDDR5X or GDDR6 and it definitely would.
i just ordered a 390x for the first time feels good
>>54661018
>it's fairly obvious that Pascal (the chip used in the 1080 and 1070) is nothing more than a refined Maxwell
How is this any problem though? Maxwell was amazingly efficient, delivering high performance at low power draw and with much less cores than GCN uses.
Why don't you bitch about AMD still being stuck on the same shitty architecture they have been on for 4 years?
What the fuck is this I've been reading? Polaris 10 is supposed to replace R9 390/X? Hasn't it been confirmed to be in the same performance level? So where's the fucking upgrade? In performance/watt? Fuck off
>>54668098
There is no such thing as GDDR6. There never was.
>>54666699
Look at the performance of Polaris 11 and how little power it consumed. The TDP for Polaris 10 is nowhere near 175w.
>>54661018
>a purely die shrunk 390X would be kicking that 1080's ass.
If it was running at 1.8GHz and in DX12 only, you mean, right?
>>54651267
>no hbm
doubt it
>>54666699
>P10 @ 1.85Ghz
Implying it can clock to 1.85 Ghz, it can't
>>54668185
There is nothing to suggest AMD can top their best card after 3 years now so i doubt it.
That will be Vega mate.
>>54668219
as opposed to the 1080 thermal throttling on stock clocks and getting rekt by the 980ti OC
>>54668227
>There is nothing to suggest AMD can top their best card after 3 years now so i doubt it.
Agreed it doesn't look like they can
>>54668225
>Implying it can clock to 1.85 Ghz, it can't
and why not?
If the nvidia meme can do 16nm 2.2ghz+ on 200watt then surely the poolaris can too
>>54668239
pretty hard to top the fury X but devs are lazy assholes and dont take advantage of GCN at all unless its games like mantle or qauntum break
Where the fuck is the OpenCL acceleration? i was promised this shit almost 10 years ago with wangblows vista
>>54668258
The architecture may not allow it
>>54668227
>as opposed to the 1080 thermal throttling on stock clocks
That's really irrelevant and only due to the stock cooler and its default fan profile. It's irrelevant because by the time we will be able to buy Polaris there will be custom 1080 models out that will easily solve the issue. It can be solved even on reference models by simply setting a more aggressive fan curve, at the expense of noise.
You should've bitched about the 216W power limit, that's actually going to require custom boards to get around most likely.
>>54668258
Pascal is a small change from Maxwell, and Maxwell was designed to clock high. No GCN arch is. The throughput on the SIMD lanes is far too high to sustain the same clocks.
Polaris 11 was demoed with a clock around 850mhz, 10 will probably top out around 1050mhz like the rest of the GCN family stock.
>>54657632
He is going to turn Nvidia HQ nto DESIGNATED.
>>54668070
Point is that AMD is targeting an untapped market who would like a low budget card that has decent performance. A 390X level performance for 250 is how it'll happen. They're also super low power (the only reason I'm considering it) which will tap the laptop/portables market. Strategy is increasing sales by a significant amount, not competing with nvidia for discrete GPUs, if that makes sense. In fact, they don't have anything to compete until vega or navi, and this is part of the strategy imo.
>>54668456
>A 390X level performance for 250 is how it'll happen.
Keep dreaming buddy
>>54651015
>Polaris update
dont you fucking do this to me AMD I've waited long enough to update from my gtx 670
>>54668476
on one side it is bullshit to expect that
on the other,i would totally upgrade with 250bucks 390x perfomance card and call amd god
>>54651213
>Then what are they competing for, last years performance?
That's the same thing Nvidia is going for, with the 1080 being on par with a 980TI OC.
And it runs at 80C to boot.
>>54668476
This is something the 1060 might even offer.
It's honestly not that hard to imagine with a die shrink
>1080: Titan X + performance
>1070:980ti stock performance
>1060:980/390x performance
>>54668506
Except Nvidia offers last years peformance of its big chip on their mainstream chip, while AMD will be offering last years mainstream performance on their mainstreamchip.
>>54668098
AMD already has a ton of experience with HMB2. Expect their navi and vega cards to curbstomp the competition.
>>54668359
>Polaris 11 was demoed with a clock around 850mhz, 10 will probably top out around 1050mhz like the rest of the GCN family stock.
Fuck sake
i mean i know amd cards are stronger clock for clock but still thats really shit
>>54668553
I WOULD IF THEY WOULD HURRY THE FUCK UP
>>54668523
>Except Nvidia offers last years peformance of its big chip on their mainstream chip,
At twice the price point.
