** The Official Nvidia GeForce 'Pascal' Thread - for general gossip and discussions **

[Baboonanza]

You are quite correct Ferrari and I shouldn't be rude when replying (stress with waiting for my Oculus Rift to be delivered) but I am basing my posting on what Tom Peterson said in the PCPer interview. Now I am not an engineer and take that there are many ways to deal with asynchronous shading like Tom has implied and this is something I have also kept a close eye on and also had my reservations about how Pascal would cope.

https://youtu.be/xtely2GDxhU?t=2h12m55s

You will forgive me if I take what Tom said as literal but truth be told, it will all come out in the wash over time.

Edit:

And I am also doing what anyone who is buying new hardware does and not looking at possible negatives

I just watched that bit (starts at 2:13:13), and he's being a little misleading I think. When he says 'Pascal has the hardware for asyncronous compute' Nvidia hasn't yet shown anything that would indicate this is true. I'm not calling him a liar, but I think are they calling some new features like dynamic-load balancing and pixel pre-emption asyncronous computing when that is definitely open to debate. I suspect AMD would strongly disagree.

This quote from WCCFTech sums it up best:

Dynamic load balancing and improved pre-emption both improve the performance of async compute code considerably on Pascal compared to Maxwell. Although principally this is not exactly the same as Asynchronous Shading or Computing. Because Pascal still can’t execute async code concurrently without pre-emption. This is quite different from AMD’s GCN architecture which has Asynchronous Compute engines that enable the execution of multiple kernels concurrently without pre-emption.

AMD has long touted the asynchronous compute capabilities of its GCN graphics architecture. The company built what it calls ACEs, Asynchronous Compute Engines, into its hardware. It’s available in all of AMD’s GCN architecture based graphics cards, including the now more than four year old HD 7970.
What Nvidia is doing with preemption and dynamic load balancing right now, while not exactly async compute, can be used to accomplish similar goals.

I do agree with the point he makes in the video though:

It doesn't matter

What Nvidia have done may be enough to get most of the performance gains from concurrent processing and if the performance is there overall arguing about technical details is somewhat pointless.

 

[Greebo]

Totally agree with you Baboonanza. It doesnt matter how it works or if it works at all.

Which is fastest? Clearly its still Nvidia.

Going forward next gen though, it will start and make a diference I suspect and Nvidia will have to change course.

 

[Rroff]

I suspect Nvidia has just gone brute force but the point is its still 10% faster than the furyx in these directx12 games and then you might gain another 20% from overclocking, does it matter that Nvidia has gone brute force? Faster is still faster no matter how you do it.

Its like somebody coming up with a new car engine totally different to the old one. It might be the better and neater design but if the old v8 with supercharger and twin turbos is still 20% faster it doesnt matter how good the new design is.

For definite proof of nVidia's async/dx12 performance we need to see beyond those 2 titles really - aots implements an almost synthetic load which doesn't necessarily represent typical game use and hitman doesn't really run well on anything.

 

[Gregster]

I just watched that bit (starts at 2:13:13), and he's being a little misleading I think. When he says 'Pascal has the hardware for asyncronous compute' Nvidia hasn't yet shown anything that would indicate this is true. I'm not calling him a liar, but I think are they calling some new features like dynamic-load balancing and pixel pre-emption asyncronous computing when that is definitely open to debate. I suspect AMD would strongly disagree.

This quote from WCCFTech sums it up best:


I do agree with the point he makes in the video though:

What Nvidia have done may be enough to get most of the performance gains from concurrent processing and if the performance is there overall arguing about technical details is somewhat pointless.

Yer, he is on a subject that clearly isn't his specialty (he is more to do with G-Sync panels etc) and does skip about a bit and truth be told, until it is implemented into games like Hitman and AoTS, we are all a little second guessing. There is no magic switch and each card has to be coded in (pretty much the same as Mantle) to the DX12 games, so I don't expect to see it right away either.

Meh, plenty of brute force there anyways

 

[IvanDobskey]

Prices should be more like:

1080 £500
1070 £300

But you can add £100+ to both for gouging purposes.

And to be honest, even £500 is too much when you consider the 980 launch price.

 

[GFX-Kid]

You are quite correct Ferrari and I shouldn't be rude when replying

Indeed. Especially when you yourself go around telling people not to be rude on a regular

(stress with waiting for my Oculus Rift to be delivered)

Ah, first world problems eh?

And I am also doing what anyone who is buying new hardware does and not looking at possible negatives

I thought you was above that kind of thing. I always thought you was the objective type, what with swinging both way's and all that

Not sure what is going on with you lately. Recently you are on record for saying GDDR5X is almost as fast as HBM, then you said nvidia spent seven billion in r&d on pascal. We chalked them down to old age instead of green shades. But now I am left wondering...

Sorry if it seems like I am giving you a tough time there fella, it's just that a lot of people here look up to what you say and enjoy your reviews. But all this is not helping

 

[mahius]

Funny to see the price argument (discussion) has now spread to the 1070. Keep expectations low, so you either won't be disapointed or you'll be pleasantly suprised. Expect much higher than what the 970 is currently at, especially considering the 970 has had competition for a while and we currently don't know whether the 1070 will have any products competing with it from AMD.

Come on Gibbo, give us a hint!

He kinda did, he mentioned a bit about exchange rates (which triggered myself to then ramble on about the economy going off-topic). Methinks it's a clue that the UK will get shafted for pricing. Alas, not a though.

