PASCAL BTFO
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
![1458841346361[1].png](https://i.imgur.com/9AFAUhV.png "1458841346361[1].png")
>www.bitsandchips.it/52-english-news/6785-rumor-pascal-in-trouble-with-asyncronous-compute-code
>Pascal will be an improved version of Maxwell, especially about FP64 performances, but not about Asyncronous Compute performances
>2016
>no async compute
>nvidiots will still buy it
>>53670077
Pascal and Polaris are going to be shit now, fuck I should have bought a 390
>>53670077
Looks like nVidia is too late to change their plans. Might as well wait for Volta instead of investing in Maxwell 3.0
Welp, looks like my 290 will last me another 2-3 years.
>New Nvidia cards not looking as good as people thought
>Same goes for AMD
And here I am happy as can be with my 390x.
>>53670077
Time to buy the Titan/980/970/960/950/Fury/390/380/370 that you've been dreaming about. Nothing special coming up.
>>53670206
All of those cards are lame tho.
Looks like I'll be hanging onto my $220* GTX 970 4GB**
>>53670339
You're so obnoxious.
And a party pooper.
Whats the big deal with async anyway. it's only useful in task where you don't mix the same pipelines. Mixing tasks where bandwidth is in use will actually be slower if using async, it really isn't the parallelism people make it out to be.
It is good but it really isn't that great.
>>53670532
it's what makes amd cards amazing at dx12 and nvidia's cards shit.
>>53670566
Mixing task in AMD card still carries a heavy penalty though.
>>53670339
Everything's lame, anon. Welcome to modern desktops, enjoy your stay.
>>53670077
>pascal seems to be a maxwell refresh
At least we don't need to wait anymore
>>53670109
wait for pascal/polaris and enjoy cheap 390's.
>>53670077
>especially about FP64
>mfw literally nobody in the world gives a single shit about FP64 on a gayming GPU
The only scientific applications that use gaming GPUs do it because of the FP32 performance because they don't care about the FP64 performance. How can nvidia be this retarded?
Here hoping Pascal will introduce native high-sample coverage masks synchronization and 64 coverage samples per pixel, hardware based obviously.
>>53670109
>>53670151
Polaris looks amazing actually. 14nm will mean double performance.
>>53671031
Expect for the fact that the latest leaks indicate that they are just going to focus on reducing power draw instead of increasing performance.
It's fucking shit m8.
>>53671057
It's unthinkable that a die shrink this big isn't intrinsically going to improve performance significantly.
>>53670109
>tfw I didn't buy a 390 when it was $220 on black friday
>>53671081
My interest is in will they still be using 2 and 4 gb vram as standard? 4K is starting to push that.
>>53671145
If it's GDDR5X it should be more. As I understand, 2/4GB is some kind of limit with HBM1, and since these cards are a new series and use GDDR5X, there's hope.
I'd settle for 6GB, but 4GB is cuts it too close in my opinion.
>>53671180
The Sapphire Nitro 390 being my go to if the the next xx70's are 4gb.
>>53671206
i have it, bought it as soon as it was available. great card
>>53670077
>www.bitsandchips.it
>rumor
>according to our sources
I should have bought a Nano.
I should get a job
>>53670077
>Novidya doesn't even have Async compute for Pascal
>Pascal is literally just a die shrink Maxwell
>AMD won't even put HBM1 on either of their Polaris GPUs
>Polaris will have a memory buswidth of 128-bit and 256-bit
Why is this generation going to be such shit?
>>53672284
you forgot
>Nvidia still jewing out on video memory with 7.5GB on at least one of their cards
>>53672284
The 128&256bit cards will POSSIBLY both be the lower-medium tier cards.
It would make sense because they've only very recently released Fiji and now announced Fiji pro - they don't want to walk all over it or damage its place in their ecosystem. Not just yet.
Also from a competition standpoint, it makes sense not to release your upper end cards straight away because that just gives the opposing company plenty of time to look at what you've got and try to make their offering better if they can.
Yes I know it'd obviously be far too late for any architectural change for the chip itself, but things like extra VRAM or only deciding to use the higher binned chips to get a better clock to improve performance that way (obviously at a cost of profitability) is also a concern.
Later in the year; possibly with Vega, we'll see the true high end cards along with the Fury successor.
>>53672284
AMD's excuse is that both the new cards will be low to mid range and will probably both have more cache and better texture compression.
Nvidia just wants to control the market with gameworks and not have to innovate in their hardware.
Still no async shaders? Gimp works to the rescue!
>>53670077
I bet 1$ that Pascal will BTFO Polaris, mark my words.
>>53673343
you should just burn the $1 and be done with it at this point, pascal looks like shit
>>53670077
AMD have certified dx 12 cards and use async compute as "feature" in gcn
Nvidia have certified dx 12.1 cards without async compute
Intel have certified dx 12.1 cards (more tiers levels than nvidia) but not have async compute
Then why AMD memers insist in Async compute is mandatory?
>>53670077
So when do the real cards come out, the flag ships?
>>53670077
Indians won once again
>>53673684
Async compute is not mandatory,
Exists because AMD incompetence with yours shaders cores
>>53673684
because amd saw the potential in it.
it's the same with HSA which vastly improved efficiency and performance, yet intel won't join on it just now because it requires further development
>mfw my 770 4gb is chugging along happily on dx11 Windows 10
Oh well waiting until Volta 2018 who else?
>>53670077
>nvidiots on suicide watch
>>53673752
not surposed with OP's news
>>53673752
>Oh well waiting until Volta 2018
maybe.
My 380 4gb will do untill then at any rate.
best be better then lol
>>53673739
That Intel dominance though.
>>53671081
>It's unthinkable that a die shrink this big isn't intrinsically going to improve performance significantly.
Why? It's not like any other processor has significantly improved performance just by shrinking the die.
If they want performance increases then they actually have to redesign the architecture.
>>53670077
>gow
no async
>RotTR
doesn't anything
>Hitman
no async
>BUT MY AotS!
This will be the only game to ebenefit from async, other games won't use it, or won't benefit from it that much. TW:Warhammer is the only title that needs async and I doubt it will use it.
no one cares about your AMD hyperthreading
The only thing I am looking forward to in this generation is price drops.
>>53671206
I got it a few months ago and its a fucking beast. So glad I didn't get caught up in "waiting for the new architecture" meme.
lol I read the url as Brits and Chips
you know like Fesh and Chips :)
Fake news posted by shitty Italian site to slander
>>53670987
>>53671031
>Polaris
>>53661407
>>53676412
>and enjoy cheap 390's
But polaris is looking significantly better than pascal right now.
>pascal
>literally just a name change
>polaris
>at least performance per watt is doubled and temperature is halved.
>>53676447
We know literally nothing about Pascal.
Not even wccftech which post anything even the more shady rumors knows something.
>>53676872
www.bitsandchips.it/52-english-news/6785-rumor-pascal-in-trouble-with-asyncronous-compute-code
