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

[andybird123]

https://www.techpowerup.com/mobile/reviews/AMD/R9_Nano/31.html

It's not arguing for arguing's sake when what you are saying is patently false. The Nano is a 200w card, not 150, and even it just barely matches the 980ti on perf/watt by being a lot slower and massively underclocked. The FuryX uses a lot more power than a 980ti to match it on performance. And before you go down the "over clocked 980tI uses moar"
https://www.techpowerup.com/mobile/reviews/MSI/GTX_980_Ti_Lightning/21.html
https://www.techpowerup.com/mobile/reviews/MSI/GTX_980_Ti_Lightning/24.html

As for the 380/390 issue, as I said; someone specifically mentioned the 380 as being a direct architecture comparison to a 960, I replied to that

 

[SimBy]

consumer price or cost to produce ^ Hawaii and Fiji are needing much larger die sizes to compete on performance, Maxwell is an incredibly efficient design by comparison

at 40-60% larger die sizes they very well should be beating the "equivalent" Maxwell cards

40-60% bigger die size? How did you come up with that? 390X is like 10% bigger than 980.

 

[andybird123]

40-60% bigger die size? How did you come up with that? 390X is like 10% bigger than 980.

the person I replied to was talking about the 960 and 380X
960 is 227mm2, the 380/X is 366mm2

the bigger maxwell's are similar in size to their comparison, but use a lot less power to get the same performance level
the 980 vs 390X in particular the 980 uses half the power in some full gaming situations whilst getting similar frame rates- to be fair there is a memory amount difference there to account for, but not double the power use

I haven't seen power comparisons for DX12, that would be interesting, but I can't see them closing that big a differential, and you would think that a GPU being more fully utilised would actually use more power rather than less

 

[Harlequin]

http://www.anandtech.com/show/9390/the-amd-radeon-r9-fury-x-review/25

20w difference under gaming load....

its common and frequently repeated that Maxwell is massively more efficient than AMD - it simply isn't

 

[egraphixstudios]

Can't wait for pascal!

 

[Mtom]

http://www.anandtech.com/show/9390/the-amd-radeon-r9-fury-x-review/25

20w difference under gaming load....

its common and frequently repeated that Maxwell is massively more efficient than AMD - it simply isn't

Here comes the Maxwell confusion. If we talk about power consumption, the ref Maxwells cited. When it comes to performance, the custom ones cited and praised with their great OC potential.

"Oh a Fury X can catch a ref 980Ti in this game? Come on everyone uses the custom ones, and ran them at 1500MHz haha...oh power consumption...here look at this chart about a ref card"

 

[andybird123]

http://www.anandtech.com/show/9390/the-amd-radeon-r9-fury-x-review/25

20w difference under gaming load....

its common and frequently repeated that Maxwell is massively more efficient than AMD - it simply isn't

Here comes the Maxwell confusion. If we talk about power consumption, the ref Maxwells cited. When it comes to performance, the custom ones cited and praised with their great OC potential.

"Oh a Fury X can catch a ref 980Ti in this game? Come on everyone uses the custom ones, and ran them at 1500MHz haha...oh power consumption...here look at this chart about a ref card"

I actually posted 1450mhz power consumption above

if you look at the anandtech charts, they use Crysis 3 as the power usage example - Anand use full system power and don't try to isolate the GPU, they also use "at the wall" power usage which again will affect the results in a different way

even so, they do use the same method for each card so lets go ahead and use that as an example

980ti = 388W at 40.9fps = 9.49
FuryX = 408W at 37.9 = 10.77

13.5% more efficient but that is for the whole system, which will skew the result, if you take off 100W for the mobo/CPU it already climbs to 16%, and looking at the techpowerup power draw results you should probably take off a bit more than that, and you are also comparing a HBM card with a GDDR card which again will be skewing the GPU Core power usage result away from a straight Fiji vs. Maxwell comparison

Techpower up have the "overclocked" 980ti going up by 31w for a power increase of 14.5%, but in the same game used by Anand, an fps increase of 16%...

to come up with a true perf/watt result you would have to run the power draw test for each and every game tested individually, but no one does that

 

[Harlequin]

its not `skewing` anything - they use the same measurement - from the wall for both cards in the same system - its an apples to apples comparison.

 

[andybird123]

Do you know how percentages work?
Or performance per watt?

They both rely on knowing what the card itself draws, not the whole system

 

[Harlequin]

Do you know how percentages work?
Or performance per watt?

They both rely on knowing what the card itself draws, not the whole system

do you understand what apples to apples comparison means? since the base line system is the same , the results are comparative - you do understand what that means?

here ill link it for you so you can increase your own , personal knowledge

https://en.wikipedia.org/wiki/Comparative

 

[muziqaz]

do you understand what apples to apples comparison means? since the base line system is the same , the results are comparative - you do understand what that means?

here ill link it for you so you can increase your own , personal knowledge

https://en.wikipedia.org/wiki/Comparative

I don't know why feed him for this useless OT conversation? Leave him alone, as he is beyond the help Ignore button does wonders, you know

 

[andybird123]

do you understand what apples to apples comparison means? since the base line system is the same , the results are comparative - you do understand what that means?

here ill link it for you so you can increase your own , personal knowledge

https://en.wikipedia.org/wiki/Comparative

Yes I understand what comparative means, however the conversation I replied to and which you joined was both architectural and perf/watt based, which your comparative does not add to

You're moving the goal posts to suit a narrative, but not the subject that was actually being discussed

 

[humbug]

Good Grief.... lol

 

[eyetrip]

Haven't been around here for a looooong time.

Good to see nothings changed

 

[bru]

Ok so Maxwell can be 5-20% more efficient Per/W depending on how it is measured, can we move on now.
Both Polaris and Pascal will put the currant architectures to shame per/w wise I expect.

 

[JediFragger]

Haven't been around here for a looooong time.

Good to see nothings changed

No **** dude!!!! Hope those Ti's are still running well

 

[muziqaz]

<...>
Both Polaris and Pascal will put the currant architectures to shame per/w wise I expect.

Exactly, it won't be even funny, my guess is.

 

[eyetrip]

No **** dude!!!! Hope those Ti's are still running well

I sold one, as doing so little gaming, it just felt like it was going to waste.

All I really manage currently is a bit of BF4 on occasion

 

[JediFragger]

I sold one, as doing so little gaming, it just felt like it was going to waste.

All I really manage currently is a bit of BF4 on occasion

Yeah, I sold one in anticipation of the new cards

But tbh I do a lot less gaming when the weather is better anyhow, so I'm gonna try and hold on for the next Ti/Titan

And tbf a single Ti prtobably handles BF4 fine @ 4k

 

[Boomstick777]

Nvidia Pascal GTX 1080 Features 8GB GDDR5X & 384GB/s Of Bandwidth, GTX 1070 Features 8GB GDDR5 & 256GB/s

Read more: http://wccftech.com/rumor-nvidia-pascal-gtx-1080-gddr5x-gtx-1070-f-gddr5/#ixzz45XvjVvRR