>while AMD will be offering last years mainstream performance on their mainstreamchip.
At half the power usage and midrange price.
>>54668578
>At twice the price point.
>980ti is 550$
>foundrs edition 1080 is 700$
>you and math
>>54668292
>missing the point
From a theoretical 2x improvement the 1080 only got 20%. I'm also not so sure how pascal is an improvement besides the higher clocks (by virtue of the die shrink/temps).
AMD has had a complete architectural overhaul. By the time it matures nvidia will be dead in the water because they would just at that time release a properly new architecture that will have quite a few bugs to sort.
>>54668508
More like
>1080: 980TI OC performance, twice the price
>1070: 980TI stock performance, costs $100 more
>1060: between 960/970 level
Of course give it a few months and they'll end up with twice the performance, once they release drivers that gimp the 9xx cards.
>>54668603
>AMD has had a complete architectural overhaul.
Where did you get that? I thought GCN4 was just a tweaked Tonga.
>>54668600
Tell that to the aussies paying $1000 for it.
>>54668636
That still isn't twice the price point, and Ausies also pay 1000$ bucks for a 980ti
>>54668614
>1080: 980TI OC performance, twice the price
Holy fuck are AMD people really this bad at math?
>>54668636
thats just the importers ripping us off.
its 'only' $599 usd but we're getting cucked.
i think alot of ausfags will just go onto newegg and tell aussie retailers to get fucked.
>>54668603
>From a theoretical 2x improvement
Who the fuck came up with that theory? Some retard who doesn't understand how business works?
>>54668631
Tonga is the name of a GPU, not an architecture.
4th generation GCN is radically different from 3rd gen.
New front end, tessellation, command processor.
New scalar unit inside the CU
New memory controller and cache structure
New UVD/VCE
>>54668689
we'll see
i guess my 390x will have to last until i rma it after i get it and it will shit itself because amd
>>54668689
Doesn't sound radically different at all, definately not as big a step as maxwell was to keppler
>>54668662
You are delusional if you think it'll be possible to get a 1080 for less than $1100 - they are doing the founders edition for a reason, it's a paper launch. There'll be fuck all retail availability; and what little there is will be bought in bulk by resellers and auctioned at ebay for twice the price.
>>54668721
>changing most of the architecture doesn't sound radically different
Can you do us all a favor and just go die somewhere?
>>54668728
>You are delusional if you think it'll be possible to get a 1080 for less than $1100
You people are fucking insane, do you honestly think this is true?
It seems you guys are adding 100$ to the price every day I am away from /g/
>>54668744
It just looks like new iterations of the same thing, thus GCN.1.4
Explain what is so radically new about it?
>>54669240
It will be sold at $900 in the UK and $1000 in the US. Keep crying that nvidia lied about LITERALLY everything at the unveiling.
>>54669258
s/US/Aus
>>54669258
Why would I cry about anything, I'm not even interested in the 1080, I'm going for a used 980ti.
The only one whimpering is you
>>54669240
>Explain what is so radically new about it?
did you not see the picture? are you having brain problems? almost the entire design is new, faster, and more efficient
>>54669344
The picture literally says nothing, that could just be pure marketing. They just put 'new' everywhere.
It's the same fucking shit Nvidia does in their slides where they try and make everything look magnificent.
Tell, me what is exactly new about this architecture and not just an iteration of old GCN they have been using for years?
>>54651281
this desu senpai
>>54658135
>not watching Gokaiger
I hope you are at least watching them in order
>>54658316
>faster than a 7870
>which probably will be on par with 1060
Is your requirement a little bit too low?
And seriously if Polaris has nothing to do with high-end market, and then we have Vega for the high-end, then when will Vega come out?
[spoiler]Or can I just buy a Polaris 10 to replace my GTX760 and call it a day? But I am still using Phenom II X6 and I'm not sure whether I go Skylake or Zen in the second half of the year.[/spoiler]
>>54670448
i7-6700k owner here. I kinda of regret it. I mean with Zen coming out soon and such with its 8 core 16thread for cheap (relative to intel's $1000 asking price), and my desire to have the PC last for 5 years, I think it was a bad purchase. But I had a shitty c2d so whatever. If you can, wait for zen.
>>54661095
there a game list for this yet?
>>54668258
nah, furyx was hamstrung by 28nm, as it was designed for 20, 28nm introduced to many bottlenecks that couldn't be overcome, but can be side stepped with async.
