http://wccftech.com/nvidia-gtx-1080-async-compute-detailed/
>1080 has full support for Async Compute
It's incredible, Nvidia has completely demolished AMD.
what about polaris
>10 frames is demolished
In what world?
>>338114706
Now if AMD released their card tomorrow and demolishes this one by 10 frames what would you say...
>>338114975
who cares?
What the fuck is Async
>>338113864
>full Async Compute
nope.
>not even 2 fps increase
>>338115358
>>full Async Compute
>nope.
Source
>tfw no GPU for 2 months now
>still awhile until 3rd party cards get released
>still nothing from AMD
>starting to wither internally
Wut do?
>>338115308
I would also like to know
>>338113864
>At 2560 x1440
I think I'm fine playing on ultra in 1080p, no need to burn money in this thing yet
Wow it beat a card that came out last year by <10 frames. Put AMDfags on suicide watch
>>338115308
>>338116173
>>338116023
buy something cheap and used for the time being
my 7770 was a nice holdover when i was saving up for my gtx 770 and only cost me $80 canadian
>>338116592
Thank you very much
So THIS is the game everyone is using as benchmark nowadays, and it struggles to look like this @ 60fps?
>>338115646
It's NVidia themselves. Pre-emption is not true async.
>>338117064
It's an unoptimized piece of shit and the only reason it's getting attention is DX12, otherwise it would be relegated to the vast shovelware pile on Steam
No one even plays it, it's simply a benchmarking tool for early DX12 performance, which so far is shit on both AMD and Nvidia
>>338117064
muh dx12
>people thinking async is the only DX12 feature that matters
>>338115646
Async Compute != Fine Grained Preemption
The dynamic load balancing is not the same either. Nvidia is just making it easier to do a context switch without the harsh penalty of having to completely reset the state of the SMs.
If it really supported Async, you would be able to run concurrent graphics and compute threads on the same SM WITHOUT context switching.
If you have no idea what I just said, you should just go back to enjoying games and not worrying about this bullshit (or go actually learn without spreading FUD). We are talking ~+10% differences here.
>>338113864
>$700 for the non-ti card
>demolished
>is just a superclocked maxwell
>>338117519
>paying an extra $100 for a shitty reference design that has always been inferior to board partner cards
I was going to get a 1080 to replace my 980ti because it suffocates in my mATX case,
then I saw that the 1080 hits 87c.
L O L.
>>338117225
>>338117239
Every day I'm more convinced that new APIs only allows developers to be lazier than ever, instead of improving something.
I'm yet to see a DX12 game and think "wow, this looks better than the past years games".
>>338113864
Look, /v/. I want to get a new card. I don't give a shit about which company it's from. I just need something. So, bullshit aside:
Is this all /g/ level faggotry or does the 1080 actually worth it?
When's AMD releasing new cards? Is it even worth waiting for based on what's going around?
>>338118463
Well the reference cards do I imagine. Maybe we'll see 86c once zotac tackles that shit.
>>338118463
on what? all games?
>>338118659
reference cards are better for smaller cases and sli because they exhaust the hot air out the back of the case
>>338118654
>Is this all /g/ level faggotry or does the 1080 actually worth it?
It's worth it if you want the fastest card on the market, but if you want a cheaper version the 1070 launches in June
Don't get the overpriced founders edition, wait for the board partner cards for cheaper
>When's AMD releasing new cards?
Polaris at Computex in a few weeks, Vega in 2017
>Is it even worth waiting for based on what's going around?
Polaris is aimed at the mainstream market whereas Vega is aimed at the performance segment, so it doesn't look like AMD will have anything to compete with Nvidia's higher end this year, but they'll have more affordable cards that still perform well and use less power
We'll have to wait for benchmarks to see how they stack up
>>338118659
The 980ti ref would hit 83-85c, with AIB cards hitting around 78-79c on default fan curves.
If the 1080 follows this, we're still going to see 80c+ on default curves, and 72-75c with aggressive ear killing curves.
I might invest in a hybrid cooler 1080, but at this point with the power limits on the 1080, and the fact that hybrids run off reference PCB it really seems unattractive. Even most waterblocks are for reference PCBs, so with this power limit it's quite shitty whichever way you put it.
>>338118617
>new APIs only allows developers to be lazier than ever
It's the opposite.
