http://videocardz.com/59845/aida64-gets-vega-10-polaris-1011-and-pascal-gp104-support
Get on the vega hype train!
>lulz why aren't u waiting for Vega? it's 0.2% more powerful!
NO
FUCK
OFF
I'm done being a waitcuck, buying a 1080 as soon as it's available.
>>54500028
good goy
>>54500028
>hbm2
>large die
>high clocks
>new arch
vega is going to btfo everything
>>54499251
the hype is real
It's going to be underwhelming and have disappointing temperatures, disappointing power consumption, and slow driver releases that don't fix problems without creating other ones, forcing you to wait for months until the actual bugfixes.
>>54500028
I mean you'll get a good card. But vega is on a smaller lithography. It could very well smash the 1080 if they price it right. The whole point of being a waitfag is to get the best stuff at the right time. If you get a 1080 on release day, you might as well have bought a 980ti months ago.
>>54502259
>But vega is on a smaller lithography
They are the same, they just chose to name it 14mn for marketing purposes and for a reason, idiots like you fall for it, the finfet technology both AMD and Nvidia are going to use is the same the only difference is the process which is where thing are going to make a difference, you can cut chipsets as small as you want as long as you are willingly to compromise.
The only company that holds true 14mn process is Intel.
>>54502423
14nm is smaller than 16nm
>>54502437
well fell for the marketing, idiot.
>>54502448
go look at apple's dies, 14nm is 7% smaller
thats even a premature 14nm process
>>54502448
14/16FF use the same 20nm BEOL production but their true feature size is much, much different.
In this comparison we see that Intel leads the pack (in absolute size), Samsung's 14LPE (not LP+ which is used for AMD's GPUs) is somewhat larger, and TSMC's 16FF+ is the same size as their 20nm production.
Nvidia is producing Pascal on the 20nm node it was originally designed for.
AMD is using a slightly inferior to Intel 14nm node.
>>54502448
What's it like getting told? >>54502615
>>54500028
>buying the 1080 when the 1070 brings out almost the same performance at a fraction of the cost
It's the 970 all over again. You fucking idiots never learn. Either be a cuck with the *70 or wait and get the enthusiast card that isn't the Titan.
Samsung nor TSMC neither Globalfoundries can do true 14mn, they have a Frankenstein mix mess and they choose the 14mn brand name to trick people, only Intel which has their own favs can do the only true 14mn process on the world.
Intel 14n sram cells: 0.0588 um2 versus 16nm TSMC 0.07 um2 versus Samsung/Global 0.064um2
TSMC's 16FF has a 64nm minimum metal pitch, 90nm gate pitch. Samsung 14nm has a 78nm gate pitch, 64nm minimum metal. Intel 14nm is at 70nm gate pitch, 52nm minimum metal.
Intel 14nm is the leader, followed by Samsung/Global 14nm, and then TSMC 16nm, but TSMC 16nm should have better scalability, in the end both TSMC and Samsung/GlobalFoundry will have to compete with chipset quality which comes down to manufacturing, tsmc 16nm its better than samsung 14nm on the Iphone 6S by almost 20%.
>>54502909
Samsung's 14LPE is not their 14LPP (also called 14LP+)
LPE, used in the iPhone SoC, is for minimum power draw on small, low power devices.
LPP/LP+ is for larger designs meant to draw more power.
We won't have any knowledge of Samsung's high power capability until products are in people's hands.
>>54502712
2 mm won't make amd win you fanboi
>>54502936
You might as well add TSMC 16FF+ to the mxi fi you're going that way.
>>54502615
They're both 20nm. There's so "slightly worse 14nm".
It'd be hilarious if they released Vega on October. I'd love to buy it and finally get rid of this 970. Polaris though, when can we expect more information? There are so many different rumors going around and I don't know where they're heading. Bringing Fury X/980Ti performance at $300 would be insane, but I highly doubt it'll happen.
Welp.
http://hexus.net/tech/news/graphics/92747-amd-vega-launch-brought-forward-october-says-report/
>>54502827
I'm hoping this wo't happen. The 1070 might be my enhanced survival option until 1080Ti otherwise I'm forced to use my 780Ti. Some games already don't work @4K, working means 60 FPS, even on lowest details. I won't be a retard this year and save the money for a 2nd 1080Ti later.
>>54502936
>LPE, used in the iPhone SoC, is for minimum power draw on small, low power devices.
>LPP/LP+ is for larger designs meant to draw more power.
LPE is Low-Power Early and was just an early version of the process.
Low-Power Plus is a tweaked/improved version of the process with better power efficiency and clockspeed, both by about 15% compared to LPE.
http://www.anandtech.com/show/9959/samsung-announces-14lpp-mass-production
>>54502940
15%better
Lpp is also denser
>>54503111
I imagine the rumors of $300 980 level performance were real for DX11 but DX12 will be even better than the 980 by a few percent probably. I don't think well see 980ti levels of performance though really at least not in DX11.
>>54505724
980ti performance is well within reach.
Both Polaris and Vega should be spectacular.
It's a shame the nVbros will have to wait for Volta for real DX12 performance, they'll still have the DX11 crown though...
>>54500308
yeah, till the next titan, which will be released like in 2 or 3 months after VEGA
>>54507161
Munted the spreadsheet up a big, fixed.
>>54507161
>>54507206
you guys have source for that spreadsheet?
>>54503111
The only person saying that rumor is you.
>>54507215
It's all fudged figures from Videocardz 1080 benchmarks and AMD's slides.
It's in no way official, just speculation based on some available data.
AMD stated 40% increase in IPC for Polaris (this assumes no extra increase for Vega over Polaris) in a few marketing slides
Polaris10 fudgy figures are:
(FijiPro Score/3584/1000)*1600*2304
>(FijiPro Score/FijiPro Cores/FijiPro Clockrate)*Speculated Polaris Clockrate*Leaked Polaris10Pro Core Count
>>54507179
A Titan will also cost 2x as much