>>54671048
Even hawaii makes more sense on 20nm than 28nm - that would've certainly curbed in the rampant power draw given the rather high clocks AMD gave it. Fiji is just a mess though as it has 45% more shaders than a 290x but tends to only be in the region of 25% faster. Even async compute implementations (notably ashes) don't let it pull ahead like it should. Best theory is its ROP starved.
>>54671108
if i remember right, its just the bandwidth inside the card itself was not made for 28, it was made for 20
a good portion of the shaders have fuck all to do because the card simply isn't fast enough.
async side steps it a bit and can pull more, but its still fucking starved.
there is a reason amd cut ties and paid the fine, missing the 20nm node fucked em hard, meanwhile nvidia made a the 900's a 28nm from the get go because they anticipated the failings of their fab.
>>54670831
Will zen actually run at a decent clock speed? I know that's not the only important thing but muh games.
>>54671249
Summit Ridge is not a highly clocked high serial throughput chip meant to compete with 4ghz Skylake and newer Kaby Lake.
It'll probably be clocked around 3ghz with a few aggressive boost states. Its meant to compete against intels Haswell-E and Broadwell-E.
>>54671249
I think it should be decent enough. Coupled with a good GPU it should last you a while if your job depends on it (video editing/rendering/3D/simulation/etc.). Otherwise a 6700k is just fine.
>>54671667
>>54671464
Hm, that's a hard decision then. I was thinking of upgrading to a 5820k since there's no real point in waiting for broadwell-e at that price point.
>>54651158
There are cards and processors that are tested in overclocking potential and you can buy the best of those for much more than normal. Sure it's not useful for random gamers, but for people who just want the absolute best, this binning is nice. Nvidia is probably doing that here. One might argue, that buying the cheaper variant is stupid, because it basically guarantees a shitty chip. Inb4 shill
>>54671923
I seriously doubt Nvidia is actively binning the chips for us mere plebians. Now I can see them doing some aggressive binning for that monster setup they showcased but Nvidia temps to cheap out on their reference (sorry, founder's) pcb for consumer grade cards.
In a twist of fate AMD's pcb for hawaii/fiji is an aboslute tank and built to handle far more than the bios actually allows (the vrms are good for well over 100c and can handle an absurd amount of current).
>>54671923
>he thinks Nvidia is actually binning chips for a reference GPU
I really hope you're joking
>>54669240
I'm not going to predict a shortage but its possible. 6700k Skylake was hugely inflated over msrp for months.
>>54651281
>what is AMDGPU
>>54657721
560Ti here, waiting for Vega.
>>54672758
>560ti
Are you homeless?
>>54668456
With my r9 290 Sapphire Tri-X would it be worth it to upgrade to their $150 option + a zen cpu?
>>54674495
Given the rule of thumb for gpus is upgrades are only worth it when offering 50-100% extra performance a 290 has no real upgrade path (especially if you overclock) until vega/big pascal based upon where polaris is likely to sit (who knows we might all be surprised).
Zen is a complete unknown at this point and is facing far stiffer competition. Until we know what sort of secondary features it supports (such as pcie lanes) nothing concrete can be suggested.
>>54673945
yeah man consumerism is great isnt it
>>54674766
I think I am being bottleneck by this 980x i7
If I overclock either CPU/GPU pretty much always crashes.
>>54674880
>If I overclock either CPU/GPU pretty much always crashes.
Assuming you aren't being retarded in your overclocking chances are your psu can't feed the increased power draw.
I'm not familiar with the nuances of overclocking that cpu, but moar voltage and an aggressive llc (with good cooling) never goes too far wrong. As for your 290 - set a custom fan curve so it ramps up to 70% at 50c and 100% at 75c will giuve you the thermal headroom to crank the volts and get the core around 1150mhz. If it needs staggering amounts of voltage ditch afterburner and use trixx. Be warned though - trixx allows for +200mv offset by default and cooling that takes series effort (pic related, a 290x at nearly 1.4v with aftermarket air cooling).
>>54674766
Yeah, I only upgrade at 150% extra performance, as in 250% of my current fps
>>54652147
>980
>midrange
when will this meme end?
>>54675305
>398 mm2
>not midrange
kek, nvidiots love drinking the koolaid.
>>54675305
>680
>cut down die sold as high end
>780 cut down die sold as high end
>titan
>cut down die sold as ultra premium
>titan black/ 780ti
>fully unlocked chip sold with price to match
Conversely
>7970
>fully unlocked chip with price to match
>290x
>fully unlocked chip with price to match
>fury x
>fully unlocked chip with price to match