The retailers like OCUK don't get too much say over the pricing. If Nvidia supplies them with FE 1080s £600 a piece, then selling them at any less than £600 would be a loss to the retailer and it would do more damage actually stocking and selling them at that price. The retailers must make at least some profit to pay the wages of the folks who work there and pay for the site/warehouse. And it allows them to occasionally do crazy deals where they might sell items at a loss. But don't expect it anytime soon for a 1080, profit margins are 'usually' better on cheaper products like the 970. Other businesses need to make a profit, not just Nvidia lol.

 

[Freddy1765]

Async is disabled for all nVidia GPUs in Ashes. Check this forum post:
https://forum.beyond3d.com/threads/...nd-analysis-thread.57188/page-64#post-1915300

 

[fs123]

I suspect Nvidia has just gone brute force but the point is its still 10% faster than the furyx in these directx12 games and then you might gain another 20% from overclocking, does it matter that Nvidia has gone brute force? Faster is still faster no matter how you do it.

Its like somebody coming up with a new car engine totally different to the old one. It might be the better and neater design but if the old v8 with supercharger and twin turbos is still 20% faster it doesnt matter how good the new design is.

My point was that the 1080 is no faster in DX12 than DX11. It goes from 53.7fps to 53.3fps in AoTS and from 47.7fps to 47 fps in Hitman which means it actually drops a bit in DX12 just like the Maxwell cards.
The clockspeed pushes it past the other cards but the alleged addition of Async does not do anything for it unlike the FuryX which appears to show improvements over DX11. AoTS doesn't have Async enabled for the 1080 and maybe thats true for Hitman too.
The FuryX gains 1 fps in DX12 while the 1080 loses almost 1 fps.

 

[markoboyo11]

^ I wouldn't worrie yerself too much about ocuk's profit margins, I'm sure they have a large % to play with for these gpu's aswell.

Early tax is a payday for everyone involved.

 

[Ayahuasca]

Quite right, there's a reason Gibbo gets so excited around new GPU launches, particularly nVidia ones.

 

[Gerard]

Well a FE card costs $699 which equates to about £478. Rounding it off to £500 would probably be the way to go but i really doubt that will happen, probably £550 plus.

 

[Bazzards]

Its all about performance I couldn't care less how it is achieved.

I am not even convinced that Async will a: be used that often and if it is it will only be on a select few titles and b: will be fairly irrelevant to the performance achieved by brute forcing on Nvidia hardware. Therefore a purchasing decision using Async as a major positive seems pointless to me.

 

[Legion]

Well a FE card costs $699 which equates to about £478. Rounding it off to £500 would probably be the way to go but i really doubt that will happen, probably £550 plus.

£600+ for early adopters, inc price gouging.

 

[Freddy1765]

Well a FE card costs $699 which equates to about £478. Rounding it off to £500 would probably be the way to go but i really doubt that will happen, probably £550 plus.

Pretty sure MSRP does not factor VAT into it.
The prices you see on American e-tailers like Newegg are before any kind of tax.
MSRP * (1+VAT) * (some factor taking exchange rates into account) is a more likely approximation.

 

[bishi]

Sorry if this has already been answered but does anyone know how long until AIB cards will be released?

 

[Greebo]

Its all about performance I couldn't care less how it is achieved.

I am not even convinced that Async will a: be used that often and if it is it will only be on a select few titles and b: will be fairly irrelevant to the performance achieved by brute forcing on Nvidia hardware. Therefore a purchasing decision using Async as a major positive seems pointless to me.

Agreed. Maybe next generation it will be a factor worth consiering but to me I just want to know which is faster no matter how they do it.

Yes, its great AMD cards are 16% faster in dx12 over dx11 but thats still not enough for them to overtake the 980ti sometimes and certainly not the 1080.

Just comes down to whether you can afford a 1080 or prepared to pay that much.

 

[Greebo]

Sorry if this has already been answered but does anyone know how long until AIB cards will be released?

same day as founders card, the 27th May.

Doesn't mean to say that every variation the in AIB lineout will be out that day. I'm sure I have read EVGA are planning up to 6 versions of the 1080 but I doubt all 6 will be out on the 27th.

 

[harney]

ppl better get used to £400+ for 1070 and £600+ for 1080
take $ price convert to £, add 20% VAT, then add 20-30£ shipping and about 10-20£ for some other crap
699$ > 479£ + 20% > £575 + shipping & stuff = ~ £620 1080 founder edition
449$ > 308£ + 20% > £369 + shipping & stuff = ~ £410 1070 founder edition

for custom £379 1070, and £549 1080, these are MSRP prices that you wont see but during special deals, the actual price will be £30-£50 more than this 90% of the time.

I think is worse than this ...
from what i have seen in that nvidia ask questions vid and read ups on line...the founders to me is looking more like a buffer card for the partners to come in at without looking too high ....
there is no way a custom board is going to be cheaper than the so called ref card...i can only see the cheap arse plastic blowers being cheaper

Will see

i hope for everybody who are buying that its going to be fair but i am afraid does not look good thus far

 

[harney]

I'm getting my 980 ti back.

I'm putting it under water and overclocking it and waiting for 1080ti.

Are you being serious or joking ??? as this was my plan all along how dare u i should have copy righted it

got a 980 g1 ti cheap getting 1491 core just shy of 8000 mem with +24 with stock bios

with custom bios not finished testing as of yet i am upto 1545core mem is same but at max volts testing more good temps so far below 68 69c fan curved getting 69 to 72% fan speed

but i am sticking with air ..